UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The geography of Canada - United States trade Andresen, Martin Alexander 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2006-199312.pdf [ 20.82MB ]
Metadata
JSON: 831-1.0092867.json
JSON-LD: 831-1.0092867-ld.json
RDF/XML (Pretty): 831-1.0092867-rdf.xml
RDF/JSON: 831-1.0092867-rdf.json
Turtle: 831-1.0092867-turtle.txt
N-Triples: 831-1.0092867-rdf-ntriples.txt
Original Record: 831-1.0092867-source.json
Full Text
831-1.0092867-fulltext.txt
Citation
831-1.0092867.ris

Full Text

T H E G E O G R A P H Y OF C A N A D A - U N I T E D S T A T E S T R A D E by M A R T I N A L E X A N D E R A N D R E S E N B A . , Simon Fraser University, 2000 M A . , Simon Fraser University, 2001 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T OF T H E R E Q U I R E M E N T S F O R T H E D E G R E E OF D O C T O R OF P H I L O S O P H Y in T H E F A C U L T Y OF G R A D U A T E STUDIES (Geography) T H E U N I V E R S I T Y OF BRITISH C O L U M B I A August 2006 © Martin Alexander Andresen, 2006 11 ABSTRACT My dissertation is an empirical analysis of the geography of international trade that is primarily focused on the interregional trade flows within Canada and the United States, but begins with an analysis of international trade flows at the international level to place Canada - United States interregional trade in a global context. Throughout my dissertation empirical methods are not only employed, but extended and developed to address issues with past research. At the international level, I find that globalization defined as an integrated global economy is not present. Rather, international trade flows are intensely regionally focused and the intensity of that focus has not decreased over time. In fact, distance and proximity have become more important in determining trading regions since 1981. At the national level, I find that there was substantial change occurring within the Canada - United States trading relationship prior to the establishment of the Canada -United States Free Trade Agreement (CUFTA). Prior to the establishment of the CUFTA, Canada appeared to be moving into the lower-end of product quality trade relative to the United States. However, since the establishment of the CUFTA that pattern has reversed. Now, Canada appears to be moving into higher-end product quality trade with the United States. This is altering the spatial division of quality-based production in North America. And at the regional level, Canada - United States interregional trade flows have altered their spatial configuration since the establishment of free trade. Prior to the establishment of the free trade agreements, the Canadian provinces were overwhelmingly focused on interprovincial trade. However, with the removal of trade barriers to the U.S. states, Canadian provinces are now trading significantly more with the southern neighbours. The levels of interprovincial trade remain high, but the shares of provincial trade to and from the United States are now increasing. T A B L E OF CONTENTS Abstract ii Table of Contents , iii List of Tables v List of Figures ..viii Acknowledgements ix Dedication x CHAPTER 1 Introduction 1 1.1. Introduction , 1 1.2. Why does international trade (and its geography) matter? 2 1.3. Geography and trade 5 1.4. Spatial unit of analysis: the sub-national region 8 1.5. My general research questions and chapter outlines 14 CHAPTER 2 In search of the global triad: does trade have its own geography? 18 2.1. Introduction ; 18 2.2. Past research on the regionalization of the world economy 23 2.3. Data, measurement, and methodology ..32 2.4. Empirical results 39 2.5. Summary and conclusions 51 CHAPTER 3 A history of Canada - United States trade relations 54 3.1. Introduction 54 3.2. Early attempts at trade liberalization 54 3.3. The Auto Pact 58 3.4. Free trade in North America 64 3.5. Conclusion 82 CHAPTER 4 A national analysis of Canada - United States international trade, 1979-2003 84 4.1. Introduction .....84 4.2. Data and methodology 87 4.3. The intervening years 94 4.4. Canada - United States trade patterns, 1989 - 2003 102 4.5. Summary and conclusions .113 iv C H A P T E R 5 Canada - United States integration: the effects of the national border and the N A F T A 116 5.1. Introduction 116 5.2. The border effect 120 5.3. Data and estimation methodology 133 5.4. Border effect results 138 5.5. N A F T A effect results 146 5.6. The direction of future research on Canadian trade 154 5.7. Conclusions 156 C H A P T E R 6 Canada - United States interregional trade flows, 1989 - 2001 159 6.1. Introduction 159 6.2. Past research on Canada - U.S. Interregional trade 161 6.3. Data and methodology 165 6.4. Canada - U.S. Interregional trade flows, 1989 - 2001 167 6.5. Conclusion 183 C H A P T E R 7 Testing for change in the spatial pattern and identifying trading regions for Canada - United States interregional trade 187 7.1. Introduction 187 7.2. Quasi-points and spatial point pattern tests 190 7.3. The spatial (quasi-) point pattern test and its index 194 7.4. Results : 197 7.5. Reciprocal trading regions in Canada and the United States 203 7.6. Conclusion 214 C H A P T E R 8 Conclusion 216 8.1. Introduction 216 8.2. Directions for further research 222 8.3. The conclusion of the conclusions 225 Bibliography 349 Appendix The measurement of international trade: sensitivity to trade overlap and product quality 383 V LIST OF TABLES 1.1. U.S. Regions and their International Trade Destinations, 1987 226 2.1. Share of World Exports and Imports 227 2.2. GDP, Export, and Import Shares 228 2.3. Percentages of Bilateral Trading Relationships Relating to the Global Triad 229 2.4. Summary of Trading Relationships, 1.30 Threshold 230 2.5. Regional Assignments, 1981, 1.30 Threshold 231 2.6. Regional Assignments, 1991, 1.30 Threshold 232 2.7. Regional Assignments, 2001, 1.30 Threshold 233 2.8. Regional Assignments, 1981, 1.20 Threshold 234 2.9. Regional Assignments, 1991, 1.20 Threshold 235 2.10. Regional Assignments, 2001, 1.20 Threshold 236 2.11. Regional Assignments, 1981, 1.10 Threshold 237 2.12. Regional Assignments, 1991, 1.10 Threshold 238 2.13. Regional Assignments, 2001, 1.10 Threshold ... 239 2.14. Summary of Trading Relationships, 1.20 and 1.10 Thresholds 240 4.1. Canadian International Trade Flows, Billions of Constant 1997 Canadian Dollars, 1980- 1988 241 4.2. Canadian International Trade Shares 242 4.3. Total and Industrial Sector Export Levels, Billions of Constant 1997 Canadian Dollars 243 4.4. Canada-U.S. International Trade Flows, Increase Factors, 1979 - 1988 246 4.5. Industrial Sector International Export Trade Shares 247 vi 4.6. Industrial Sectors, by Trade Type 250 4.7. Canadian and U.S. Average Tariff Rates, Estimated A d Valorem Equivalent 255 4.8. Canadian International Trade Flows, Billions of Constant 1997 Canadian Dollars 257 4.9. Canadian International Trade Shares 258 4.10. Total and Industrial Sector Export Levels, Billions of Constant 1997 Canadian Dollars :".;.„ ; ...259 4.11. Canada-U.S. International Trade Flows, Increase Factors, 1989-2003 262 4.12. Industrial Sector International Export Trade Shares 263 4.13. Industrial Sectors, by Trade Type 266 5.1. Standard Gravity Equation Variables 271 5.2. Border Effects 272 5.3. Border Effects, Sensitivity Analysis 273 5.4. Provincial Border Effects 274 5.5. Canada - United States Trade 276 5.6. Canada - United States Trade, Provincial N A F T A Effects 277 5.7. Canada - United States and Interprovincial Trade 278 5.8. Canada - United States and Interprovincial Trade, Provincial N A F T A Effects 279 5.9. Factors Behind the N A F T A Effects 280 5.10. Provincial Trade Flows, Billions of Constant 1997 Dollars, and Increase Factors 281 5.11. Shares of Provincial Trade Flows 284 6.1. Spatial Units of Analysis 286 6.2. Provincial Exports, Millions of Constant 1997 Dollars, and Increase Factors 287 6.3. Provincial Imports, Millions of Constant 1997 Dollars, and Increase Factors 293 vii 6.4. Provincial Export Shares 299 6.5. Provincial Import Shares 303 6.6. Export Share of Provincial G D P 307 6.7. Changing Interregional Patterns 311 6.8. Proximity and Economic Sizes Regression Output 312 6.9. Increase Factors for Ontario's Automotive Trade 313 7.1. Spatial Interaction Data and Quasi-Points ..314 7.2. Indices of Similarity, Exports and Imports 315 7.3. Effective Provincial Tariff Rates ...316 7.4. Provincial Trade Shares and Export - G D P Ratios with the United States .317 7.5. Trade Location Quotients, Exports, 1989 and 2001 318 7.6. Trade Location Quotients, Imports, 1989 and 2001 322 7.7. Provincial Reciprocal Trading Regions 326 A . l . Trade Overlap and Product Quality Thresholds 393 A.2 . Descriptives -394 A.3 . Model Selection 395 A.4 . Sensitivity Analysis 396 LIST OF FIGURES 1.1. Global International Trade Flows and Output 328 1.2. Regional Trade Agreements: Annual and Total 329 2.1. Exports and Imports Trade Location Quotients, 2001 330 2.2. Exports and Imports Trade Location Quotients, 1991 331 2.3. Exports and Imports Trade Location Quotients, 1981 332 6.1. Canada, the United States, and the EU-15 333 7.1. Indices of Similarity, Exports and Imports 334 7.2. Trade Location Quotients 335 7.3. Reciprocal Trading Regions in Canada and the United States, 1.30 Threshold 346 7.4. Reciprocal Trading Regions in Canada and the United States, 1.20 Threshold 347 7.5. Reciprocal Trading Regions in Canada and the United States, 1.10 Threshold 348 ACKNOWLEDGEMENTS I would like to take this opportunity to thank my committee: Professor Trevor J. Barnes, Professor David W. Edgington, Professor Michael D. Buzzelli, and Professor Stephen T. Easton. Their help, encouragement and expertise that has proven invaluable for the completion of my dissertation. I would also like to thank Katrina for her unwavering support and patience during the completion of my education that took longer than expected. DEDICATION For my wife and son, Katrina and Matthew. 1 CHAPTER 1 Introduction 1.1. INTRODUCTION This dissertation examines the economic geographical effects of international trade flows between Canada and the United States over the past fifteen years. More specifically, the purpose is to investigate the sub-national (regional) effects of the establishment of free trade between Canada and the United States, 1989 - 2003. I wil l argue that only through a geographical approach can Canada's international trading relationship with the United States be properly understood. Despite the sweeping nature of current international trading agreements, the trading relationship between Canada and the United States began modestly and pre-dates Canada's confederation in the 19th century. It only began to intensify from the 1960s forward. This intensification began formally in 1965 with a trading arrangement involving trade in automotive products (The Canada - United States Automotive Products Agreement of 1965, hereafter known as the Auto Pact). In 1988, the Canada -United States Free Trade Agreement (CUFTA) was signed, and five years later the North American Free Trade Agreement ( N A F T A ) , essentially adding Mexico to C U F T A . These trading agreements all had provisions for reducing barriers to trade (both tariff and non-tariff barriers), increasing cross-border investment, and more recently, enhancing the temporary cross-border flow of skilled labour. 2 As a result of the Auto Pact, and reinforced by these later free trade agreements, Canada and the United States are each other's top international trading partner for both exports and imports, with automotive products trade dominating trade flows on both sides of the border. While it is true that the trading relationship between Canada and the United States occurs at a national level, the argument of this dissertation is that the national scale of analysis shrouds a more geographically differentiated (regional) nature. Canada, as well as the United States, has a geography of industrial production. And because of this geography of production, trading patterns of sub-national regions within each country differ (different sub-national regions trade different products and services), translating into a geography of trade. These two geographies led the Canada - United States' free trade agreements to have regionally differentiated consequences. Tariff and non-tariff barriers, for example, are lowered at the national level but do not impact equally for all sub-national regions within each country. Consequently, there are geographical differences implicitly embedded within free trade agreements. The purpose of my dissertation is to present a sub-national (regional) analysis of Canada - United States trade, 1989 - 2003. It is only at the sub-national scale, my dissertation argues, that the full impact of free trade agreements between Canada and the United States trade can be appropriately assessed. 1.2. W H Y DOES I N T E R N A T I O N A L T R A D E (AND ITS G E O G R A P H Y ) M A T T E R ? In 2004 total global merchandise exports were valued at over US$9.1 trillion, with more than half of that international trade flowing between developed countries (such as Canada 3 and the United States), less than 15 percent between developing countries, with the remainder flowing between developed and developing countries (Helpman 1999; International Monetary Fund 2005a, 2005b). As shown in Figure 1.1, the average growth rate of international goods and services exports is more than twice the average growth rate of gross domestic product from 1985 to 2001. Additionally, services1 occupy an increasing share of total output for most industrialized nations (particularly in North America and Europe) and have even a faster export growth than merchandise trade flows (Dicken 2003). Perhaps most important, international trade flows are positively associated with economic growth (Frankel and Romer 1999; Noguer and Siscurt 2005). For these reasons international trade flows are central to national economies. <See Figure 1.1, page 328> The rapid growth in international trade flows is largely regarded a result of successive rounds of negotiations of the General Agreement on Tariffs and Trade (GATT, established in 1947), now known as the World Trade Organization (Helpman 1999). Central to the G A T T was the principle of non-discrimination in trade relations among the G A T T members, also known as the principle of multilateralism-—that is global free trade or the move toward it (Bhagwati 1999). Trading blocs or zones, usually placed in opposition to multilateralism (protectionism is truly the opposite of multilateralism), represent the move toward free trade within a limited set of countries in 1 In the context of international trade, services include telecommunications, financial services, management, and advertising (Dicken 2003). 2 There are, of course, potential downsides resulting from countries being too open, such as vulnerability to international market crises. •4 the world, but maintaining tariffs for all remaining countries not in the trading bloc. However, trading blocs are better viewed as limited forms of free trade: a form of multilateralism that does not include the entire global economy. Trading blocs are not all the same. We can recognize five general types, what is order of complexity they comprise: preferential trading agreements, free trade areas, customs unions, a common market, and an economic union—which includes a common currency for the region (OECD 1993). The first, preferential trading agreements, are a mild form of integration. They involve lower trade barriers to member countries, which may or may not include free trade for certain goods. Second, a free trade area, however, entails the elimination of barriers, tariff and non-tariff, to trade between member countries, with or without the requirement that there be a common barrier to all non-member countries. These agreements may include investment liberalization and harmonization of technical standards, but do not usually include the free movement of productive factors, particularly labour. Third, a customs union extends the free trade area so that all member countries agree to a common tariff scheme against non-member countries. Fourth, a common market extends a customs union to include the free mobility of factors. And finally, an economic union extends the common market to include the harmonization of economic policies among member countries, which may include a common currency (OECD 1993). One should not consider these categories as rigid however, but as points along a continuum that measures degrees of regional economic integration. In 1985, there were 26 supra-national Regional Trade Agreements (RTAs) in 3 The Asia-Pacific Economic Cooperation (APEC) is another form of regionalism that relies upon non-binding commitments to facilitate economic growth, international trade, and investment (APEC 2006). 5 force in the global economy. As shown in Figure 1.2, that number rose to 189 by the end of 2003, with 100 of the RTAs established during the past seven years.4 <See Figure 1.2, page 329> The thesis of this dissertation is that international trade flows in the global economy and the simultaneous rise of regional trade blocs demonstrate the importance of geography. Patterns of international trade flows are highly related to spatial proximity (contiguity) and distance (Frankel et al. 1995; Poon 1997). Such features are also found in RTAs . By definition, RTAs have a high degree of intra-RTA trade flows with some of the more well-known examples including the European Union, M E R C O S U R (South America), the Association of Southeast Asian Nations ( A S E A N ) Free Trade Area, and with more relevance to this dissertation, first the Canada - United States Free Trade Agreement (CUFTA) and later the North American Free Trade Agreement ( N A F T A ) . This high degree of intra-RTA trade flows heightens the economic importance of RTAs . Indeed, intra-RTA trade growth is typically higher than the average growth for international trade as a whole (Frankel 1997). 1.3. G E O G R A P H Y A N D T R A D E The relationship between RTAs and trade growth has spawned a plethora of research by both geographers and economists. Economists typically investigate the non-spatial determinants of intra-RTA trade flows asking what drives these trade flows (see Frankel 1997, 1998), whereas geographers typically investigate the geographical dimensions 4 Over 250 RTAs have been notified to the GATT/WTO as of December 2002. This number was expected to approach 300 by the end of 2005 (WTO 2006). 6 asking where that trade flows (see Dicken 2003, Michalak and Gibb 1997, and Poon et al. 2000). Curiously, most of the attention geographers have paid to international trade flows has been at the nation-to-nation scale (see Dicken 2003; Gaile and Grant 1989; Hanick 1987, 1988, 1989; Michalak and Gibb 1997; Nierop and De Vos 1988; and Poon et al. 2000), despite the fact that sub-national regional,production systems are at the heart of economic geography (Walker 2000).5 In this light, the geographer Tony Hoare (1993) notes that " a little-explored facet of the global economy is the way regions within nations and different parts of the international community interact through trade flows. ... Given the well-established tendency for any one country to trade more with some overseas nations than with others we should expect at least as much and probably more trading-partner specialization on the part of that country's constituent localities" (Hoare 1993: 701). However, even studies concerned with sub-national regions (see Bauer and Eberts 1990; Hoare 1985; Smith 1990) tend to examine trade flows of those regions with the "rest of the world" rather than with other sub-national regions (Hoare 1993: 702). In short, within the geographical trade literature, the sub-national region is a relatively unexplored research area. A n exception is the research of Erickson and Hayward (1991) that examines the destinations of United States' regional export flows. Through their recognition that the United States is an "economy organized over [a] large...geographical area...stretching from one side of the continent to the other, [and that] it is probable that substantial differences in export destination patterns do exist," their regional analysis of export flows shows that regional differences in trade patterns are strong (Erickson and Hayward 1991: 372). 5 It should be noted that most international trade data is measured at the national level only. 7 <See Table 1.1, page 226> A s shown in Table 1.1, after aggregating the U.S. states into nine regions defined by the United States' Bureau of Economic Analysis, Erickson and Hay ward (1991) find that spatial proximity to markets is a critical factor in determining the spatial pattern in export destinations: U.S. region to country international trade flows. For example, the United States' Western region (comprising Washington, Oregon, California, and Nevada) trades more than average with Asia, the Great Lakes region (comprising Wisconsin, Illinois, Indiana, Ohio, and Michigan) trades more than average with Canada, the New England region (comprising Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, and Vermont) trades more than average with Europe, and the Southwest region (comprising Arizona, New Mexico, Oklahoma, and Texas) trades more than average with Mexico (Central America). This geographical pattern should not be surprising given the force of proximity and distance. Erickson and Hayward's (1991) results, of course, pose the question: why is the nation the unit of analysis in most North American studies of international trade flows, and not the region? 6 Canada, too, undoubtedly has great variation in the destinations of its provincial exports, but it is most often treated as a single spatial unit of analysis. More recently, economists have begun to pay attention to the sub-national regional scale when investigating international trade flows between Canada and the United States, particularly in the study of changes in international trade flows due to the N A F T A . Most of the research done on the effects of the N A F T A are conducted at the 6 They pose this question despite the fact that they only disaggregate the U.S. states and not their trading partners. This lack of disaggregation, however, is likely a limitation of their data. 8 national level and conclude that the N A F T A has had an insignificant effect on the national aggregates of Canada - United States international trade flows: This conclusion is the result of a lack of significant differences between the C U F T A and the N A F T A (Cox and Harris 1992; Gould 1998; Krueger 2000). However, allowing the effect of the N A F T A to vary regionally, Wall (2003) finds that there is a strong geographical component to the effects of the N A F T A . Similarly, Coughlin and Wall (2003), by treating Canada is a single spatial unit, find geographical differentiation in the effect of the N A F T A on U.S. states. What seems to be happening is that a national scale of analysis washes out the regional effects of the Canada - United States free trade agreements, leading to false conclusions—positive and negative effects of the N A F T A that are present at the sub-national level are averaged out when the analysis is at the national scale. As a result, there is a compelling need for a comprehensive two-way sub-national study of the geography of Canada-United States trade flows in order to reveal the regional effects of trade agreements. For Canada - United States' international trade flows, the sub-national region is the place where change manifests itself and, hence, it is only at the sub-national (regional) scale that the full impact of free trade agreements between Canada and the United States trade can be appropriately assessed. 1.4. SPATIAL UNIT OF ANALYSIS: THE SUB-NATIONAL REGION Since the early 1980s, there has been substantial interest in the role played by the sub-national region in the process of economic growth and development—the resurgence of regional economies (Storper 1997). The sub-national region has always been considered important, but increasingly it is seen as central to the economic performance of the nation 9 as a whole. However, these dynamic sub-national regions are not omnipresent. They are found in particular places, and never permanent. There are several different literatures that discuss the importance of regional economies: economics, economic geography, and social-political policy. Each is discussed in turn. Despite the recent focus on sub-national regions and international trade flows showing strong regional effects of trade policy, regional economies are an old idea in the study of international trade in economics, though seldom now discussed. Over 65 years ago, Bertil Ohlin stated that "[s]pace is important in economic life for two chief reasons: the factors of production are to some extent confined to certain localities and move only with difficulties; and costs of transport and other impediments prevent free movement of commodities" (Ohlin 1967: 5). Ohlin goes further to state that the region, the spatial unit of analysis in his trade theory, must be a congruent spatial unit such that any differences within the region must be smaller than the differences that exist between regions. In this context, the analysis of interregional trade is only "helpful in a study of trade between nations...[when those nations]...fulfill...the conditions established for regions" (Ohlin 1967: 49). In these cases, international trade is a specific, or special, case of interregional trade. The important implication for my dissertation is that the nation is not the natural spatial unit of analysis for international trade flows, even at the abstract level used in the economic theory of international trade. The second literature operates at a more concrete level of analysis that has its intellectual beginnings with the work of Alfred Marshall, a nineteenth century economist, who investigated concentrations of economic activity that were specialized in particular industries—industrial districts (Amin 2000). These districts typically contained many 10 small and highly specialized firms in close proximity to one another, and were the centres of economic activities in their industries. The geographic proximity minimized transportation costs, provided access to specific forms of labour, produced a division of labour at the level of the firm, and what Marshall called spillover effects that are economic gains from close proximity and interactions with other firms. This work by Marshall, however, was done in the early part of the twentieth century, not attracting much attention until recently. Marshall's work on industrial districts was revitalized though the work of Piore and Sabel (1984) on industrial restructuring in Italy. Amin (2000) indicates their work as the turning point, or catalyst, for the discussion of the resurgence of regional economies in academic research. Historically, industrialization resulted in intra-national spatial unevenness because of market structures and resource availability (Scott 1988; Scott 1998), and in turn producing marked regional differentiation (Walker 2000). The industrial restructuring that occurred during the 1970s and 1980s, the concern of Piore and Sabel (1984), was expressed spatially as "old" manufacturing regions losing their dynamic character in favour of "new" manufacturing regions. As a result, the regional geography of advanced capitalist countries changed, significantly. The new industrial regions, formerly peripheral (or non-existent) to the economy, currently function as the "new engines of the global economy" (Scott 1998: 22). The work of Scott (1988; 1998) further shows that as the resource needs of industry changes in the new global industrial economy, so do the places in which industry operates. Building on the work of Piore and Sabel, Storper (1999; 2000) further argues that critical to understanding why the region has become central in dynamic capitalist 11 economies are cultural factors, what he calls untraded interdependencies that generate regional-specific assets in production. As the technological aspects of production become increasingly complex, the ability to perform one aspect of production is predicated on knowing how to do another—technological spillovers. This knowledge, or know-how, is the essence of the interdependencies that exist both within and between firms. These interdependencies are largely intangible aspects of the production process— labour market operations, customs, understandings, and values—that makes them untraded and, therefore, regionally specific (Storper 1999). These untraded interdependencies are precisely why there have been difficulties transplanting technology into another region, only to have that technology fail or flounder (Gertler 1997). Though this literature is confined to examining high technology sectors of the economy, its relevance to my dissertation is in highlighting the region-specific assets of production. The new industrial economy is not only located in particular places or regions that are different from the old industrial economy, but contains forces that continually reproduce the regional character of production and, therefore key for my dissertation, promote international trade flows. Regardless of the sector of the economy, it is regions not nations that trade. The final relevant literature on the region comes from social-political policy, and is framed in terms of the decline of the nation state. The primary argument is that while the economic significance of the nation state has withered through the globalization process, sub-national regions such as the Third Italy, California's Orange County, and Route 128 have become critical for economic growth. This hypothesis is particularly associated with Kenichi Ohmae (1993, 1995), who defines regions as "natural economic 12 zones...[that]...may or may not fall within the geographic limits of a particular nation" (Ohmae 1993: 78 - 79). Ohmae (1993: 78) has even gone as far to say that the nation is a "dysfunctional" spatial unit of analysis for human activities. He writes: "[e]ach [nation] is a motley combination of territories with vastly different needs and vastly different abilities to contribute" (Ohmae 1995: 12). Certainly, Canada and the United States are both defined by tremendous regional economic diversity. For Ohmae (1993) a sustainable region must be small enough to exhibit similarity in both its human and physical geographical spaces (a criterion not so different from Ohlin), but large enough to have the infrastructure to be able to participate in the global economy. By their very nature, regions are the centre of economic action within nations, are international in scope, and are not dependent on the nation state for the necessities of production (investment, industry, information, and individuals)—Ohmae's 4 "I's" of the global economy. The days of the nation state setting policy for regional economic development are no longer with us. Regions, through their own economic dynamism are the places that attract investment, industry, information, and individuals (Ohmae 1995). These regions possess the most dynamic trading relationships. In summary, the three literatures I have briefly reviewed each point in different ways to the centrality of the region to international trade. It is not nations as such that trade, but regions. M y purpose in this dissertation is to take this insight and apply it to understanding the geography of international trade flows between Canada and the United 7 Also relevant to the concepts outlined by Kenichi Ohmae is the cross-border trade region. Such regions, by definition, contain a national border. See Edgington (1995) and Wright (1999) for North American examples of the cross-border trade regions: Cascadia and the Mexican maquiladoras, respectively. 13 States. The theory behind the importance of the sub-national region is not itself evaluated or extended, but that literature is used as a point of departure to show the importance of a sub-national approach to the study of international trade. Additionally, the sub-national approach to the study of trade flows between Canada and the United States is an under-developed literature. The volume of literature investigating the effects of free trade agreements on Canada - United States international trade flows is quite large. However, these studies analyze trade at the national (Krueger 2000, Clausing 2001), industry (Trefler 2004), and product (Romalis 2005) levels. These are only four studies of trade flows between Canada and the United States, but are typical in the way this trading relationship is studied. This is despite the fact that each province within Canada has a different industrial mix and, therefore, tariff and non-tariff structure. Consequently, when these barriers to trade are reduced or eliminated through national policy, there are going to be differential effects across the economic landscape. Very little research analyzes the geographical aspects of Canada - United States international trade flows. Coughlin and Wall (2003) and Wall (2003) investigate the geographical effects of the N A F T A , finding strong support for their respective geographical approaches. Polese (2000) and Acharya et al. (2003) also take a geographical approach to Canada - United States international trade flows, but use aggregated sub-national regions rather than individual provinces and states. There are a few other geographical studies of Canada -United States international trade (see Brown 1998, Norcliffe 1996, and Brown and Anderson 1999), but they do not investigate this geography over time. Consequently, very few studies have investigated this phenomenon. Therefore, there are not only 14 theoretical motivations for studying trade at the sub-national region, but motivations from a lack of relevant research in this area. 1.5. M Y GENERAL RESEARCH QUESTIONS AND CHAPTER OUTLINES Given the nature of my goal, the dissertation is overwhelmingly empirical involving a comprehensive sub-national numerical study of the geography of Canada - United States international trade flows. More specifically, my intent is to uncover the regional trading patterns between Canada and the United States to establish if these trading patterns have changed as a result of the C U F T A and/or the N A F T A . Stemming from this general intent are four sub-questions, each of which raise different concerns: (1) How important is geography to international trade in general? (2) What is the current state of knowledge regarding the trading relationship between Canada and the United States? (3) What is the geography of trading patterns between Canada and the United States? And (4), how has the changing geography of Canada - United States international trade flows impacted the trading relationships within these two countries? (1) How important is geography for international trade in general! The answer to this question is first undertaken at the international scale by placing Canada - United States international trading relationship within a global context (chapter 2). M y point of departure in this chapter is to search for a geography of international trade that is distinct from that of the geography of production. If the geography of international trade is simply the mirror of the geography of production (the countries that produce the most trade the most) then the study of the geography of international trade is merely the study 15 of the geography of production: international trade is governed by production, so the factors that produce the geography of production should be analyzed. Once the distinction between the geography of international trade and production is established, bilateral trading relationships are analyzed to uncover the importance of geographical relationships in establishing bilateral trade. Is international trade determined by economic and/or political factors, or is international trade more a result of geographical factors? If so, what are those factors? (2) What is the current state of knowledge regarding the trading relationship between Canada and the United States? This second sub-question is dealt with in chapter 3, 4, and 5. The answer to this question is first undertaken as a review of the history of Canada - United States international trade flows, beginning before Canada's confederation and ending with the Canada - United States Free Trade Agreement and the North American Free Trade Agreement. The literature investigating the effects of these two recent free trade agreements is also reviewed to provide context for the changes from 1989 - 2003. Additionally, in order to compare with the regional results, a national level of analysis is carried out in chapter 4. International trade flows between Canada and the United States are investigated at the industrial sector level to search for changing patterns at the national-industrial level. This analysis both complements and extends beyond research on the study of Canada - United State international trade flows, by analyzing international trade itself. Finally, chapter 5 resolves two outstanding issues contained within the literature on international trade between Canada and the United States: the border effect and the N A F T A effect. 16 (3) What is the geography of trading patterns between Canada and the United States? With the baseline study at the national level, and outstanding issues resolved, the analysis turns to the regional geography of Canada - United States international trade flows in chapter 6. Rather than using the nation as the spatial unit of analysis or the sub-national region composed of many Canadian provinces or U.S. states, this geographical analysis takes advantage of the most geographically disaggregated data available for Canada's international trade: international trade flows measured at the Canadian province and U.S. state levels of analysis. By undertaking this geographical approach, the varied effects of the free trade agreements are uncovered, revealing that not all provinces are experiencing the same effects from the free trade agreements. Additionally, through this geographical approach, I show that the factors behind the changing spatial distribution of international trade flows since the establishment of the free trade agreements are not restricted to changes contained within the free trade agreements themselves. Rather, changing spatial patterns are, in part, a result of industrial restructuring that pre-dates the negotiations of the free trade agreements. (4) How has the changing geography of Canada - United States international trade flows impacted the trading relationships within these two countries? This last sub-question is addressed in chapter 7 that contains two analyses. First, through the development of a new spatial pattern change test, the degree of spatial change in interregional trade for each Canadian province is investigated, uncovering that different provinces have undergone different degrees of spatial change with regard to their trading relationships within Canada and the United States. Using this spatial differentiation of change as a stepping stone, the analysis of the changing geography of Canada - United 17 States international trade flows turns to trading regions within these two countries. Using the methodology developed in chapter 2 at the international level, trading regions are identified within Canada and the United States for both 1989 and 2001 to show the changing geography of their trade. After these questions have been answered in chapters 2 through 7, chapter 8 evaluates the answers to these four questions. I conclude that only through a geographical approach can a true understanding of the trading relationship between Canada and the United States take place. Both Canada and the United States are too large to be analyzed each as one spatial unit, and that such analyses hides the dynamic nature of the sub-national regions. The chapter finally notes some of my overall conclusions and provides some directions for future research. The answers to the questions posed above provide the first comprehensive account of the sub-national (geographical) trading patterns between Canada and the United States. Essential to this account of Canada-United States international trade flows is geographic information systems (GIS). The management of the different data, particularly the topographical relationships between the many regions in these two countries, is greatly simplified within a GIS, allowing for a richer analysis of the topic at hand. The result is an appropriate understanding of the geography of international trade, a greater understanding of the historical circumstances of the trading relationship between Canada and the United States, a detailed account of the geography of Canada -United States international trade flows, and an understanding of the changing geography of those international trade flows. CHAPTER 2 18 In search of the global triad: does trade have its own geography? 2.1. I N T R O D U C T I O N In this first analytical chapter of my dissertation, the geography of global trade flows is investigated to place Canada - United States international trade flows within a global context. As discussed above in the introduction, global international trade flows are increasing rapidly, so any changes in the patterns of Canada - United States international trade flows are not occurring in isolation. Though the geography of Canada — United States international trade flows is the result of particular local and historical outcomes in these two countries, some of these outcomes may be resulting from more global trends such as an increasing presence in the international economy. Therefore, understanding the nature of the geography of global international trade flows may be instructive to understanding the geography of international trade flows between Canada and the United States. As stated in the introduction, over the past 20 years the international exports of merchandise more than doubled while world output increased by approximately 50 percent (IMF 2005a; I M F 2005b), and trade in services, though a relatively smaller portion of international trade, is growing at an even faster rate (Dicken 2003). Yet, as stated by Dicken (2004), this geographical process is relatively unexplored by 19 geographers. International trade is an increasingly important component (both directly and indirectly) of our economic livelihood at every scale and international trade volumes are increasingly centred on Western Europe, North America, and East Asia. This regionalization of the world economy led Ohmae (1985) to describe the world economy as a global triad: international trade and investment dominanted by the United States Germany, and Japan—Ohmae (1985) was the first of many to describe the world economy in this manner. Both the importance of international trade and its regionalization makes international trade an obvious avenue of research for (economic) geographers despite Dicken's (2004) finding that geographers are largely not included in the discussions around global trade. Globalization, however, is not a process occurring impromptu, but a process that is increasingly institutionalized through international negotiations regarding international trade and investment, manifesting in regional trading agreements. With both international trade volumes and a large portion of these regional trade agreements centred on Europe, North America, and East Asia the global triad is not only a de facto regionalization project, but a regionalization project that is now largely formally institutionalized. Poon et al. (2000) outlines the major formal institutionalization in each of the three triad regions since 1985: the Single European Act of 1986, the 1992 single European market programme, the launch of the Euro currency in 1999, the establishment of the North American Free Trade Agreement ( N A F T A ) in 1994, the advent of the Southern Common Market (Mercosur) in 1991, and the expansions of the Association of South East Asian Nations (ASEAN) with its affiliated free trade agreement. Despite this 20 large number of significant moves toward formally institutionalizing the global triad from 1985 to 2000, the six years since Poon et al. (2000) wrote have also seen considerable movement toward formal institutionalization in all three global triad regions. North America has seen the least formal expansion of its triad region, but has been in negotiations to form the single most important economic region in the world—the Free Trade Agreement of the Americas (FTAA) . Encompassing thirty-four countries in North, Central, and South America, the economic size of the F T A A would dwarf that of the current European Union—the countries forming the N A F T A are economically larger than the current European Union. Containing some of the largest international trade flows in the world, the F T A A would be formidable in the world economy. Because of massive protests to the F T A A in most negotiating countries, its implementation at the end of 2005 had been delayed, however ( F T A A 2006). The European Union has now incorporated ten countries in Central and Eastern Europe, the CEEC-10, expanding its membership to twenty-five nation-states. Though the economic component of this expansion is smaller than the expansion to occur with the F T A A the expansion solidifies the European Union's place in the world economy as the second largest free trade area in the world. In East Asia, Japan and Singapore formed Japan's first region-specific trade agreement in 2002. Singapore also established free trade agreements with New Zealand (2001) and Australia (2003)—Thailand also entered into a free trade agreement with Australia in 2005 (WTO 2006). However, probably the most significant free trade agreements established are those involving China, a country that is increasingly becoming an economic behemoth in the global economy. In 2004, China signed free 21 trade agreements with two of its special economic zones, Hong Kong and Macau. But in 2003, China entered into a free trade agreement with A S E A N (WTO 2006). A l l of these regional trade agreements, aside from those involving New Zealand and Australia, are particularly significant to East Asia because of its military and economic history. The Japanese military domination in East Asia in the first half of the 20 t h century and the concern over the potential future economic domination by China, has made many East Asian countries leery of formal regional economic integration (Korhonen 1998; Kwan 1997; Narine 1999). Therefore, these moves toward the formal institutionalization of East Asian regional trade agreements is significant, and, as with North America and the European Union, lends support to Ohmae's (1985) thesis. There are, of course, critics of the global economy as a global triad. The increasing presence of the triad regions leads some to warn of possible protectionism (Aho 1994) as well as the possibility of triad-based international agreements superseding bilateral discussions not covering the same international concerns (Lawrence 1994; Wei and Frankel 1997). Such fears cause some to believe that the formation of the global triad in particular, or regional trading agreements in general, are deleterious for economic welfare (see Bhagwati (1999) for a discussion on these issues). However, the presence of the global triad or any other regional trading agreement is not a sufficient condition for these concerns. The W T O , and the G A T T preceding it, requires all of its members entering into regional trade agreements to satisfy particular criteria. For example, all 8 It should be noted that many free trade agreements have also been established between countries in each of the triad regions, triad region - non-triad region countries, and non-triad region - non-triad region countries. 22 countries in a regional trade agreement cannot on average increase their trade barriers to non-member countries. Of course there are interpretations involved in many of these rules set out by the W T O and any given country may chose to act unilaterally despite the W T O rules, but there are no a priori justifications for these concerns. Given the importance of the international economy to most countries and the attention paid to issues surrounding regional trading agreements (see Bhagwati 1999; Gibb and Michalak 1996; and Michalak and Gibb 1997) one would expect there to be a great deal of attention paid to the formation of the triad regions themselves by both geographers and economists. As noted by Poon et al. (2000), however, this is not the case. 9 Very few empirical studies have been undertaken to investigate the existence of the global triad. And those studies that have investigated the existence of the global triad (Anderson and Norheim 1993a; O'Loughlin and Anselin 1996; Poon 1997; and Poon et al. 2000) find little evidence to support its existence—evidence of clustering in international trade and trading regions are found, but not the global triad itself. This chapter adds to the existing limited literature in two ways. First, I analyze the geography of international trade flows from a different perspective than past research. Past research has focussed on international trade shares without appropriately considering the economic sizes of trading economies. The present analysis suitably accounts for these economic sizes. Second, I identify those regions in the world that trade intensely with each other to evaluate whether the global triad exists by using a methodology to establish economically meaningful trading regions. The following section reviews the 9 There are many studies investigating international trade patterns within each of the triad regions. The studies relevant for North America are discussed in chapter 6. 23 empirical literature investigating the global triad. Section 3 presents the data, measurement, and methodology employed. The empirical results are reported in Section 4. And Section 5 concludes that the global triad is the result of not considering trade intensity in terms of both the intensity of trade shares and trade volumes. 2.2. PAST RESEARCH ON THE REGIONALIZATION OF THE WORLD ECONOMY The literature attempting to establish empirically the existence of the global triad, or the regionalization of international trade flows in general, has two fundamental issues to review: the measurement of trade and the method of regional assignment. The measurement of international trade is critical to establishing whether or not the global triad exists and, hence, varies from study to study as different authors attempt to capture international trade flows. Similarly for the method of regional assignment, depending on the method used to assign nations to regions the resulting regions may differ substantially. Each issue is discussed, in turn. 2.2.1. Measurement Trade intensity refers to how intensely one country trade with another country. The type of measurement used to investigate the global triad is the trade intensity index. The index is based upon the degree to which any two or more countries trade with one another more or less than expected. If two countries have an intense trading relationship, this represents a geographical bias in international trade flows: one country sends and/or receives a larger portion of its trade to and/or from one particular place than would be expected. 24 The basic trade intensity index—used by O'Loughlin and Anselin (1996), Poon (1997), and Poon et al. (2000)—is computed as: where Xy is the share of country f s exports to region j and m, is the share of world imports destined for region j. Essentially, this index captures whether or not a country sends a disproportionate share of its exports to a particular region, relative to the rest of the world. If / y is equal to one, country / exports proportionately to region j, relative to the rest of the world; if ly is greater than one then country i exports a disproportionately larger share of its exports to region j, relative to the rest of the world; and if 7 y is less than one, country i exports a disproportionately smaller share of its exports to region j, relative to the rest of the world. 1 0 Therefore, this index measures the degree of trade specialization of one country with another. As an index, it has the same interpretation as the location quotient. However, this index has two limitations stemming from the fact that any geographical bias found in country f s exports is measured relative to all other countries' trade. The first limitation is that i f all (or most) countries exhibit a geographical bias in their trade with region j, as may be expected if the world is indeed breaking into triad-based regions, then country i may not appear to trade intensely with region j even i f it does so at the same degree as the rest of the world. For example, i f all of the countries in the world export 50 percent of their trade to one country, a clear indication of a geographical bias in international trade, and country i also exports 50 percent of its trade to this country, then ly does not indicate any trade intensity or specialization. But clearly 1 0 Region j may be a country. (2.1) 25 trade specialization is occurring, country i simply is not different than the rest of the trading countries of the world. The second limitation is that only considers export flows to identify a geography of international trade. However, this is despite the fact that the spatial distribution of production shows that particular countries produce disproportionate shares of the global output. For example, in 2001, the United States contained approximately 5 percent of the world's population but accounted for over 27 percent of the world's gross domestic product. This geography of production is a long-standing fact in economic geography that operates at all scales, whether it be at the local, national, or international (Walker 2000). Therefore, the geography of production shows a small proportion of the world's countries producing the vast majority of the world's output, one would expect that global trade is dominated by these same countries simply due to their economic size. If the geography of international trade follows the same spatial pattern as the geography of international production, then it is very likely that there is no geography of international trade independent of the geography of international production—countries simply export their largest shares of trade to the economically largest countries in the world, and similarly to the economically smallest countries in the world. If it is the case that there is no independent geography of international trade, then to understand the geography of trade flows we need to understand the geography of international production. However, i f an independent geography of international trade does exist, there is a need to study its specific spatial pattern separate from the spatial pattern of production. Herein lays the limitation of lj/. this measurement only considers export flows without incorporating the geography of international production. 26 In an attempt to control for the geography of international production using economic size, Anderson and Norheim (1993a; 1993b) and O'Loughlin and Anselin (1996) compute the following trade intensity index: where J, is the ratio of country /'s exports to the world to country f s gross domestic product, GDP,, and ly is the same as Equation 2.1. However, using Xy to represent the volume of exports from country i to region j, Equation 2.3 shows that Py is problematic for interpretation: The first term in Equation 2.3 does have the interpretation of country f s dependence on region j through trade, but the second term in the equation is the inverse of rtij and which confounds the interpretation of Py. Therefore, ly does have a natural interpretation with its location quotient form, but Py does not have a natural interpretation. Py changes due to country f s changes in its economic dependence on region j through trade, but also from changes in global exports to region j and total global exports. The consequence, as recognized by O'Loughlin and Anselin (1996) themselves, is that the value of this index cannot be meaningfully compared between regions that are of a different size. The use of ly, by Anderson and Norheim (1993a) and O'Loughlin and Anselin (1996), does find that exports are focussed on geographically close countries, particularly (2.2) V J (2.3) x y Total Worl d Exports GDP. World Exports j 27 in North, Central, and South America, but there is not any evidence of the increasing presence of the global triad over a time period spanning 25 years (1968 - 1992). This time period may not include much of the more recent formal institutionalization discussed above, but it does span the time period which Ohmae (1985) made the first claim of the global triad. With regard to Ptj, there is evidence of this index increasing over the same study period but, as stated above, the interpretation of this trend must be done with caution. The total exports from country i to region j and world exports to region j may have stayed exactly the same given the fact that total world exports have grown faster than G D P over this entire time period—such a phenomenon would actually indicate a decrease in the regionalization of global trading patterns. Therefore, as stated by Anderson and Norheim (1993a) and O'Loughlin and Anselin (1996), the use of these indices alone provides no evidence of an increasing presence of the global triad. 2.2.2. Regional Assignment The five empirical investigations of the regionalization of the global economy discussed above, use their respective trade intensity indices to aggregate individual countries into trade regions. These trading regions are then compared to the global triad countries to see whether or not the claims of the global triad's existence are supported empirically. This research investigating the formation of the global triad has used two methods of regional assignment. The first, only used by O'Loughlin and Anselin (1996), is a spatial statistical method used to search for clustering in the export and import shares of the United States, Germany, and Japan—with these three countries representing the global triad regions. And the second, used by Poon (1997) and Poon et al. (2000), is the 28 Intramax method in conjunction with the trade intensity index, ly, that assigns every country in the world to a regional trade bloc. Each is discussed in turn. O'Loughlin and Anselin (1996) use two spatial statistical measures to assess the existence of the global triad, one global and one local. 1 1 The global measure is the commonly used Moran's I, a global measure of spatial autocorrelation, and the local measure is the G,* statistic of Getis and Ord (1992), one of the local indicators of spatial association (Anselin 1995). That the use of Moran's / indicates evidence of the clustering of exports for each of the United States, Germany, and Japan, is consistent with their prior findings when using both ly and Py. However, as with the trade intensity indices there is no evidence of increased spatial clustering over their study period. Clustering is present but it existed long before any claims of the global triad with no evidence that clustering increases over time. The results from the analysis of the G,* statistic are qualitatively similar to that of Moran's I, though more pronounced: strong statistical evidence of the clustering of exports, but no significant changes since the late-1960s. Though the use of spatial statistics does seem appropriate for analyzing a spatial phenomenon, the spatial statistical methods are used to search for the global triad in cartographic space. Therefore, spatial contiguity and cartographic distance are used for the identification of regions. However, as noted by Allen et al. (1998), regions are not 1 1 Global spatial statistical measures, in the present context, will identify the clustering (or lack thereof) of international trade flows for the world as a whole. The concentration of international trade flows for all countries does not vary. Local spatial statistical measures, however, allow for the identifications of clustering of international trade flows in multiple places on the same map, with that clustering being able to vary in intensity across space. 29 necessarily spatially contiguous arrangements of economic space, they may have holes within them such that the region is not continuous—Allen et al. (1998) are working in the context of the sub-national region, but the concept applies at the supranational scale as well. This regional typology is particularly instructive for East Asia. Many of the countries in what some consider an East Asia region share no common boundaries with any of their fellow neighbours. Consequently, searching for regions in cartographic space may not identify the formation of regions (the global triad) because not all countries are spatially contiguous, both within and outside of the global triad regions. The second method of regional assignment does not rely on the strict cartographic relations of contiguity and distance and it called the Intramax method. The Intramax method used by Poon (1997) and Poon et al. (2000) is a clustering algorithm that uses the trade intensity index, /„, to form trading regions. This algorithm has the advantage of allowing the data to determine the nature of the regional formation, rather than relying on an a priori designation of regions—the algorithm determines trading regions endogenously. As such, i f the global triad exists, regional trading blocs centred on the United States, Germany, and Japan wil l emerge, rather than being imposed on the data. This method identifies two countries that have the greatest intensities of trade and then collapses them to form a region—they are hereafter treated as a single spatial unit with their exports aggregated together. With this reduced set of countries, the trade intensity index is recalculated and the next greatest intensity of trade between two countries (or regions) is identified and they are collapsed again to form a larger region. This second region may or may not include the first two countries that themselves form a region. This procedure is repeated iteratively until all of the countries in the data set are 30 assigned to a region (Poon et al. 2000) (see Masser and Brown (1975) and Fischer et al. (1993) for more detailed accounts of this algorithm). The general result employing from this procedure is an increase in the regionalization of global trade since 1985, but still a global triad had not been established by 1995. In 1985, the year of Ohmae's (1985) claim of the global triad, there were seven trading regions in the world, with three of those regions being centred on countries in the European Union: the United Kingdom, Germany, and France. This leads Poon et al. (2000) to claim that the evidence for the establishment of the global triad is weak, but there is a definite increase in the regionalization of global trade since Ohmae's (1985) first claim of the global triad's existence. As with the spatial statistical methods of O'Loughlin and Anselin (1996), there are some methodological concerns around the Intramax method. The first is that all countries are assigned to a trading region. And the second is the use of trade intensity to establish those regions. The necessity of all countries being assigned to a region does not explicitly conflict with the endogenous nature of regional assignment using Intramax (no regions are imposed a priori), but it does go against the principle. It is possible that all countries are involved in a trading region, but more likely there are countries that are effectively not part of the global trading system. This is not to say that these countries have no trade, but that they are not part of a group of countries that trade intensely among themselves. And this is precisely why there is concern regarding the regionalization of the global economy in the first place (see Bhagwati 1999). If the global triad is indeed established, and the global triad regions are, as claimed, becoming insular with respect to trade, the countries outside of the strict global triad regions (such as South America, Africa, and 31 South Asia) do not likely belong to a global triad-based trading region. Therefore, there must be a point at which countries are no longer assigned to regions. Trade intensity is clearly an important criterion for the establishment of a region, but that intensity alone may be misleading. For example, in Poon et al. (2000), the first region to be identified is the Ivory Coast and Burkina Faso. As shown below, these two countries do form a region in 1981 and 2001, but the volume of trade flowing between these two countries is incredibly small in the context of world trade. The Ivory Coast and Burkina Faso may trade very intensely with each other, but their presence in the global economy is of little consequence. Poon et al. (2000, 431) notes that the absolute volume of trade may "obscure regional patterns because they do not normalize for country size," but this is an issue with the measurement of trade intensity and should not be ignored in regional assignment. Not considering the volume of trade in regional assignment led Poon et al. (2000) to assign France and Germany to different trading regions in 1985, when these two countries had the second largest bilateral trading volumes in the world in that year. It may be the case that France and Germany each have different countries that are highly dependent on trading with them due to their different histories, but any economically meaningful trading region must have France and Germany together because of the magnitude of their trade with each other. In other words, there is a great difference between economic dependence and economic integration. If international trade only flows in one direction between two countries, then one of those countries is likely dependent on the other. But if those two countries trade intensely with each other such that each country is dependent on the other country through trade, then the two countries are integrated. This issue of trade reciprocity is discussed further below. 32 Previous analyses of the existence of the global triad have shown that there is some evidence of clustering in international trade flows. However, little evidence of increased clustering and the global triad itself is present. The measurement of trade intensity and the methodology for regional assignment are problematic in terms of interpretation and economically meaningful trading regions, respectively. These limitations are both dealt with in the present analysis, discussed in the following section. 2.3. DATA, MEASUREMENT, AND METHODOLOGY In order to address the issues of the interpretation of trade intensity and the establishment of economically meaningful trading regions, the appropriate data are needed to make empirical measurement of trade intensity and the method of regional assignment must be developed. As such, data, measurement, and the regional assignment methodology are respectively discussed here. 2.3.1. Data The data on aggregate cross-national and bilateral international trade used in the present empirical investigation of the regionalization of the global economy come from Statistics Canada's (2004a) World Trade Database, 1980 - 2001 and data for gross domestic product (GDP) come from the International Monetary Fund's (2005a) International Financial Statistics. Further data selection is based on the availability of both bilateral trade flow data and G D P for each country for the years 1981, 1991, and 2001. For reasons of comparability over time, the countries that formed the former Czechoslovakia, Yugoslavia, and Soviet Union are combined, as are China and Hong Kong. The final 33 data set consists of 152 countries, covering the vast majority of global international trade flows—as shown in Table 2.1, more than 96 percent of global imports and 97 percent of global exports are represented for all years. With respect to the global triad, all North American and European Un ion 1 3 countries are represented. The A S E A N - 1 0 , 1 4 China, South Korea, Japan, Australia, and New Zealand represent East Asia ( E A - A N Z ) . Hong Kong is included in China for data comparability and Taiwan is absent due to the unavailability of gross domestic product measures. Due to the importance of these two special economic zones in China and their significantly different trading behaviours than China as a whole, this absence is problematic. However, this is a limitation in the data that is unavoidable. <See Table 2.1, page 227> 2.3.2. Two Measures of Trade Intensity The measurements used in this analysis are similar to those used in past research investigating the regionalization of the global economy. However, the particular measurements used in this analysis provide a much clearer interpretation with respect to the geography of international trade that is distinct from the geography of international production. The first measurement, though not a location quotient in its pure form, is interpreted in the same way as the location quotient so it is referred to as the trade location quotient, LQT, . The LOT, only measures a single country's exports and imports 1 2 Within the country aggregations, there are actually 178 countries represented. 1 3 The European Union countries represented are the EU-15: Belgium-Luxembourg, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, and the United Kingdom. 1 4 Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, and Vietnam. 34 (as opposed to a bilateral relationship) to investigate how intensely a country is involved in the global economy in general, and is calculated as: imports), GDP, is the gross domestic product of country i, and GDPW is the gross domestic product of the world. Therefore, the LOT, index measures the percentage of world exports (or imports) from country i relative to its share of world G D P . Herein lies its ability to distinguish a geography of international trade as distinct from the geography of international production. If the LOT, index is equal to unity, a country exports (or imports) proportionately to its share of world GDP; if the LOT, index is less than unity, a country exports (or imports) disproportionately less than its share of world G D P ; and i f the LQT, index is greater than unity, a country exports (or imports) disproportionately more than its share of world G D P . Therefore, a geography of international trade distinct from that of the geography of production exists i f countries exhibit L Q T , values different from unity. From a theoretical standpoint, the LQT, index is able to account for the well-established (see Walker 2000) geography of production. As discussed above, the United States is the largest economy in the world, so one would expect, a priori, that the Untied States' international trade volumes are also the largest in the world. The LQT, index accounts for the size of the United States and measures its trade activity relative to its economic size rather than only considering trade volumes. For descriptive and mapping purposes, the categories for the trade location quotient follow Miller et al. (1991) used in their analysis of regional economic development: if the LQT, index is less than 0.70, a country is very underrepresented; i f (2.4) where x, is the value of exports (or imports) from country /, Xw is total world exports (or 35 the LQT, index is 0.70 - 0.90, a country is moderately underrepresented; i f the LQT, index is 0.91 - 1.10, a country has average representation; i f the LQT, index is 1.11 -1.30, a country is moderately overrepresented; and if the LQT, index is greater than 1.30, a country is highly overrepresented. A l l of the maps presented below used these ranges as legend categories for ease of comparison between different regions of the world. This first measurement is instructive to search for the overall pattern of the geography of international trade, but it is limited because it does not consider bilateral trading relationships. Hence, the LQT, index cannot be used for regional assignment because it does not measure each individual country's geography of international trade with all other countries of the world. In order to mediate this limitation and allow for regional assignment, a bilateral version is calculated as: where xy is the value of exports from country / to country (or region) j, xiw is country i's total exports to the world, GDPj is the gross domestic product of country j, and GDPW is the gross domestic product of the world. The LQTy index, then, measures the percentage of country f s exports sent to country j relative to country /'s share of world GDP. If the LQT,y index is equal to unity, country i exports to country j proportional to country fs share of world GDP; i f the LQT// index is less than unity, country i exports to country j disproportionately less to country j relative to country fs share of world GDP; and i f the LQTy index is greater than unity, country i exports to country j disproportionately more to country j relative to country fs share of world GDP. Therefore, the LQTy index measures the geography of international trade as distinct from the geography of international production from each country to each (2.5) 36 individual country in the world. In the present analysis, each country has 151 trading partners and, therefore, 151 LQTy index values. For each country, the LQTy index measures the degree of trade intensity with each individual country in the world, with the geography of production considered in each and every case. The primary advantages of the LQT,y index for regional assignment are that it maintains the interpretation of the /y index, but it also controls for regional economic size. This controlling for regional economic size was captured by Py, but at the cost of interpretability. 2.3.3. The Regional Assignment Algor i thm As discussed above, the Intramax method has two limitations: all countries are assigned to a trading region and only trade intensity is used to determine the order of regional assignment. The algorithm developed for the present analysis addresses these two limitations while maintaining the strength of the Intramax method: the algorithm used to form trading regions does not use the global triad as its point of departure, and it is an endogenous data-driven algorithm that assigns countries to regions based only on the criteria set out below. In order to address the first issue of all countries assigned to a trading region, the concept of a reciprocal bilateral trading relationship is introduced. A reciprocal bilateral trading relationship is established i f and only i f both the LQTy and LQT,, for two countries are greater than a pre-determined threshold value. This criterion requires not only for one country / to export disproportionately more to another country j more than would be expected based on country j's share of world GDP, but also for country j to export disproportionately more to country i based on country f s share of world GDP. The pre-determined threshold value of 1.30 is used in the analysis below to capture 37 highly overrepresented reciprocal bilateral trading relationships, followed by a sensitivity analysis using the values of 1.20 and 1.10 to ensure that the finding presented here are not sensitive to changes in the critical threshold value. This reciprocal bilateral trading relationship is the natural point of departure to search for the global triad. Large volumes of trade flows originating from the global triad regions are not enough to establish the existence of the global triad. Each triad region contains the largest economies in the world so it is expected that large volumes of trade originate and are destined for these countries. Therefore, reciprocity in intra-triad regional trade must be established to confirm or deny the existence of the global triad. Otherwise, countries are simply exporting large volumes of trade to other large economies that happen to be in their " t r iad" region. 1 5 Once a reciprocal bilateral trading relationship is established, the L Q T y index is no longer used in the analysis. Whether a country exports disproportionately more to another country by 30 percent or 100 percent does not matter. The two countries simply have to have an intense trading relationship. More important at this stage is to incorporate the volume of trade between these two countries. 1 5 This condition of reciprocity in trade flows is particularly important for the determination of a trading region. For example, suppose that Canada exports to the United States without reciprocity, the United States exports to Mexico without reciprocity, and Mexico exports to Canada without reciprocity. Also suppose that each of these exports is considered "intense". Would it make sense to consider these countries a region? Each country may be dependent on another country through trade, but to classify this, albeit connected, dependence as a trading region is suspect. It is possible that such a relationship may be a trading region if the connection is through a production chain. However, such a claim cannot be made with relatively coarse aggregate trade flow data. 38 As stated above, not considering the volume of trade generates a regional assignment algorithm making some curious regional assignments: France and Germany are in separate regions, and the first regional assignment involves two countries with very low volumes of trade. To prevent such curiosities, the volume of trade within reciprocal bilateral trading relationships is used to assign the regions. The reciprocal bilateral trading relationship with the largest volume of international trade is used to assign countries to a region rather than the magnitude of their trade intensity. Therefore, the iterative regional assignment algorithm employed is as follows: using the LQTy index, identify all reciprocal bilateral trading relationships; aggregate the trade flows within each reciprocal bilateral trading relationship and rank them by the magnitude of their aggregate trade flows; classify the largest magnitude aggregate trade flow as a region; recalculate the LQTy index treating the region as one spatial unit; repeat until there are no reciprocal bilateral trading relationships. The temporal component of this analysis (1981, 1991, and 2001) serves to show any changes in trading regions over time. The combination of the LQT indices and the regional assignment algorithm provides a trade intensity measure and regions that are economically meaningful—the largest bilateral trading relationship in the world, Canada and the United States, is the first region identified. Rather than only considering the relative intensity of international trade flows, the methodology presented here uses relative intensity to establish reciprocal trading regions and then incorporates trading volumes into regional assignment. In this manner, the present methodology uses both relative and absolute measures of international trade flows to establish trading regions. Given the nature of the regional 39 assignment algorithm, using reciprocal bilateral trading relationships, the trading regions established are referred to as reciprocal trading regions in the results below. 1 6 2.4. EMPIRICAL RESULTS 2.4.1 The Trade Location Quotient, Global Imports, and Global Exports The foundation for the claim of a global triad is shown in Table 2.2. Despite the spike in 1991, the share of world imports and exports of North America, EU-15, and E A - A N Z is increasing steadily such that more than 80 percent of international trade flows move to and from these nations. However, such an interpretation is premature. As stated above, it is the bilateral relationships that define a global triad and all the concerns associated with such a relationship. Also shown in Table 2.2 is the world share of G D P for these three regions. As with exports and imports, they comprise of more than 80 percent of the world's economic activity. Therefore, it should be no surprise that these same countries comprise the majority of global international trade flows. The United States, for example, is the largest economy in the world (27.1 percent of world G D P in 2001), so it has the world's greatest capacity to absorb global imports (18.6 percent in 2001) and to 1 6 This method of regional assignment does have a bias toward larger economies in region formation, much like the Intramax method has a bias toward smaller (though extremely open) economies. The present regional assignment algorithm may exclude some small economies from regions because they do not have sufficiently large trade volumes with the economically large countries. However, as shown below, the vast majority of countries are assigned to a trading region or reciprocal bilateral trading relationship, including many small economies. Therefore, this potential bias toward larger economies is believed to be of lesser concern than the issues of bias found within the Intramax method. 40 ship global exports (12.2 percent in 2001). However, the United States will appear to be a focal point of international trade simply due to its massive economic size. <See Table 2.2, page 228> The existence of a unique geography to international trade is evident in Figures 2.1 - 2.3. From 1981 to 2001, Canada, Northern Europe, and East Asia maintained their overrepresentation in both exports and imports with East Asia increasing its overrepresentation in the most recent years. Africa has seen the most radical changes having their representation of exports and imports fall significantly in Western and Sub-Saharan Africa for exports and essentially the whole continent for imports. With respect to the three regions of the triad, North America is rather underrepresented in both imports and exports. Canada, the only North American nation consistently labelled as "highly overrepresented" has relatively low L Q T index scores with respect to other countries in the world. Mexico has only recently (1996) moved into the status of high overrepresentation, and the United States is consistently very underrepresented for both imports and exports. The United States' lack of relative standing in international trade flows is likely due to the vast size of its internal market. <See Figures 2 .1 -2 .3 , pages 330 - 332> The EU-15 countries, with the exception of Italy, are all overrepresented with respect to imports; and with the exception of Greece, Italy, and Spain all EU-15 countries are overrepresented with respect to imports. Related to the finding with the United States, the larger EU-15 countries (such as France, Germany, and the United Kingdom) have lower magnitudes of the L Q T indices, whereas the smaller EU-15 countries (such as 41 Belgium-Luxembourg, Ireland, and the Netherlands) have the highest magnitude of the L Q T indices. The E A - A N Z countries are far less consistent. The economically largest country in E A - A N Z , Japan, is highly underrepresented for both imports and exports in all years. But one of the smallest countries, Myanmar, is also highly underrepresented in exports and imports for all years. Notably high representation in exports and imports are Indonesia, (exports only), Malaysia, Singapore, and Vietnam (only in 2001). The correlations for the L Q T indices over time and between exports and imports are all positive and significance at the one percent level—though qualitatively similar to the parametric correlation coefficients, the nonparametric results are presented here due to concerns regarding the normality of the data. Not surprisingly, due to substantial differences in Figures 2.1 - 2.3, the Spearman's correlation coefficients decrease in magnitude over time for both imports and exports. For imports, the correlation coefficients range from a high of 0.773 to a low of 0.610, and for exports the high is 0.683 and the low is 0.656. Perhaps more interesting is that the correlations between the L Q T indices for imports and exports of the same year also fall over time: decreasing from 0.617 to 0.546. Given the relatively high correlation coefficients, one measure of representation is still a good predictor of the other, the similarity of import and export representation is decreasing over time. Though instructive on its own to show the specific geography of international trade not performed in previous research, the above analysis tells us nothing of the existence or establishment of the global triad. As such, the analysis now turns to the establishment of bilateral and regional trading relationships. 42 2.4.2 The Trade Location Quotient and Bilateral/Regional Trading Relationships As shown in Table 2.3, using the methodology outlined above, relatively few reciprocal bilateral trading relationships are found—10.0, 10.2, and 11.2 percent of all 11 476 possible trading relationships for 1981, 1991, and 2001, respectively. Of course, by definition, there cannot be 11 476 reciprocal bilateral trading relationships—this can only occur if all countries traded with all other countries proportionally more than their gross domestic product—but given the attention to globalization and the increases in international trade, this number is surprisingly low. As a percentage of reciprocal bilateral trading relationships, the global triad international trade dominates overwhelmingly. Very little trade flows between the three regions of the global triad, but a vast majority of trade occurs within each of those three regions. There has been a sharp increase in international trade within the three regions of the global triad from 1981 -1991, but little has changed since then. The remaining dominant portion of international trade flows from the global triad to countries outside of the global triad, with very little trade (less than 8 percent in all years) flows to and from non-triad countries. Once again, the global triad appears to be a dominant force in the international economy, and contrary to the findings of O'Loughlin and Anselin (1996), international trade is increasingly clustered around the three regions of the triad. <See Table 2.3, page 229> This view of international trade is a very limited one, however. As mentioned above, if the point of departure is the existence of the global triad—with each triad region centred on one of the three largest economies in the world—it should come as no surprise that the global triad dominates in international trade. Having the reciprocal bilateral 43 trading relationships as the point of departure and assigning countries to regions using the methodology described above, an entirely different view of the international relationships become manifest. <See Table 2.4, page 230> Table 2.4 summarizes the results from the regional assignments. The results clearly indicate no evidence of a global triad, based on reciprocal trading relationships—a result much closer to the spatial statistical results of O'Loughlin and Anselin (1996) than the movement toward a global triad found in Poon (1997) and Poon et al. (2000). With there being 24, 24, and 26 reciprocal trading regions in 1981, 1991, and 2001, respectively, and over 100 countries involved in these reciprocal trading regions, the average number of countries per region remains relatively constant over the entire study period. Additionally, there are very few bilateral trading relationships operating outside of the reciprocal trading regions, leaving approximately 20 countries without any reciprocal trading partners. The reciprocal trading regions are shown in Tables 2.5, 2.6, and 2.7. Overall, geography and history clearly play the dominant roles in regional assignment. Particularly in 1981 (see Table 2.5), colonial patterns are evident for France, Portugal, Spain, and the United Kingdom. France (Reunion), Portugal (Angola and Guinea-Bissau), Spain (Cuba, Equatorial Guinea, and Mexico), and the United Kingdom (Cyprus, Saint Kitts and Nevis, and Sierra Leone) all have reciprocal trading regions in 1981 based on their colonial past. Though most of these formal colonial ties were broken only between 10 and 20 years prior to 1981, Cuba and Mexico had been independent from Spain respectively for almost 100 and 200 years. Clearly, historical ties remained 44 important at least until 1981 for some countries with regard to international trade flows. Evident in all years, however, is that geographically close countries tend to have reciprocal trading relationships and geographically distant countries do not. This phenomenon, though present in most of the reciprocal trading regions is particularly present in Central and South America. The Central American countries and their geographically close Caribbean countries form one region, the more northern Caribbean countries form a region, the northern countries of South America and their geographically close Caribbean countries form a region, and the central and southern countries in South America form another region. Clearly, geographic distance is a powerful force in the establishment of reciprocal trading regions. Though a very intuitive outcome for some, this result shows that the forces of globalization by and large are bound by geographical relations, not only encompassing distance and proximity, but geographically defined relations involving social, political, and institutional aspects of each country with many of these relations seemingly related to former colonial relations. <See Table 2.5, page 231 > Turning to the regional assignments in the specific years, there is significant spatial reorganization occurring within some of the reciprocal trading regions despite little change in their numbers. North America experienced the least change over the study period. Canada and the United States are always a region—the largest in magnitude reciprocal bilateral trade flow in all years—with Mexico joining the North American region by 1991 (see Table 2.6). This joining of Mexico with Canada and the United States (though primarily with the United States) precedes the N A F T A by three years and is only one year after the United States and Mexico began to move toward a 45 free trade agreements of their own (Cameron and Tbmlin 2000). Therefore, the integration of these economies had begun years before any formal institutionalization had taken place. The European countries have experienced substantial change over the 20 year study period. In 1981, the European-based reciprocal trading regions were centred on Germany, the United Kingdom, and Norway, with a few colonial ties (see Table 2.5). Spain and Portugal, although they did not join what was to become the European Union until 1986, each had no ties back to Europe, or each other, aside from Portugal-Iceland. Italy, a founding member of the European Union, had no reciprocal trading relationships with any other European countries. But unlike Spain and Portugal that have their 1981 trading relationships defined through former colonial ties, Italy has trading relationships defined geographically by its larger region: the Eastern Mediterranean, the Middle East, and Northern Africa. Curiously, in the twenty-five years since the beginnings of the European Union, little integration has actually taken place. <See Table 2.6, page 232> By 1991, the European countries began to show more reciprocal trading relationships (see Table 2.6). The regions centred on Germany, Norway, and the United Kingdom changed little, but Spain and Italy shared a common region as did Austria and Switzerland with Hungary and the former Czechoslovakia—the latter country joining the European Union in 2004. Portugal continues to remain outside of any European region, and Greece, a member of the European Union since 1981, is in a region that is mainly composed of its immediate neighbours, none of which are part of the European Union. In 2001, Italy, Portugal, and Spain are all common to a region, and Greece emerged in a 46 region with Cyprus that joined the European Union in 2004. Needless to say, Europe does not show itself to be a region of any global triad, but even in 2001 has six reciprocal trading regions involving European Union countries (see Tabje 2.7). One may interpret this set of reciprocal trading regions involving European Union countries as not challenging the global triad thesis, but supporting it. Perhaps not all European countries trade intensely with each other, but the six reciprocal trading regions in 2001 encompass the European countries and, therefore, show the European triad region. This interpretation, however, would be premature. Many of the countries involved in the European reciprocal trading regions are outside of the European Union, and there is little evidence of that pattern decreasing over time. Many of the non-European Union countries in the years of regional assignment have now become members of the European Union (the CEEC-10), but these are countries traditionally considered distinct from the European triad region by those that speak of the existence of the global triad. East Asia is also undergoing significant adjustment. The first phenomenon to note is that South Pacific countries (namely Australia and New Zealand, with respect to the above analysis) should be grouped together with East Asia cautiously. Aside from 1991 (see Table 2.6), there is no overlap between East Asian and South Pacific countries, but the South Pacific countries are small in a global context so any chances of bias are small. Japan, the largest country in East Asia (3rd largest in the world), only has a reciprocal trading relationship with Indonesia. However, Japan's relationship with Indonesia is dominated by petroleum products. Therefore, this international trade relationship is likely related more to its ties with oil-producing nations for natural 47 resources, than a relationship based on the global triad thesis. South Korea has a similar relationship in 1981 and 1991 with petroleum-producing countries, but by 2001 does settle into an East Asian trading relationship with China and Mongolia (see Table 2.7). The most pronounced regional grouping in East Asia involves the A S E A N countries. In 1981, five of what was to become the A S E A N - 1 0 were in a reciprocal trading region, and by 2001 that number had increased to seven - other non-ASEAN countries are also in this reciprocal trading region making it the largest in East Asia (see Tables 2.5, 2.6, and 2.7). By 2001 there appears to be both a North- and a South-East Asian region, but neither of them includes Japan. Of course, Japan was the first country in East Asia to experience rapid industrialization and to establish relationships outside of East Asia with other industrialized countries. Indonesia, a large component of the A S E A N , lies outside of these reciprocal trading regions due to its being a petroleum-producing nation. And China, now a formidable presence in the global economy, once part of a reciprocal trading region involving a number of the A S E A N countries, is involved in a reciprocal trading region with only South Korea and Mongolia by 2001. This result is a prime example of a country with many large trade volumes that flow in one direction, and few reciprocal trading relationships. <See Table 2.7, page 233> Though not included as one of the three regions in the global triad, Central America, South America, and the Caribbean are some of the largest reciprocal trading regions—Central America being the largest. In 1981, South America was separated into three regions, and both Central America and the Caribbean formed a fourth region (see Table 2.5). In 1991, there were the same number of reciprocal trading regions, but the 48 countries had spatially reorganized (see Table 2.6). Part of this spatial reorganization involved the Mercosur countries plus their associated members (Bolivia and Chile) forming a reciprocal trading region of their own. Incidentally, 1991 was the year Mercosur was established and the remaining countries in this area became much more tightly clustered in space, forming a Central American region, a northern South American region, and a Caribbean region. By 2001, South America had essentially been split into a northern and southern region (see Table 2.7). And by that same year, the Caribbean and Central America formed the largest reciprocal trading region in the world in terms of the number of members. The former Soviet Union and former Eastern European Socialist countries form a reciprocal trading region in 1981, but by 1991 that region began to break apart (see Tables 2.5 and 2.6). Hungary and the former Czechoslovakia form a region with Austria and Switzerland, whereas the former Yugoslavia joined Greece's region. Both of these moves are likely due to history (Austria-Hungary) as well as geographical proximity. However, by 2001 the former Yugoslavia joined the reciprocal trading region centred on Austria, leaving Poland as the only country maintaining a reciprocal trading relationship with the former Soviet Union (see Table 2.7). Greater detailed bilateral trade data that separates all of the former Soviet Union may prove to be instructive in Europe's spatial reorganization of international economic relations, particularly with the Central and Eastern European countries (CEEC-10) joining the European Union in 2004. One of the most dynamic areas with regard to changing reciprocal trading relationships over the study period has been Africa. In 1981, the eastern coast of Africa was in a reciprocal trading region, two other trading regions on the western coast and the 49 island nations off the south eastern coast formed a fourth reciprocal trading region (see Table 2.5). By 1991, the reciprocal trading region on the eastern coast all but disappeared, while the trading regions on the western coast (though losing some members) solidified into one trading region, the trading region off the south eastern coast expanded, and South Africa emerged as a reciprocal trading region from a bilateral trading relationship (see Table 2.6). Ten years later much had changed again. The reciprocal trading region centred on South Africa grew substantially since 1991, likely due to the abolition of apartheid and the accompanying trade liberalization. The eastern African trading region has re-emerged with another smaller trading region to the north, and the reciprocal trading regions on the north western coast have largely split into eastern and western reciprocal trading regions. No doubt, as the African economies continue to advance and political conditions continue to stabilize spatial reorganization will also continue. And finally, we turn to the Middle East and South Asia. Very little regionalization materialized in these two areas in 1981. One reciprocal trading region emerged centred on India, but the development of reciprocal trading regions had not occurred until 1991 (see Table 2.6). By that year, in addition to a reciprocal trading region centred on India, the majority of the small independent states in the Middle East formed a trading region centred on Pakistan. In 2001, there were four reciprocal trading regions in the Middle East, all centred on petroleum-producing nations and India (see Table 2.7). Though most of these countries are not considered a region—outside of the Organization of Petroleum Exporting Countries—this area's dynamic likely changed 50 shortly after the 11 September 2001 terrorist attacks and the U.S. presence in the Middle East. 2.4.3. Sensitivity Analysis The critical threshold value of 1.30 used to establish reciprocal bilateral trading relationships is admittedly arbitrary. A 30 percent overrepresentation in international trade flows is unlikely to be considered too low a threshold but it may be considered too high. As such, the critical threshold values are lowered to 1.20 and 1.10 for a sensitivity analysis. Tables 2.8 - 2.13 show the regional assignments for 1981, 1991, and 2001 using both 1.20 and 1.10 as critical threshold values. <See Tables 2.8 - 2.13, pages 234 - 239> Curiously, changing the critical threshold values barely changes the results. Overall, in all three years and all critical threshold values, there is some minor re-shifting of countries between reciprocal trading regions in Central America, South America, and East Asia. Europe does undergo some re-shifting of countries as well, but the European Union does not emerge as v vFortress Europe" even when the critical threshold value is 1.10. The most striking consistency is in the overall nature of the reciprocal trading regions themselves (Table 2.14). <See Tables 2.14, page 240> Using lower critical threshold values to establish reciprocal bilateral and regional trading relationships one would expect more countries to be involved in reciprocal trading regions and the that the reciprocal trading regions themselves would have more country members. This, however, is not the case. As shown in Table 2.14, there are essentially no changes in the total number of countries involved in reciprocal trading 51 regions (120 - 130) or in the average number of countries per region, approximately five. With such small change in both the number of countries within the reciprocal trading regions, and the nature of the reciprocal trading regions themselves, all of the analyses and interpretations discussed above are considered robust. 2.5. S U M M A R Y A N D C O N C L U S I O N S This chapter, using a new metric to measure the spatial relationship of international trade that controls for country size has shown that there is a geography of international trade distinct from the geography of production. By 2001, the geography of international trade indicates that the three regions of the global triad engage in intense interactions, but they are clearly not the only areas of the world with a high intensity of international trade. Central America, South America, and parts of Africa also exhibit very strong international trade intensities, higher in magnitude than many countries in the three regions of the global triad, though not in terms of volume. Extending the international trade location quotient into reciprocal bilateral and regional trading relationships finds absolutely no evidence of the existence or movement toward the global triad. Instead, reciprocal trading relationships are governed quite strictly by geographical space and, to some extent in the early years of the study, by history (i.e. colonialism). Additionally, the number of countries per reciprocal trading region remains relatively constant over time indicating no increased clustering of regionalization from 1981 - 2001. With respect to reciprocal trading regions and international trading agreements, there are three reciprocal trading regions that correspond well to their respective 52 international trading agreements by 2001. The N A F T A countries are involved in a reciprocal trading region by 1991, three years before the N A F T A entered into force—the United States and Canada were already involved in a free trade agreement in 1989, however. And this reciprocal trading region is the largest in magnitude in the world despite there only being three members. The Canada - U.S. and U.S. - Mexico bilateral trading relationships are the two largest international trade flows in the world since 1991, dwarfing most other bilateral trade flows by orders of magnitude. The Mercosur countries and associated members (Bolivia and Chile) form the second international trading agreement that corresponds to reciprocal trading regions. This region is one of the largest economic entities in the world and has clearly been successful in establishing and maintaining reciprocal trading relationships since its inception. And contrary to previous research, this region has emerged as a trading region independent of its North American neighbours. Though all of the A S E A N - 1 0 countries are not represented in a single region by 2001, the majority of the member countries are involved in a reciprocal trading region (see Table 2.7). A north eastern reciprocal trading region has also emerged by 2001, but Japan is essentially absent from East Asia with respect to reciprocal bilateral trading relationships. The only exception to this result is Japan and Indonesia. But as discussed above, this is not surprising given Japan's relatively poor abundance of energy resources. Overall, this analysis has shown that the geography of international trade not only exists but is incredibly dynamic. Political, historical, and economic forces continue to change, leading to a spatial reorganization of the international economy. Though some areas of the world appear to be stabilizing with respect to reciprocal trading regions, 53 many areas are undergoing significant spatial reorganization. Additionally, areas commonly grouped together, such as the European Union member countries, are shown to not be nearly as integrated through trade as other areas such as the NAFTA and Mercosur countries. The European Union has also recently expanded into Central and Eastern Europe, potentially altering the geography of international trade in that area. With the establishment of the NAFTA as the top reciprocal trading region in the world, the analysis now moves to a more detailed study of the top bilateral trading relationship in the world operating within the NAFTA—Canada and the United States. Chapter 3 54 A history of Canada - United States trade relations 3.1. INTRODUCTION Canada and the United States not only have the top bilateral trading relationship in the world, based on international trade volumes, but they were found in the previous chapter to be part of the most significant reciprocal trading region in the world—by 2001 the North American reciprocal trading region contained the largest flows of international trade in the world. This trading relationship, however, is defined through formal institutionalization (free trade agreements) at both the industry and national level. It is the purpose of this chapter to outline this formal institutionalization. The chapter is organized as follows. The next section briefly outlines early attempts at free trade between Canada and the United States. Section 3 outlines the Canada - United States Automotive Products Agreement of 1965. Section 4 reviews the socio-economic and political context of the free trade agreements, the free trade agreements themselves, and the empirical studies investigating the effects of these free trade agreements. Section 5 summarizes and concludes. 3.2. EARLY ATTEMPTS AT TRADE LIBERALIZATION The Canada - United States international trading relationship began in the middle of the nineteenth century before Canada was even an independent nation. With the United 55 Kingdom bargaining on behalf of Canada in 1854, the Elgin - Marcy Reciprocity Treaty (EMRT) was signed and entered into force between the United States and the colonies of Upper and Lower Canada. These colonies included the contemporary provinces of Ontario, Quebec, New Brunswick, Nova Scotia, PEI, and Newfoundland, which among themselves had entered into a free trade agreement in 1850 (Crookwell 1990). This free trade agreement was established for two primary reasons: first, the economic depression in the 1840s led the United Kingdom to fear that the provinces would move toward a union with the United States; and second, the Agreement was expected to resolve a North Atlantic fishing rights dispute with the United States. In order to mitigate the potential merging of the Canadian colonies with the United States, the United Kingdom entered into negotiations with the United States for a trading agreement. The hope of the United Kingdom was to provide the Canadian colonies with the benefits of access to the U.S. market, without severing their ties to the United Kingdom. As a colony of the United Kingdom, the Canadian colonies were directly linked to the economy of the United Kingdom, providing valuable primary resources. If the Canadian colonies formed a union with the United States this valuable economic link would be at risk, however. This issue, along with the need for negotiations regarding North Atlantic fishing rights prompted the negotiation of free trade so the United Kingdom could maintain its colonial presence in North America (Crookwell 1990). The E M R T was concerned with only natural resources and agricultural products, but as a portion it represented two-thirds of all Canadian merchandise trade at the middle of the nineteenth century. During the tenure of the E M R T , the issue of a customs union 56 emerged repeatedly, predominantly urged by the United States. But the Canadian colonial tie to the United Kingdom was a major stumbling block to any such union. Despite its popularity, the E M R T was terminated by the U.S. Senate in 1866. This was partly because of the United Kingdom's support for Confederates during the United States Civ i l War, and partly because of the general rise of protectionism (Crookwell 1990; Fry 1987; Stevens 1987; Velk and Riggs 1987). Canada repeatedly tried to re-establish the treaty without success. But when the United States moved to re-establish the treaty in the 1880s, Canada had begun its own protectionism as it strove to develop its own indigenous manufacturing industries rather than fostering international trade (Crookwell 1990). Within the Colonial Conference of 1907, having the goal of stimulating international trading agreements with other countries, the Canadian government reconfigured its tariff structures with the world. This reconfiguration instituted a new tariff level for all countries with which Canada did not have an international agreement (bar the United Kingdom). Canada had some success in establishing international trading agreement with France, Italy, and Japan during this time. But because of the colonial nature of the conference, the United States was not included in any of Canada's international trading agreements. Consequently, the United States accused Canada of discriminatory trading practices, setting off a tariff war between the two countries. In order to resolve the dispute, the two nations entered into negotiations for an international trading agreement of their own, covering tariff and non-tariff barriers, to be established in 1911, after a Canadian election. The Conservative Party that won the 1911 election never put the agreement to a vote, although the Agreement was passed in the United States. 57 The Canadian public view was that a free trade agreement with the United States would threaten Canadian nationality and the Canadian connection to the United Kingdom (Stevens 1987). This was despite the fact that the 1911 agreement was more modest than the E M R T , though broader in scope (Crookwell 1990; Fry 1987). The 1930s saw two trade agreements between Canada and the United States (1935 and 1938). Both these agreements were largely negotiated to counter the extreme protectionism that both countries were practicing in the early 1930s following the Great Depression, rather than a move towards free trade as such (Fry 1987). The extreme protectionism that arose in the 1930s was the result of trying to insulate local economies from the oscillations in the global economy. It was believed that the effects of the Great Depression could be mitigated by trade protectionism that isolated the national economy from the troubled international economy outside (Fry 1987). Although there was a failed attempt at free trade negotiations with the United States in 1948, Canada's international trading relationships in the immediate post-war years focussed on the General Agreement on Tariffs and Trade (GATT). The G A T T was established in the wake of the Second World War (1947) to foster a multilateral trading system. Initially, the G A T T mainly dealt with tariff reductions, but later moved into non-tariff barriers to trade, as well as rules of international competition (WTO 2006). From 1947 through to the mid-1960s, Canada was involved in the six G A T T negotiations on tariff reductions. Only when industry crisis emerged in the mid-1960s did Canada's international trade attention turn from global free trade to the possibility of free trade with the United States. This occurred through negotiations of the Auto Pact. 58 3.3. THE AUTO PACT 3.3.1. Automotive Trade Before the Auto Pact The automobile industry in Canada had modest beginnings in 1904 when the Ford Motor Company of Canada came into being. In order to protect the new industry from international competition, Canada imposed a tariff rate of 35 percent on the United States (though this number appears to be arbitrarily high, it was inherited from the carriage building industry). Additionally, Great Britain was not immune, having a lower, though significant, 22.5 percent tariff imposed by Canada. These tariff rates were extremely high compared to tariff rates today, but they were comparable, or lower, than the rates imposed at the time by other automobile producing nations such as the United States, Great Britain, and France. Though there were small changes, the tariff rate scheme for the industry was completely overhauled in 1936, and these tariffs remained fundamentally unchanged until the 1960s (Bladen 1961). The new scheme granted tariff free status to automobiles and parts originating from Great Britain and a 17.5 percent tariff for other countries, including the United States. Some Canadian automotive production sectors deemed sensitive to import competition, however, were protected by a 22.5 percent tariff rate for countries aside from Great Britain (Wonriacott 1965). While the Second World War disrupted demand in general, once hostilities ceased the pre-war pattern resumed in which Canada and the United States essentially possessed two separate automotive operations although dominated by the same firms. With high tariffs still holding, Canada in effect had a miniature version of the automotive industry in the United States (Holmes 1983; Holmes 1992). General Motors, Ford, and Chrysler (hereafter referred to as B ig Three) invested in Canada's industry because of a potentially 59 lucrative market, despite inefficient production levels resulting from the high degree of protection and a small market size. The Big Three all found this satisfactory: firms had an oligopolistic market structure and high profits; automotive workers' unions and automotive parts producers did not compete internationally; and the Canadian government had a national industry. The losers, however, were the consumers who paid high prices for finished automobiles (Nelson 1996). However, the market would not stay lucrative for long. Because of the high degree of protection and corresponding lack of investment, the Canadian automobile industry's international competitiveness faltered during the 1950s. Though the Canadian industry was protected, its productivity was too low to compete with imports, even after the tariff was considered. Without investment in Canada because of its small market and lack of sufficient investment returns, the productivity of Canadian automotive producers fell significantly behind that of other countries—high unit costs in Canada were also due to short production runs. Simultaneously, and most likely because of the rapid economic expansion in the post-war period, strong domestic demand forced Canada increasingly to import automobiles, despite the high tariffs (Holmes 1993). Even with the high tariffs imposed on outside automobile manufacturers selling to Canada, they were still able to compete. The result was that even though domestic demand for automobiles was rising, Canadian output and employment in the automobile sector fell: from 1955 - 1960, output fell from 375 000 to 326 000 units and employment fell from 53 000 to 42 000 (Holmes 1983). For the first time since its inception, the Canadian automobile industry was facing international competition. Because of the importance of the automobile industry to 60 Canadian employment, income, and international trade, political pressure mounted for a Royal Commission to investigate the competitive position of the Canadian automobile industry. The result was Report: Royal Commission on the Automotive Industry, hereafter referred to as the Bladen Report. The Bladen Report made seven recommendations for the automotive industry. They were based on the assumption that it was necessary for the number of firms within the industry to decline in order to increase the length of production runs. In turn, this would allow such firms to take advantage of economies of scale and reduce their unit cost, becoming more internationally competitive. The first six of these recommendations involved changes in the automotive industry at both the domestic (sales and excise taxes) and international levels (tariff rate changes). Although the first six recommendations were important, the most important and original recommendation was the "extended content plan" (Johnson 1963: 212). This plan was meant to increase the level and efficiency of protection for the automotive industry through a balance of free trade and protectionism. As the name of the recommendation suggests, the (Canadian) content of the imports determined whether or not the item imported fell under free trade or protectionism. If sufficient Canadian content was present in the imported item, it would receive duty-free status. If the Canadian content requirement was not satisfied, the standard customs tariff would apply. The content requirement was "extended" because it did not matter where the automobile or automotive parts were produced, only that a certain portion of the final factory cost was produced in Canada. The purpose of this aspect of the recommendation was to promote Canadian exports of automotive parts that 61 would be used in final automotive product imports (Bladen 1961). Following this Report, important changes to the automotive industry were implemented with positive results for Canada., In Canada, both employment and output increased, pleasing the Canadian final assembly firms and their labour unions. In the United States, however, automotive parts manufacturers complained about the new Canadian competition. These firms lobbied the United States government for a countervailing duty (Nelson 1996). The upshot was that both governments began negotiations for an automotive trade agreement: the Canada-United States Automotive Products Trade Agreement of 1965. 3.3.2. The Auto Pact The Auto Pact was a ""managed trade agreement that permits the duty-free movement of automotive parts and assembled vehicles between the two countries [Canada and the United States] subject to certain safeguards" (Holmes 2000, 651). The result was the Big Three were able to rationalize and integrate the production of automobiles and their parts within a single tariff-free market. Firms that qualified under the Auto Pact were not restricted to the Big Three; there were over one hundred such firms, such as Volvo Canada and American Motors (Canada) Ltd., but the dominant players were the Big Three. The Auto Pact did far more than simply create a tariff-free region. The safeguards in the agreement specifically protected automotive production in Canada. The United States granted duty-free access for Canadian automobiles and automotive parts provided they met a 50 percent North American content requirement—only original equipment parts were included, not replacement parts. Canada, on the other hand, granted 62 duty-free access to the United States on the condition that a certain proportion of production and value added was guaranteed to take place in Canada. Specifically, American firms needed to achieve a certain level of production to domestic sales ratio (in the case of passenger cars, the required level is more than 100 percent); the amount of value-added for automotive production in Canada must be at least the value achieved in 1964; in each year, automotive production was to achieve a certain level of Canadian value added ( C V A ) , which is not limited to the value of Canadian car components, but includes services supplied in Canada (generally, more than 60 percent of the production is required); and finally, there would be a one time increase in the Canadian value added to occur over a three year period (Fuss and Waverman 1992). These production safeguards clearly distinguish the Auto Pact from a free trade agreement, hence my use of the more restrictive term tariff-free region, above. These stipulations were viewed as quite onerous by the United States. In fact, the United States argued that they should be considered temporary and phased out over time. But, no such phasing out was agreed upon. Despite these concerns by the United States, restrictions on the automotive industry have not been binding. In fact, the production to sales ratio had generally increased from approximately 100 percent in 1980 to almost 250 percent by 1996—the latest available data. The Canadian value-added requirement, though more volatile than the production to sales ratio, has also remained well above its minimum level. It should be noted, however, that the only stipulation in the Auto Pact were the minimum threshold levels. These thresholds could be exceeded, but any increases over the minimum levels were riot to become binding. 63 International trade in the automotive industry for Canada saw significant expansion from the mid-1960s to the mid-1970s, and is continuing today. The particular nature of the international trade flow expansions favoured exports over imports, improving Canada's international trade balance with the United States, and the world. With the implementation of the Auto Pact, the Canadian industry experienced a significant rationalization of production. The number of models produced in Canada, and specifically within each firm, drastically reduced in order to take advantage of economies of scale. As a result, the Canadian automotive industry became relatively competitive and efficient. Despite this competitive advantage, however, the Canadian industry was limited (Holmes 1996). The government provision of medical health care, and later the relatively low value of the Canadian dollar, provided Canada with a relatively inexpensive labour force (15 - 20 percent labour costs savings over the United States). Consequently, Canada attracted a disproportionately large share of labour intensive production such as final assembly and particular labour intensive automotive parts, whereas the high value added production of body stampings, engines, and transmissions were located in the United States. This geographical division of labour of operations led to a ""distinctive pattern of trade between Canada and the United States" (Holmes 1993, 26): an international trade surplus in automobiles and an international trade deficit in automotive parts for Canada. The decision making processes were also geographically segregated because the head offices of the Big Three were all in the United States. Despite Canada's position in the geographical division of production in the automobile industry, this first move toward integration with the United States through 64 free trade agreements have been a success (Holmes 1996). Production, employment, imports, and exports have all expanded significantly since the inception of the Auto Pact, increasing the wealth of the Canadian economy. These expansions dominated in particular regions, namely central Canada, and Southern Ontario especially. However, the development of economic instabilities beyond that of the automobile industry led Canada to push for a free trade agreement at the national level with the United States rather than being circumscribed to one (albeit important) industry. 3.4. FREE TRADE IN NORTH AMERICA 3.4.1. Why Free Trade? Why Now? The early-1980s were a time of great socio-economic and political uncertainty for Canada. Four factors were important. First, the 1981 - 1982 recession with its corresponding high interest rates, high unemployment, and particularly important for Canada, volatility in commodity prices. By the early 1980s, Canada - Untied States international trade flows were the largest international trade flow in the world, in large part because of the Auto Pact. Consequently, 70 percent of Canadian exports and imports were sent to and from the United States which meant that almost 20 percent of the Canadian economy was directly tied to the United States through international trade flows—a great deal of Canadian economic dependency on the United States (International Monetary Fund 2005a, 2005b). And because of this dependency, the need for open and assured access to the U.S. market for exports was reinforced. Second, during the recession, the United States began to move toward protectionist trade policies. United States' industries that were considered vulnerable to 65 international competition fell under protectionist policies call "trade remedy" laws. Many of these industries were particularly important for Canadian exports such as softwoodiumber, fish, pork, and steel. Though there was no immediate threat for any of these exports at the time, Canada clearly needed to limit the scope of any unilateral trade restrictions made by the U.S. policy-makers. Third, on the international front, there was a perceived decline in the G A T T ' s ability to regulate world trade effectively as well as protecting Canada's international trade interests. Specifically, the Tokyo Round of the G A T T had failed to advance trade rules to foster international trade, and most of the "trade remedy" laws against Canadian exports were consistent with the G A T T through an escape clause that allowed G A T T members to temporarily escape from their G A T T responsibilities i f domestic industries were suffering serious injury from import competition. And finally, the emergence of new international competition in world markets because of multilateralism that was successful through the G A T T forced Canada to consider its place in a globally competitive world (Smith and Stone 1987). Canada may have had a strong economic relationship with the United States, but its exports were threatened by new competition from East Asia. As with the automotive sector in the 1950s and 1960s, Canada's manufacturing sector was partially confined to its small domestic market. Therefore, Canada needed to expand its free trade relationship with the United States far beyond the narrow, though significant, scope of the automotive industry. This would potentially increase economies of scale and encourage the rationalization of many industries in order to benefit them as the automotive industry benefited from these same changes years before. Therefore, Canada either had to become 66 more global or tie itself more directly to the United States. The Canadian Conservative government at the time chose the latter. In 1985, in hopes of reducing Canada's socio-economic and political uncertainties, the Canadian government requested a comprehensive free trade agreement with the United States. The request was made despite the fact that in 1985, 85 percent of all Canadian exports to the United States crossed the border duty free, with the remaining average (value-based) tariff rate being 4 percent—these zero and low tariff rates were in large part due to the Canada - United States Automotive Products Agreement of 1965 and Canada's dedication to the successive rounds of the G A T T negotiations. Given the high degree of economic dependence on the United States, Canada was vulnerable to unilateral action from the United States through foreign trade policy changes without the protection of the G A T T , and sought to minimize this vulnerability through a negotiated free trade agreement so Canadian exports to the U.S. economy would not be lost (Coffey etal. 1999). Thus, Canada's primary objectives in free trade negotiations were greater certainty in U.S. trade laws and market access to the United States through decreased tariff rates. The United States, on the other hand, was not faced with the socio-economic and political uncertainties of Canada. Consequently, the United States was more focused on rule-making in trade relations: no new trade barriers; agreements on services, investment, and intellectual property rights; and resolutions to long-standing trade disputes, particularly automotive subsidies to non-Auto Pact automotive producers in Canada. These issues were of prime concern for the United States because of its increasing trade in services, the desire to be able to invest without restrictions in Canada, 67 to protect the ever-increasing role of less tangible subject matters such as performances and digital media, and to mitigate international competition in the automotive industry that was channelled through Canadian investment incentives to Japanese automobile producers. In general, both countries hoped the C U F T A would further the multilateralism process by stimulating more free trade agreements, as well as functioning as a "fal lback" if the G A T T (WTO) process broke down. At the domestic level, both Canada and the United States also aimed for inter- and intra-industry specialization (the rationalization of production) in order for both countries to benefit from the Agreement. This rationalization would, in turn, increase the competitiveness of each country in a global context to promote export-led growth and reduce each country's trade imbalance with the rest of the world, particularly the United States (Schott 1991). Additionally, the rationalization of production would mitigate the threat of U.S. foreign trade policy changes, decreasing the need for the G A T T to intervene in Canada-U.S. trade relations. 3.4.2. The Canada-U.S. Free Trade Agreement At the time, the C U F T A was the most comprehensive free trade agreement negotiated and implemented between any two nations. For the first time, the C U F T A established a bilateral and contractual institutional base to manage a bilateral trade and economic relationship. The Agreement also introduced the concept of non-discrimination: a firm's nationality would not determine competitive bidding prospects (Hart 1989; Waverman 1991). And importantly, for the international trading regime and multilateralism in general, the C U F T A is consistent with the G A T T / W T O . Tariff barriers between Canada and the United States were negotiated to fall, but non-member tariffs remained 68 unchanged. Despite the obvious discrimination in tariff rates, no new barriers were created with the rest of the world, so the average tariff rate for the world fell (see Bhagwati (1999), for a critical discussion of this issue). In fact, because of increased efficiency and productivity through rationalization, and the Agreement countering the worldwide short-run protectionist trend resulting from global economic recession, some II argued that the C U F T A ' s " new agreements on dispute settlement, services, and investment...[could be a basis]...on which G A T T negotiators c[ould] build and develop better multilateral accords" (Schott 1991: 81). The C U F T A as a free trade agreement can be broadly classified into three conceptual categories: trade liberalization, rule-making, and standstill. Trade liberalization is probably the most "v i s ib le" portion of the C U F T A . Tariffs on most goods were to become zero within ten years—less progress was made with non-tariff barriers. More government contracts were to be opened to competitive bidding, which is the so-called national treatment of firms that states any firm from either Canada or the United States is to be treated without consideration of their nationality. With respect to rule-making, the C U F T A established: legal frameworks that allow businesses to operate in both Canada and the United States; a trade dispute resolution mechanism; investment policies; government policies that prevent national bias (similar to the national treatment of firms); and, cross-border travel by businesspersons to stimulate the trading and investment relationships between the two countries. Finally, standstill refers to aspects of the C U F T A that preclude any return to protectionist policies by stating that any hew restrictions and/or barriers cannot be greater in magnitude than before the Agreement. It should be noted that the C U F T A does not 69 prevent future attempts at protectionism, but requires that policies be transparent (Schott 1991). Despite these great strides forward in trade liberalization, rule-making, and standstill the C U F T A has significant limitations and comes nowhere near the level of trade/economic integration present in the European Union (Waverman 1991). Admittedly, the purpose of the C U F T A was not to attain the level of integration in the European Union, but to solidify an already existing trading relationship. Nevertheless, there was still much work on trade relations to be done: agriculture is only partially covered; the textiles industry is severely constrained, and beer is excluded from the Agreement completely; the liberalization of services and investment is quite limited; the C U F T A is not concerned with the liberalization of the factors of production, particularly labour mobility; and barriers to international trade with third parties are not covered (Waverman 1991). However, whether the C U F T A is judged by its great strides forward or its limitations, it is a document that has a significant role in the nature of the Canadian economy because of the high degree of economic interaction with the United States. 3.4.3. The Effects of the CUFTA Because the Canada - United States Free Trade Agreement (CUFTA) is still quite recent, research on the effects of this free trade agreement on the members' national economies is not in great abundance. However, research on the C U F T A has emerged in the past few years. One of the prime concerns regarding regional trade agreements, or regionalism, such as the C U F T A and the N A F T A , is that member countries will concentrate their international trading relationships solely with member countries, leading to decreased 70 multilateralism and decreased world welfare (Bhagwati 1999). With respect to the C U F T A , very little meaningful evidence is found to date to justify this concern. Clausing (2001) finds that there is substantial trade creation (intra-regional international trade growth) resulting from the C U F T A , with little evidence of any trade diversion. Supporting this claim, Krueger (2000) finds that not only is there little evidence of trade diversion in the aggregate, but this result holds across almost all commodity categories as intra-North American trade increased as a whole. Using a finer degree of data aggregation, Fukao et al. (2003) finds that there is significant evidence of trade diversion for U.S. international trade flows in textiles, apparel, and footwear products, particularly to Asia. Rather than previously importing these goods from Asia, they are now being sourced from N A F T A countries, predominantly Mexico. However, as pointed out by Coughlin and Wall (2003) and Wall (2003), a preferential trading area such as the C U F T A or the N A F T A may alter the spatial distribution of production in one of two related ways. First, with the addition of Mexico to the preferential trading area in North America, the mean centre of the N A F T A countries' consumers shifted south by a large degree. In order to minimize transportation costs to all potential consumers a firm may choose to relocate further south, from New Jersey or Ontario to Arizona, for example. In so doing, the firm alters its transportation costs to non-partner areas as well such as to Asia and Europe: reduced transportation costs to Asia and increased transportation costs to Europe—of course, this all depends on where the new centre is located. This adjustment, in turn, may alter the sourcing of imports giving the illusion of trade diverting behaviour. And in the context of Canada and the United States, the effect of this spatial re-distribution of firms is only amplified i f 71 the focus is the provinces and states. So, even i f non-partner country international trade does not change, there will be the illusion of trade diverting behaviour at the subnational level (provinces and states) (Coughlin and Wall 2003; Wall 2003). Secondly, the firm may relocate directly into Mexico without being punished by tariff barriers. This alters the flows of commodities potentially turning exports from the United States to imports to the United States from Mexico. If the United States initially imported that commodity from outside of the N A F T A and then exclusively imported its " o w n " commodity from Mexico after the N A F T A due to a lower cost, trade diversion wil l also appear to be taking place (Coughlin and Wall 2003; Wall 2003). Despite the geographical complications to measure and assess trade diversion it is commonly investigated in studies on the effect of the N A F T A (see Fujita et al. 1999, Hanson 1996, 1998a, 1998b, Krugman 1998, and Krugman and Hanson 1994). The phenomenon of measuring trade creation and trade diversion is also particularly curious because neither the C U F T A nor the N A F T A stipulated any changes in non-member tariff rates. In fact, tariffs between the N A F T A countries and the rest of the world continue to fall through the trade negotiations within and through the World Trade Organization (WTO 2006). Therefore, the appearance of any trade diversion is likely due to the geographical complications in measuring trade diversion outlined above. As such, the investigation of trade diversion is not undertaken in the present analysis. With regard to the expected rationalization of production resulting from decreased trade barriers (Schott 1991), Head and Ries (1999) find that Canadian manufacturing output per plant has increased dramatically (about 34 percent) and the number of plants has decreased dramatically (about 21 percent) during the six years after the C U F T A came 72 into effect. However, the C U F T A is only partially responsible. The substantial rationalization in Canadian manufacturing is partially due to the measurement error on the part of Statistics Canada and also because of industrial re-organization and the reduction of (U.S.) tariff rates. Consequently, the C U F T A was only found to have been responsible for less than 10 percent of the increased labour productivity in Canadian manufacturing firms. Overall, the measured effect of the C U F T A on Canada is positive at the national level. Trefler (2004) notes that manufacturing output and employment decreased in the years following the C U F T A , but it is difficult to assess the independent effect of the C U F T A because of the recession that ensued at the same time it entered into force (Gaston and Trefler 1997). International trade flows, on the other hand, have increased in magnitude much greater than expected, with more than one-half of the international trade flow increases attributable to the C U F T A (Clausing 2001; Schwanen 1997). The industrial sectors that achieved the most significant tariff decreases typically had the greatest growth in international trade flows. And for these industries that were substantially impacted by the C U F T A , the tariff cuts not only explain almost all of Canada's increased international trade flows with the United States, but also the increased share of the United States in Canadian international trade flows (Clausing 2001; Schwanen 1997; Trefler 2004). Not surprisingly, Canada experienced unprecedented import and export expansion during the 1990s that cannot be explained without considering the C U F T A . 3.4.4. Another Free Trade Agreement in North America? 73 No sooner did the C U F T A enter into force then negotiations began over another free trade agreement in North America. In March 1990, the United States and the United Mexican States (Mexico) announced their intention to establish a free trade agreement of their own (Cameron and Tomlin 2000). Superficially, the agreement was of little concern to Canada. In 1993, the year before Mexico entered into a free trade agreement with the United States, the vast majority of Canada's 1993 international trade flows were with the United States (exports = 79.8 percent, imports = 67 percent), followed by East Asia (exports = 9.0 percent, imports = 14.8 percent), the European Union (exports = 6.2 percent, imports = 9.7 percent), and finally Mexico (exports = 0.5 percent, imports = 2.3 percent). Trade access with Mexico was simply not an issue for Canada. However, access to the U.S. economy was important. If the United States and Mexico established a comprehensive free trade agreement, similar to that of the C U F T A , Canada's access to the U.S. economy may have been threatened. Just as the C U F T A had the potential to be trade diverting for non-CUFTA countries, so could the United States -Mexico free trade agreement. Mexico had a significant labour cost advantage over Canada that would only be increased through efficiency and productivity gains resulting from the rationalization of production (Cadsby and Woodside 1993; Hart 1989; Waverman 1993; Weintraub 1991). Consequently, in January 1991, Canada sought negotiation status with the United States and Mexico to ensure access to the U.S. market yet again. Rather than having foreign trade policies of other nations dictate Canada's role in the global economy, Canada played an active role in those foreign policies potentially achieving further gains 74 in efficiency and productivity from scale economies and rationalization, improving Canada's overall position in the global economy. (Coffey et al. 1999; Whalley 1993). 3.4.5 The North American Free Trade Agreement The result of the negotiations between Canada, Mexico, and the United States was the North American Free Trade Agreement ( N A F T A ) . The document is over 1000 pages long and was negotiated during a recession among very unequal (in terms of economic power) nations. Historically speaking, the N A F T A became the most comprehensive free trade agreement negotiated between regional trading partners, superseding the C U F T A , and is the first free trade agreement negotiated between an industrializing country and industrialized countries (Hufbauer and Schott 1993). Moreover, the N A F T A extended the scope of the C U F T A with respect to Canada - United States international trading relations. The negotiations of the C U F T A went rather unnoticed by the U.S. public, with lively debate on the Agreement in Canada—the C U F T A became an election issue in 1988. The N A F T A , on the other hand, was different. The Canadian public saw it as just another international trade agreement with the United States. The U.S. public saw it differently. The most remembered public debate on the N A F T A in the United States revolved around Ross Perot's attacks on the N A F T A and the "giant sucking sound" of U.S. jobs moving to the Mexican economy. For Canada, the N A F T A incorporated the C U F T A and allowed the Canada - U.S. negotiations to be re-opened on subsidies, countervailing duties, anti-dumping duties, and intellectual property rights (Coffey et al. 1999). The first three of these trade issues were particularly important for Canada, whereas the last issue was particularly important for the United States. 75 The most notable aspect of the N A F T A is the trade liberalization through tariff reduction between the three countries: tariff rates eliminated over a 10 - 15 year period, with products that were deemed more sensitive to import competition maintaining tariff protection for a longer period of time. Despite this lengthy transition, 50 percent of all tariff rates were eliminated as of 1994, and Canada - United States tariffs continued on their complete tariff rate phase-out according to the C U F T A , and completed in 1998. The Agreement also included a commitment to decrease non-tariff barriers such as quotas and import licenses, particularly with United States - Mexico trade in agriculture, and to permit an easier flow of business and professional people across both national borders through the use of temporary migration VISAs (Coffey et al. 1999; Hufbauer and Schott 1993; Weintraub 1993). As with the C U F T A , the N A F T A contains rules of origin in determining whether or not a product qualifies for the N A F T A tariff rate—essentially, a certain portion of the product must be produced within the free trade area to qualify. These rules of origin increased significantly in the automotive and textiles/apparel sectors—viewed as increases in the degree of protectionism between members of a free trade agreement (Cadsby and Woodside 1993). In the automotive sector, the "domestic" content of the automobiles and engines went from 50 percent (1994) to 56 percent (1998) to 62.5 percent (2002). This increase is believed to be significant because any domestic content requirement above 60 percent requires large investments in power train manufacturing (engines and transmissions), potentially having a deleterious effect on Japanese automotive investment in Canada (Waverman 1993). Due to increases in Japanese automotive investment in Canada, however, those concerns were premature. Similarly 76 with textiles and apparel, the rules of origin became more restrictive as a result of the NAFTA. Even though there has been the elimination of tariff rates and non-tariff barriers on NAFTA trade, very few products qualify under the NAFTA within textiles and apparel. For example, textiles and apparel products must be produced from yarns made in one of the NAFTA countries to qualify for the full NAFTA benefits. This is considered a "schizophrenic result" of the Agreement (Hufbauer and Schott 1993: 3). Other aspects of the NAFTA include: agreements on labour and the environment, though considered more "symbolic gestures" than substantive agreements (Cadsby and Woodside 1993); an expansion of the dispute settlement procedures to include Mexico, but to include a permanent supranational institutional body that may be effective in government trade relations for the member countries; a slight broadening of the scope for financial service liberalization, as well as services in general; moderate liberalization in land transportation services; and the explicit protection of intellectual property rights, a U.S. objective in the CUFTA negotiations (Cadsby and Woodside 1993; Hufbauer and Schott 1993). Notable omissions from the NAFTA include the energy sector and the exclusion of future tariff and non-tariff abrriers to international trade flows. The energy sector, aside from moderate access to the Mexican oil and gas market (Hufbauer and Schott 1993) and a provision for U.S. utility companies to honour existing contracts with Canadian energy providers, is generally immune to free trade. Significant impediments to international trade flows, also unresolved in CUFTA negotiations, such as anti-dump and countervail procedures against member countries continues to be an issue in Canadian trade—softwood lumber, for example. Regardless of the fact that a large 77 portion of the N A F T A documents deals with exceptions to free trade and the restrictions/limitations therein, the N A F T A is a "move toward greater freedom in economic relations among the three [member] countries" (Weintraub 1993). The obvious question to ask now is what effect the N A F T A has had on Canada - United States international trade flows 3.4.6. The Effects of the N A F T A Gould (1998) is the first study to test the independent effect of the N A F T A . Though Gould (1998) only has international trade data through to 1996, his data are measured quarterly to obtain a large number of post-NAFTA observations for statistical testing. In order to control for the general rise in international trade flows over the study period, Gould (1998) investigates the quarter-to-quarter changes in international trade flows. Additionally, Gould (1998) only uses trade flows both originating and destined for the N A F T A countries. Using a 90 percent confidence interval to determine whether or not the N A F T A has impacted international trade, Gould (1998) does find statistical support for the N A F T A producing increases in United States - Mexico international trade flows, particularly imports from Mexico, but finds that the impact of the N A F T A has been statistically insignificant for Canada with both the United States and Mexico. Despite this statistically insignificant finding, the effects of the N A F T A on Canada are found to have positive signs and are likely due to the relatively small number of post-NAFTA observations. Krueger (1999) comes to a similar conclusion for Canada using a gravity model of international trade. This study differs from Gould's (1998), aside from the statistical model, in that it contains data for the N A F T A countries' international trade flows with 78 non-NAFTA countries, and the Canada - United States, United States - Mexico, and Canada - Mexico effects are not measured separately. Though finding that the N A F T A has decreased N A F T A country imports from non-members, the N A F T A has an insignificant, though positive, effect on the N A F T A countries. The results for Krueger (1999) are substantially different for the case of United States - Mexico international trade flows than those found by Gould (1998). This difference is likely due to aggregation bias. Gould (1998) finds that the effects of the N A F T A on Canada are all statistically insignificant, whereas United States' exports to Mexico are marginally significant and United States' imports from Mexico are statistically significant—if Gould (1998) had used a 95 percent confidence interval, United States' exports to Mexico would not have experienced a statistically significant positive effect from the N A F T A . The entire effect of the N A F T A is insignificant for Krueger (1999) and Canada - United States international trade flows dominate the international trade flows within North America as a whole. Therefore, within Krueger's analysis, the insignificant effect of the N A F T A on Canada - United States dominates Krueger's (1999) total intra-NAFTA effect. Additionally, Krueger (1999) only had four yearly post-NAFTA observations in her data set, likely contributing to the insignificant results. The last national analysis of the effect of the N A F T A , by Romalis (2005), is the most comprehensive—Fukao et al. (2003) investigate the effects of the N A F T A , but only regarding trade diversion. Romalis (2005) states that empirical studies investigating the effect of the N A F T A have had difficulty identifying such an effect using nationally aggregated trade data for two reasons. First, Mexico's economy began to internationalize in 1986 and, second, the Peso devaluation that occurred in 1994 - 1995. Consequently, 79 Romalis (2005) uses commodity level data to bypass these national effects. Using a sophisticated theoretical model and corresponding estimation procedure—sophisticated relative to the previously discussed studies—Romalis (2005) finds that the N A F T A has had a substantial impact on international trade flows, but only a modest impact on welfare in the N A F T A countries. This establishment of the positive impact on international trade flows is critical because as stated by Gould (1998), the N A F T A can only impact welfare i f it alters international trade flows. Romalis (2005) is the first national level analysis of the effect of the N A F T A to find an overall positive effect, but it does have its limitations—the unit of analysis for Romalis (2005) is the commodity, and he aggregates those results to the national level. First, Romalis (2005) states that the use of commodity level data avoids the difficulties with the internationalization of the Mexican economy since 1986 and the Peso devaluation from 1994 - 1995. This may be the case, but there are other ways to control for these effects and maintain a more aggregate level of analysis. Most of the studies discussed above use either a time trend variable or year specific dummy variables to control for the general increase in global international trade flows and, hence, the opening up of the Mexican economy during that period. If the Mexican growth of internationalization is considered as potentially biasing the statistical results, a time trend variable or year specific dummy variables specific for the Mexican economy may be included, and tested for significance. Similarly, most studies of the effects of the N A F T A either include an exchange rate variable or they adjust trade volumes based on the exchange rate. Either of these techniques, particularly the former, would capture the Peso devaluation. The use of commodity level data then is not strictly speaking a 80 limitation but it was not necessary to control for the effects of Mexico's internationalization and currency issues. Therefore, the problems associated with Mexico's case are easily avoided. The second limitation lies in the specification of Romalis' (2005) trade model and relates back to the difficulties in measuring trade diversion. Romalis (2005) does find a positive effect for the N A F T A , but also finds significant trade diversion as well. Previous analyses of the effects of the N A F T A have found little support for any trade diversion under the N A F T A , a result that is intuitive given that neither the N A F T A , nor the C U F T A before it, contain any stipulations regarding non-member tariff barriers. Romalis' (2005) finding of significant trade diversion is most Hkely linked to his specification of product variety. He assumes for simplicity in modelling that product varieties are horizontally differentiated based on the country of origin. This means that all varieties are considered to be the same quality, but differentiated only on the "Made in . . . " label—the same type of product differentiation is used in Krugman's (1979, 1980, 1981) model of international trade flows with increasing returns to scale, a type of horizontal product differentiation that is usually put in terms of blue cars versus red cars. This is precisely the difficulty discussed above regarding product quality, commodity classifications, and trade diversion. Therefore, though instructive with the positive effect of the N A F T A at a national level, Romalis (2005) introduces just as much potential for error in his use of commodity level data as he tries to avoid from the use of aggregate trade data. Coughlin and Wall (2003) is the first paper to study the geography of the effects of the N A F T A , albeit on the U.S. states. Employing a rather parsimonious gravity model 81 of international trade, Coughlin and Wall (2003) find that the overall effect of the NAFTA on United States' exports to Canada are positive and statistically significant, a 15 percent increase in international trade flows. Though a modest increase in international trade flows resulting from a free trade agreement, this increase is significant as it is in addition to the gains from the CUFTA. Geographically, thirty-six U.S. states had a greater than 10 percent increase in international trade flows with Canada, while eleven U.S. states exhibited little change (- 10 to 10 percent), and four U.S. states had a decrease of more than 10 percent. When the individual U.S. states are aggregated into the nine Bureau of Economic Analysis regions, however, all U.S. regions indicate positive change for international trade flows with Canada resulting from the NAFTA. This is clearly a result of aggregation bias and the modifiable areal unit problem that shrouds the geography of the effects of the NAFTA, that is, the sub-national regional results are not genuine. And lastly, with a Canadian geographical focus on the effects of the NAFTA, Wall (2003) investigates the effects of the NAFTA on three Canadian regions (western, central, and eastern Canada) on international trade flows with the nine Bureau of Economic Analysis U.S. regions using a parsimonious gravity model of international trade similar to that of Coughlin and Wall (2003). Overall, Canada's exports and imports to and from the United States are up respectively 29 and 14 percent, as a result of the NAFTA representing a large magnitude change for Canadian exports to the United States. Geographically, the effects of the NAFTA vary substantially. Eastern Canada undoubtedly suffered in both their exports and imports to and from the United States resulting from the NAFTA respectively down 9 and 13 percent. In contrast, central 82 Canada exhibits large increases in both exports to and imports from the United States resulting from the N A F T A respectively, up 43 and 18 percent. And western Canada demonstrates a 0.9 percent increase in exports to the United States and a 0.5 percent decrease in imports from the United States leading Wall (2003) to claim that the overall impact of the N A F T A on western Canada is insignificant. Though extremely instructive regarding the geography of the effects of the N A F T A on Canada's economic landscape, Wall (2003) uses relatively coarse Canadian regions to investigate the effects of the N A F T A on Canada. When Coughlin and Wall (2003) aggregated the individual U.S. states into nine regions their results changed significantly, no longer showing that some U.S. states traded less with Canada after the N A F T A entered into force—likely shifting their trading relationships southward to Mexico. Likewise, a similar effect can be expected to be present for Canada because the provinces within each of Wall 's (2003) Canadian regions are also individually diverse economies.1 7 3.5. CONCLUSION This chapter has provided a brief outline of the history of Canada - United States international trade relations with the presentation of empirical evidence of the effects of the C U F T A and the N A F T A for reference in the following chapters. The formal institutionalization of the relationship between Canada and the United States dates to the Another issue regarding the political geography of Canada - United States international trade, undertaken most notably by McCallum (1995) and Helliwell (1996, 1998), is the so-called "border effect". However, this issue is covered in chapter 5. 83 19th century before Canada's confederation, but began its present trajectory in 1965. The Auto Pact and the C U F T A have been found to have a positive impact on international trade flows between Canada and the United States, but the effects of the N A F T A n Canada have been more elusive. With this background of the international trading relationships between Canada and the United States in mind, the following chapters of my dissertation wil l analyze the impact of free trade in detail at both the national and sub-national scales of analysis. The first analysis is undertaken at the national scale to act as a benchmark for further geographical analyses. 8 4 Chapter 4'8 A national analysis of Canada - United States international trade, 1979 - 2003 4.1. INTRODUCTION It should be clear from Chapter 3 that profound economic regulatory changes have occurred with establishment of the free trade agreements between Canada and the United States. Previous research has investigated the effects of these free trade agreements on the Canadian and U.S. economies, but such analyses are typically undertaken using international trade flows aggregated to the national level, or aggregated to the industry level for the nation as a whole. Though such aggregate analyses are instructive and undertaken here, within such aggregations great variation in the type of trade exists. The international trade in goods is generally separated into inter-industry trade and intra-industry trade. Inter-industry trade is based on comparative advantage: each country exports the good it has a comparative advantage in to the other country. Therefore, the goods traded are distinct from each other, originating from different industries, trading food for clothing for example. This type of trade is a reciprocal 1 8 A version of this chapter has been accepted for publication. Andresen, M.A. (2006) The effect of North American trade liberalization on the nature of Canadian trade, 1989 - 2002. American Review of Canadian Studies, forthcoming. 85 relationship, but because each country imports and exports different goods, this type of trade is referred tp as one-way trade: each good only travels in one direction. In contrast, intra-industry trade is not based on comparative advantage. Rather, the more similar the countries are the more intra-industry trade the two countries wil l partake. The goods traded are generally not distinct from each other, but differ based on characteristics. Those characteristics may be based on quality, trading a Mercedes-Benz for a Yugo, or based on other attributes not related to quality, such as colour, shape, sound, etc. Those intra-industry goods differentiated based on quality are called vertically differentiated goods, whereas goods differentiated based on non-quality attributes (similar quality goods) are called horizontally differentiated goods. Many analyses investigating the effect of free trade agreements such as those established between Canada and the United States have used measures to distinguish between inter-and intra-industry trade, but none to date have made any distinctions between vertically and horizontally differentiated goods. This chapter employs a measurement system of international trade that distinguishes intra-industry trade flows based on quality, providing a rich assessment of the changing patterns of international trade flows between Canada and the United States. The primary purpose of this chapter is twofold: first, to provide a comparison of Canada - United States international trading patterns before and after the implementation of free trade; and second, to provide a national analysis of Canada - United States international trading patterns to act as a baseline for explicitly geographical analyses in future chapters. The first task is undertaken using detailed commodity data for the two periods 1979 - 1988 and 1989 — 2003, allowing for differences in trading patterns before 86 and after free trade agreements to be analysed. The second task is implicit in the first. The 1989 - 2003 analysis provides the foundation for showing the benefits of a geographical approach to international trade flows between Canada and the United States. The national level statistics presented here represent the national averages of the individual Canadian provinces. Therefore, the geographical approach taken in later chapters wil l show the regional differences of the impact of free trade on Canada. Unfortunately, geographically disaggregated trade data at the provincial level are not available prior to the implementation of free trade between Canada and the United States. Canadian trade data measured at the provincial and U.S. state levels became available for the first year only in 1988, with the first analysis using these data being published seven years later (see McCallum 1995). The primary limitation imposed by this lack of data is that no comparison of the geography of Canada - United States international trade before and after the establishment of the Canada - United States Free Trade Agreement can be undertaken. This chapter is organized as follows. The data and measurement methodology to distinguish the different types of trade is presented in the following section. The 10 years of Canada's international trade prior to the Canada - United States Free Trade Agreement (CUFTA) is discussed in section 3. Section 4 presents Canada's trading patterns since the inception of the C U F T A . And section 5 concludes that an analysis of international trade flows that includes the changes in the different trade types based on quality enriches the analysis. 87 4.2. D A T A AND M E T H O D O L O G Y 4.2.1. Data The data used in the analysis of 1980 - 1988 international trade patterns for Canada with the world's countries are provided by Statistics Canada (2004a). These data measure Canadian international trade flows using the Standard International Trade Classification (SITC), containing yearly trade values at the four-digit level of aggregation for both imports and exports, indicating the origin and destination at the country level. The data used in the analysis of 1979 - 1988 international trade patterns for Canada with United States are detailed commodity-based international trade flow data provided by The Center for International Data at the University of California-Davis (see Feenstra 1996, 1997 for a detailed account of these data). These data are also provided using the Standard Industrial Trade Classification (SITC), an old classification system for international trade data. For comparability with the 1989 - 2003 time period, the SITC classifications are converted into their equivalent Harmonized Tariff Schedule classification. These data contain the yearly dollar value and quantity of products traded, allowing for the international trade flows to be decomposed into their component parts (discussed below) in order to obtain a better picture of the dynamics of Canada's and the United States' changing international trade relations. The data used in the analysis of 1989 -2003 international trade patterns for Canada with the United States and the world are also detailed commodity-based international trade flow data, but provided by Statistics Canada (2004b). These data are measured using the Harmonized Tariff Schedule with yearly dollar values and quantities for products traded at the 10-digit (imports) and the 8-digit (exports) levels of 88 aggregation. In order to facilitate comparisons with exports and imports, the 10-digit import classifications were recoded to match the 8-digit export classifications. A l l data are converted into constant 1997 Canadian dollars. 4.2.2. The Measurement of Trade Types The analysis below uses the common measures of export, import, and total trade constant dollar values as well as international trade shares to analyze the changing trading patterns for Canada from 1979 to 2003. Though these aggregate measures of international trade flows are instructive for the investigation of international trade in general, and trade policy changes such as free trade agreements, in particular, they may still hide changes in international trade patterns. The first measure used to uncover changes in the pattern of international trade is the often-used Grubel and Lloyd (1975) index. Secondly, the Canada - United States aggregate international trade flows are disentangled into their component parts: one-way trade (inter-industry), two-way trade (intra-industry), and the specialization in the quality of that two-way trade, vertically differentiated and horizontally differentiated intra-industry trade. Within vertically differentiated trade, both high- and low- quality goods can be identified, using the methodology below, identifying Canada's role respectively in high-quality and low-quality market trade, respectively. Horizontally differentiated intra-industry trade is referred to as medium-quality trade as neither Canada nor the United States imports or exports goods with distinctive quality differences. This terminology of medium-quality trade, however, is more of a convenience in reference to high- and low-quality market trade because both countries may be importing high-, medium-, or low-89 quality goods. Due to data limitations, no absolute measures of quality can be ascertained within horizontally differentiated intra-industry trade. Though these trade measures are being used in a national analysis of international trade flows in North America (see Fontagne et al. 1997 for a study on the European Union), they are particularly instructive for a geographical approach to studying that same trade. Though not undertaken in the present chapter, there is no reason to believe that subnational regions have the same spatial distribution of trade in low-quality goods as they do for high-quality or medium-quality goods. Additionally, and arguably more interesting, by separating the various classifications of trade into high-, medium-, and low-quality goods, one can uncover the role of a subnational region within an international production system. As discussed in Chapter 3, the removal in the automotive industry in central Canada of trade barriers led to the rationalization of production. In the case of the automobile industry, Canadian firms assumed their new role in labour intensive production due to the relative cost advantage in labour over the United States. Similar changes have likely occurred across the country. Therefore, the removal of tariff barriers may not only alter the destinations and origins of a province's exports and imports, but alter the very nature of that trade as well. In the present chapter, these measures of trade are used to provide an assessment of the effects of free trade agreements on the international trading patterns of Canada and the United States. The first distinction in the type of trade that needs to be made is that between inter-industry and intra-industry trade. The Grubel and Lloyd (1975) Index is the most 90 widely accepted measure of intra- versus inter-industry trade. The Grubel-Lloyd Index 1 9 captures the degree of overlap in trade between two countries. In the present study, trade overlap is measured at the commodity level. If the Grubel-Lloyd Index has a value of zero, there is no trade overlap and all trade is then classified as one-way trade. In this situation, each country may import and export goods to each other, but there are no overlaps between the type of goods exported and the type of goods imported. If the value of the Grubel-Lloyd Index has a value of one, then there is perfect overlap in trade between the two countries. Although the Grubel-Lloyd Index measures the degree of trade overlap between two countries, it does not indicate when trade becomes two-way trade. If the definition of two-way trade is taken literally, the simultaneous import and export of the same commodity classification, any commodity that has a Grubel-Lloyd Index greater than zero is two-way trade. More generally, we can consider trade within a commodity classification to be two-way trade when there is significant overlap 2 0 for exports and imports in the same commodity classification. In the present analysis, significant overlap is 20 percent, but previous applications have considered 0 and 10 percent significant overlap in exports and imports. The choice in these previous applications has been arbitrary, so a statistical justification of the 20 percent overlap is presented in the technical appendix to this dissertation. Below this level, the overlap is 19 G £ Xi Mi _j XJ Mi , where j is the industry studied, X represents exports, and M represents imports. 20 X w 0 _ w a y trade if ' p t ' ^ > 20% > where p represents the product and t represents the year. - Max{Xl>„Mj 91 not considered significant and, therefore, does not represent the structural feature of trade (Abd-el-Rahman 1991). Using this criterion, an index of two-way trade can be calculated. This index represents the share of trade that is classified as two-way trade and was proposed by Fontagne and Freudenberg (1997). As with the share of trade volumes, this index varies from zero to one, with one representing perfect two-way trade. It should be noted that despite the similarity in interpretation with the Grubel-Lloyd Index, the two-way trade index (TW) is fundamentally different from the Grubel-Lloyd Index; rather than measuring the degree of trade overlap between two countries, the T W index considers all trade over the 20 percent threshold to be two-way trade. However, when these two indices are compared, they are quite similar (Fontagne and Freudenberg 1997). The difference between the two indices is that T W is a measure of all trade flows that are considered to be two-way trade, whereas the Grubel-Lloyd Index only measures the degree of trade overlap with no indication of when trade becomes two-way trade. Both indices are presented below for comparative purposes. Thus far, only one- and two-way trade have been differentiated from each other. Two-way trade, however, can be separated into three further classifications based on the quality of exports and imports. As discussed above, these quality classifications are horizontal product differentiation (characterized by products with similar quality levels with different attributes), and vertical product differentiation that is characterized by products with significantly different quality levels (high and low quality). Following Stiglitz (1987), prices are assumed to represent quality differences. Consequently, 2i Y(x, + M,) Share of Two- Way Trade = ^ f-[ LV 92 differences in the unit values (UV) or prices of these commodities are always assumed to represent quality differences. Unit values are defined for each commodity classification as the value of trade divided by the quantity traded, giving an average price of the goods traded in this category. Clearly, the more disaggregated the classification system the better this method represents the price of the commodities—a classification system such as the 8-digit Harmonized Tariff Schedule with approximately 10 000 commodity 00 classifications captures this well. Horizontal product differentiation can then be thought of as the unit value of exports in a commodity classification being similar to the unit value of imports in that same commodity classification. However, there must be some allowance for similar quality goods to have some price variation. Therefore, a threshold of similarity needs to be established. At what point is the price difference between exports and imports considered to represent quality differences? If the ratio of the export and import unit values is within a certain threshold2 3 they are considered of equal quality. As with the degree of overlap, some justification for the quality threshold needs to be undertaken. Past research has used 15 and 25 percent quality thresholds, but the threshold in the present analysis is 15 percent.24 This trade is considered horizontally differentiated (HD). If the ratio of the export and import unit values is greater than the threshold, there are 2 2 The categories in the Harmonized Tariff Schedule are so specific that different commodities have different quantity measures: litres, kilograms, number, etc. while the SITC classification system is more general and uses tonnes as its quantity variable for all commodity categories. 1.15 UV M 2 4 This threshold value has also been determined in the statistical analysis presented in the technical appendix to this dissertation. 93 quality differences between the goods exported and the goods imported.25 If Canada's export unit value is greater than its import unit value for a commodity classification, Canada trades a relatively high-quality product, and if Canada's export unit value is less than its import unit value for a commodity classification Canada trades in a relatively low-quality product: vertically differentiated high-quality trade (VDHQ) and vertically differentiated low-quality trade (VDLQ), respectively. These three classifications can be used to create indices representing their shares in overall trade flows. The sum of HD, VDHQ, and VDLQ equal the total value of two-way trade (TW).26 The primary limitation of the data used in this study is the non-existence of quantities for every product category. Sometimes this non-existence of quantities is due to confidentiality when particular products are produced by only a few firms. However, quantity information is also missing when the same product category is recorded with multiple quantity units—different exporters and importers measure their trade using different units. Despite the low degree of aggregation in these data, many products are reported using multiple quantity units, with no standardization imposed. As a result, the percentage of horizontally- and vertically-differentiated trade is often not equal to two-way trade. Therefore, the proportions of trade based on quality must be viewed as a sample of all two-way trade in most industrial sectors. To aid in interpretation, the two-way trade index (TW) is supplemented with a restricted two-way trade index (TWR) that 25 UVX , UV* 1 — > 1.15 or — < UVM UVM 1.15 ^ - i + A / 1 2 6 TWHD - Pi*j ^HD^ p'' / M , where HD represents horizontally differentiated trade, Z represents all trade types, p, e j represents product / in industry j, and / represents the year. A similar formula is used in the calculation of the shares of two-way trade in vertically differentiated products (VDHQ and VDLQ). 94 only includes two-way trade for those quantities reported for both the import and export value—both indices are reported in Table 4.13, below. This limitation in the data does not decrease the value of implementing the above measurement methodology. In order to assess the effects of free trade agreements it is necessary to decompose trade into its component parts. This type of analysis allows for a better insight into the integration of the Canadian and U.S. economies. This insight is gained through not only identifying the changes in the volume of international trade flows either at the national of industry level of analysis, but identifying what types of trade have increased and/or decreased. Has Canada's increases in international trade flows been from increases in low-quality trade at the expense of high-quality trade, or vice versa? If Canada's low-quality trade has increased, Canada's role in the international division of labour would be to provide cheap, labour-intensive goods originating from relatively low-paying jobs. This type of increase in international trade volumes may lead to decreases in welfare for many people due to lost wages. On the other hand, increases in high-quality trade at the expense of low-quality trade would have the opposite effect. Before turning to the changes In these indices since the inception of the free trade agreements, the preceding 10 years of Canada's international trade flows are presented to provide a context. 4.3. T H E I N T E R V E N I N G Y E A R S Almost twenty-five years separate the implementation of the Auto Pact and the C U F T A , but the intervening years were not without change. Despite the efforts to diversify the international trading portfolio of Canada by the Trudeau Liberal government in Canada, 95 Canadian exporters became increasingly more reliant on U.S. markets (Hufbauer and Schott 1998), tying the Canadian economy to U.S. economic performance. In the ten years preceding the implementation of the C U F T A , Canadian real G D P grew by a factor of 1.3 and international trade with the United States grew slightly faster, by a factor of 1.33, with exports (1.35) growing faster than imports (1.30). By way of comparison, Canada's trading patterns with the world (1980 - 1988) and with the United States (1979 - 1988) before free trade are presented. Additionally, a detailed industrial sector analysis of Canada - United States international trading patterns at the industrial sector level is presented. Both are instructive because knowing the spatial distribution of international trade flows before the free trade agreements allows an interpretation of the post-free trade spatial distribution. 4.3.1. Canada's Trading Pattern with the world, 1980 - 1988 Regionally, Canadian export volumes exhibited decreases to East Asia ( E A - A N Z ) and the European Union ( E U l 5 ) 2 8 during the early 1980s, but exhibited strong increases to both regions in the late 1980s just before the C U F T A entered into force (see Table 4.1). The United States showed little change in exports from Canada in the early 1980s, but did increase in the late 1980s. Mexico and the Rest of the World (ROW), however, remained constant with their exports from Canada with a low degree of exports that decreased a small degree over the 1980s. 2 7 East Asia is defined as the ASEAN 10: Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, Vietnam, plus China, Hong Kong, Taiwan, South Korea, and Japan plus Australia and New Zealand. 2 8 European Union-15: Austria, Belgium-Luxembourg, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, United Kingdom. 96 <See Table 4.1, page 241 > Both Mexico and the R O W show similar declining patterns for the levels of exports and total international trade flows. The decreases for the R O W are substantial, falling by 50 percent in most cases. However, both the EU15 and E A - A N Z essentially doubled their volumes of Canadian imports from 1980 - 1988. Overall, international trade flows for 1980 - 1988 with the European Union increased by a factor of 1.15, slower than Canadian real G D P growth. In contrast, total international trade flows with East Asia increased by a factor of 1.7, significantly higher than GDP. With respect to international trade shares (Table 4.2) both Mexico and the R O W exhibit the same patterns as the levels of international trade flows remaining respectively constant and decreasing. East Asia remained relatively constant in its Canadian export share until the late 1980s. However, after this point it started to increase. The European Union, though regaining some of its Canadian export share lost in the mid-1980s, lost significant ground in Canadian exports over the entire study period. With regard to Canadian imports, East Asia and the European Union had steady increases. As a consequence of large increases in imports, their total shares in Canadian trade increased overall. The United States, maintaining its dominance in Canadian international trade relations, increased its Canadian export and total trade shares from 1980 - 1988 and its Canadian import share remain relatively constant. <See Table 4.2, page 242> Therefore, at a national level from 1980 to 1988, Canada was significantly adjusting the spatial distribution of its international trade. Both the destinations of Canada's exports and the origins of Canada's imports changed significantly in the years 97 before the C U F T A . This changing spatial division of international trade with respect to both the world, and specifically the United States, is most likely due to the successive rounds of G A T T negotiations (Kennedy Round, 1964-1967; Tokyo Round, 1973 -1979; and the last and most comprehensive, Uruguay Round, 1986 - 1994) (World Trade Organisation 2006). 4.3.2. Canada - United States Trading Patterns, 1979 - 1988 Let me now turn to Canada - United States international trade flows, 1979 - 1988. The industrial sectors (see appendix for the definitions of the industrial sectors) that exhibited above average export growth were Beverages and Tobacco, Plastics and Rubber Products, Printing and Publishing, Primary and Fabricated Metals, Electrical Machinery, Motor Vehicles and Parts, Other Transport, and Other Industries (Tables 4.3 and 4.4). A l l of these industrial sectors except Other Industries had export growth far outstripping import growth over this period. In fact, aside from Beverages and Tobacco (1.35), only Other Industries (8.01) had above average (1.30) growth in imports. Unfortunately due to the aggregate nature of the data in this time period (Standard International Trade Classification, Revision 2), little can be done to uncover which industrial sectors within Other Industries had such phenomenal import growth. <See Tables 4.3 and 4.4, pages 243 - 246> With regard to relative trade shares for Canada - United States international trade flows for 1979 - 1988, there was also significant adjustment (see Table 4.5). Animal Agriculture, Vegetable Agriculture, and Mining Quarrying Petroleum all decreased their shares in Canadian exports. However, Printing and Publishing, Leather, Textiles, Clothing, Electrical Machinery, and Motor Vehicles and Parts all increased their 98 Canadian export shares to the United States—Motor Vehicles and Parts almost doubled its Canadian export share. Almost the opposite occurred within Canadian import shares: Textiles, Wood Products, Printing and Publishing, Leather, Clothing, Primary and Fabricated Metals, Non-metallic Mineral Products, and Mining Quarrying Petroleum all lost Canadian import shares. Only Electrical Machinery and Other Industries increased their Canadian import shares, and Electrical Machinery's growth was moderate. Overall, Vegetable Agriculture, Chemicals, Wood Products, Non-metallic Mineral Products, and Mining Quarrying Petroleum showed significant decreases in their total trade shares over the time period, with Motor Vehicles and Parts and Other Industries showing the only impressive growth. <See Table 4.5, page 247 -249> Turning to the indices for the decomposed trade types, Table 4.6 shows that two-way trade has increased its share steadily in the ten years prior to the free trade agreements. The level of low-quality market trade has remained relatively constant, whereas medium- and high-quality market trade have both increased their shares. Prior to 1984, high-quality goods exhibited significant increases in their share ( 7 - 3 1 percent of trade, 1979 - 1984), but that share subsequently dropped to just below 20 percent with medium-quality market trade taking its place. Given that low-quality market trade has remained relatively constant (a slight decrease over the entire study period), the increases in two-way trade are attributed to increases in both medium- and high-quality market trade. What this means for Canada is that in the years preceding the C U F T A , higher quality goods dominated Canada - United States international trade flows. At this same 99 time, one-way trade also decreased. As discussed above, this has implication for labour in Canada. The movement toward higher quality goods in trade, and therefore production, is a shift toward the increased need for skilled labour to produce the higher quality products and the corresponding wage increases. At the industrial sector level, two-way trade increased or remained relatively constant in all industrial sectors except Animal Agriculture, Mining Quarrying and Petroleum, and Leather. Incidentally, these three industrial sectors also decreased their levels of international trade flows, decreased their trade shares, and increased their shares of low-quality market trade over the study period. Animal Agriculture and Leather also exhibit sharp decreases in their shares of high-quality market trade. <See Table 4.6, pages 250 - 254> The overall trend for the industrial sectors is an increase in the low-quality market trade. The most notable increases are in Food (17 — 44 percent), Beverages an Tobacco (2 - 25 percent), Plastics and Rubber Products (3 - 45 percent), and Textiles (9 - 30 percent). Needless to say, this common trend in so many industrial sectors does not manifest itself in the national aggregate indices. Therefore, the effects of a small number of industrial sectors are dominating the national aggregated results. These industrial sectors are Primary and Fabricated Metals, Non-electrical Machinery, Electrical Machinery and Motor Vehicles and Parts, consisting of 25 - 35 percent of all Canada-U.S. international trade flows depending on the year of measurement. Of these four industrial sectors, Primary and Fabricated Metals is the only industrial sector to decrease its share of two-way trade, 46 - 35 percent. Though a substantial drop, the low-quality trade share dropped an even greater amount, 3 0 - 9 100 percent. Making up this difference are increases in both medium- and high-quality trade. Therefore, this industrial sector was undergoing a substantial change in its export and import structures long before tariff and non-tariff barriers to international trade began to fall in 1989. Non-electrical machinery exhibits a marked increase in its two-way share of trade (54 - 84 percent), with an increase in medium-quality trade but significant volatility in both high- and low-quality trade. Overall, the trend in high-quality trade appears to be relatively constant, whereas the trend in low-quality trade is negative, but moderate. A difficulty with these interpretations in this industrial sector is its low T W R values, relative to its T W values. Consequently, much of the change in the product quality component of trade within this industrial sector is likely being lost due to a lack of appropriate data. Electrical Machinery, however, shows a significant drop in low-quality market trade (34 - 17 percent), while high-quality market trade increases significantly ( 1 9 - 4 4 percent)-—medium-quality market trade remained relatively constant. Therefore, not only do the increases in high-quality market trade account for the moderate growth in two-way trade (68 - 80 percent) in this industrial sector, but it also displaced a large portion of low-quality market trade. Motor Vehicles and Parts low-quality market trade decreased from 1 6 - 0 percent over the study period with high-quality market trade increasing significantly from 6 to almost 40 percent—medium-quality market trade also increased, but more moderately (51 - 61 percent). These changes in the nature of Motor Vehicles and Parts trade occurred while two-way trade increased substantially from an already high level (78 - 99 101 percent). Though all four industrial sectors had similar trade in their respective indices, the changes in the Motor Vehicles and Parts and Electrical Machinery industrial sectors are clearly the driving force behind the patterns at the national aggregated level. Herein lays the utility of the use of high-, medium-, and low-quality trade classifications in analyzing international trade flows. For all practical purposes, two industrial sectors are behind the nationally measured changes in trade patterns. Most notable is the disconnect between the changes in the quality composition of international trade flows at the national level and most Canadian industrial sectors. In all industrial sectors aside from Primary and Fabricated Metals, Non-Electrical Machinery, Electrical Machinery, and Motor Vehicles and Parts, low-quality trade exhibited substantial increases. Only in the four industrial sectors mentioned here did low-quality trade exhibit substantial decreases. Therefore, the vast majority of industrial sectors have likely experienced increases in lesser-skilled employment with corresponding lower wages. Only a handful of industrial sectors (that consist of a large number of manufacturing jobs, however) have moved into the need for greater skilled workers. And given the nature of the Canadian economy being a patchwork quilt of industrial sectors (Britton 1996), the industrial sectors with these positive impacts are in particular places, namely Ontario. Needless to say, Canada's place in the international economy was far from constant before free trade agreement negotiations began with the United States in the mid-1980s. The anticipation of free trade with the United States and Canada's involvement in the successive rounds of G A T T negotiations have had a significant impact on not only with which countries Canada trades, but the nature of that trade. Both the United States and automotive trade dominate Canada's international trading relations. / 102 However, natural resource and machinery industries are beginning to become more active in both imports and exports. And trading relations with East Asia are increasing significantly at the expense of the Rest of the World—the trading relationship with the EU15 has remained constant. 4.4. CANADA - UNITED STATES TRADE PATTERNS, 1989 - 2003 The results for Canada - United States international trade patterns, 1989 - 2003 are presented below. I begin with an overview of the reductions in tariff rates at the national and industrial sector levels between Canada and the United States following the free trade agreements. This is followed by an analysis of Canada's changing trade relations with the world and a detailed analysis of Canada - United States international trade at the industrial level. 4.4.1. Tariff Reductions The tariff reductions shown in Table 4.7 are simple averages of commodity-based (Harmonized Tariff Schedule) tariff rates calculated at the national and industrial sector levels. The majority of the tariff rates are ad valorem (value-based), but approximately 5 percent of the commodity categories are partially or wholly quantity-based (dollars per kilogram, etc.). As such, ad valorem equivalent rates are calculated for these tariff categories using the duty collected and the total value exported to obtain an ad valorem equivalent tariff rate (Department of Finance Canada 2004, External Affairs Canada 1987, Feenstra et al. 2002, and United States International Trade Commission (USITC) 2004). The value- and quantity-based rates are generally in the same magnitude; however, the highest tariff rates are quantity-based, by a factor of eight over the highest 103 value-based tariff rates, and found in the traditionally protected industrial sectors such as Clothing, Textiles, and Agriculture. There are a number of limitations of these tariff databases. The primary limitation is that information on quotas, such as quotas under the Multi-Fibre Agreement or special quotas under N A F T A , are not included. These databases also do not include anti-dump, countervail, or other special duties such as those imposed on the softwood lumber industry. And importantly with respect to the automotive and textiles and clothing industrial sectors, the eligibility of commodities for the preferential C U F T A and N A F T A tariff rates based on the rules of origin is not included in these calculations (Feenstra et al. 2002). Despite these limitations, these tariff databases provide a good indication of the level of protection for Canada, the United States, and their respective industrial sectors. The Canadian and U.S. average tariff rates are shown in Table 4.7. In the base year, 1988, the ad valorem equivalent tariff rate for Canada (8.91 percent) is almost one and one-half times that of the United States' ad valorem equivalent tariff rate (6.1 percent). This shows that Canada's degree of protection was significantly higher than the United States before the establishment of the C U F T A . The tariff rates of the two countries do converge by 1995 and are now approximately 0.5 percent. A curiosity to notice is that both the United States and Canada have positive, though small in magnitude, tariff rates after 1998 when tariff rates were to be zero according to the C U F T A negotiations. The dominant source of these positive post-1998 tariff rates is quantity-based tariffs in the Animal Agriculture, Food, and Beverage and Tobacco industrial sectors—traditionally, highly protected industries. <See Table 4.7, pages 255 - 256> 104 The industrial sector tariff rates for Canada and the United States are also shown in Table 4.7. A l l industrial sectors exhibit the same basic pattern of tariff rate reductions, with the Beverages and Tobacco, Textiles, and Clothing industrial sectors having the greatest decreases in tariff rates for both countries—the Animal Agriculture industrial sector in the United States also exhibited significant tariff rate reductions. The most important aspect of the industrial sector tariff rate reductions to notice is the increase in tariff rates for Canada's Animal Agriculture and Beverage and Tobacco and for the United States' Animal Agriculture and Food industrial sectors. This initially counter intuitive result—present in many industrial sectors, but quite apparent in the industrial sectors mentioned above—is present in the industrial sectors that have many quantity-based tariff rates. In these industrial sectors, international trade flows are growing at a faster rate than the tariff rate reductions. As a result, even though the quantity-based tariff rate is decreasing, causing increases in the trading flow volumes, the ad valorem equivalent tariff rate is actually rising. Therefore, these tariff rate increases are actually a consequence of decreased protection. 4.4.2. International Trade at the National Level The levels of Canadian international trade flows to the United States, Mexico, the European Union, East Asia (hereafter referred to as E A - A N Z ) , and the Rest of the World (ROW) are shown in Table 4.8. Canada's international trade with the United States has grown steadily throughout the study period, with exports growing at a rate slightly higher than imports. Concerning the possibility of trade diversion resulting from the free trade agreements, total international trade flows to all other regions of the world, including Mexico, have increased at a faster rate than the Canadian economy. This means that all 105 regions of the world have become more important to the Canadian economy over this time period. In comparison with the 1980 - 1988 period, international trade flows with the United States grew substantially post-1991, after almost 10 years of very little growth. Also the trend of Canada's international trade flows with Mexico and the R O W have reversed, with the R O W gaining substantial ground lost in the previous decade and Mexico exhibiting substantial growth ($9 - $33 billion). E A - A N Z continues to be Canada's largest trading partner after the United States. Considering imports separately, the pattern is similar to the larger trends just discussed with exceptional growth in Canada's imports from Mexico. Exports exhibit a slightly different pattern, growing significantly faster than the Canadian economy for the United States and Mexico, but slower, though positive, for the EU15, E A - A N Z , and R O W . Therefore, any concerns regarding trade diverting effects of the free trade agreements are unfounded, and consistent with previous research. As with the levels of total trade, Mexico and the R O W have regained lost growth in both imports and exports. <See Table 4.8, page 257> The relative shares of Canadian international trade flows to the same countries/regions are shown in Table 4.9. Canada is clearly adjusting its trading relationships with the rest of the world as a result of the free trade agreements. The EU15, E A - A N Z , and the R O W are all losing shares in Canadian international trade flows, particularly with respect to exports, despite gains in the value of international trade flows. This is a continuing trend from 1980 - 1988 for Mexico and the R O W , but the EU15 and E A - A N Z had increased their shares of Canadian exports in the four to five years prior to the C U F T A entering into force. Essentially Canadian exports have been 106 shifting their destinations from Mexico, the EU15, E A - A N Z , and R O W to the United States over the past 15 years: 74 percent (1989) to 85 percent (2003). Mexico has only increased its share of Canadian exports marginally. E A - A N Z began to lose some of its import share after the implementation of the N A F T A , but has since regained its losses; both the EU15 and the R O W show consistent moderate gains in their import shares, also exhibiting slight decreases around the time the N A F T A was implemented. Prior to the C U F T A , both the EU15 and E A - A N Z exhibited strong gains in Canadian import shares. The United States, however, has decreased its share in Canadian imports by 3 percent, with Mexico absorbing most of the U.S. losses. <See Table 4.9, page 258> The overall result of the free trade agreements is Canada reorganizing its international trading relationships both within and outside of North America. Additionally, this reorganization may have occurred without any losses in global welfare because Canada's international trade flows outside of North America have increased in real terms—this hypothesis, however, is not tested. In many cases, that growth in international trade flows has outstripped the growth of the Canadian economy. 4.4.3. International Trade Growth at the Industrial Sector Level As shown in Table 4.10, almost all industrial sectors have experienced significant growth in the volume of international trade flows in real terms—all values are in constant 1997 Canadian dollars for exports, imports, and total trade. The only exception to this pattern is the Leather industrial sector that exhibited almost zero growth—the norm in the 1979 -1988 period. The Food Products and Clothing industrial sectors exhibited the strongest growth, with trading volumes five times the 1989 value, or more, in 2003- Generally 107 speaking, all industrial sectors in the Canadian economy expanded their levels of international trade flows with the United States after the C U F T A was signed. <See Table 4.10, pages 259 - 261 > Separating international trade flows into exports and imports, export growth outstripped import growth in all but four industrial sectors. With increase factors—see Table 4.11—as high as 5.18 (Textiles), 6.60 (Food), and 9.40 (Clothing), and overall exports (2.55) and imports (1.97), Canada's trade balance with the United States improved. This outcome satisfies one of the goals set out by Canadian negotiators of the C U F T A . From 1979 - 1988, export growth also tended to be stronger than import growth, but the magnitude of growth for exports, imports, and total trade after 1989 dwarf the corresponding values before free trade. <See Table 4.11, page 262> Regarding the timing of the expansion in international trade flows, most industrial sectors show smooth expansion over the study period. However, Vegetable Agriculture, Mining Quarrying and Petroleum, Wood Products, Primary and Fabricated Metals Products, Other Transport, Professional Goods, and Other all appear to have accelerated their growth at times coinciding with the implementation of the N A F T A . Despite this appearance, caution should be exercised with any interpretations given that only five years separate the C U F T A and the N A F T A and no inferential testing is done to confirm or deny this appearance in the data. Lastly, as shown in the national levels of imports and exports, as well as most of the industrial sectors, the levels of post-2000 international trade flows have fallen slightly, likely due to the changing political climate in the United States resulting from the 11 September 2001 terrorist attacks (9/11). Since 9/11 the 108 United States government has increased security at its borders, increasing the costs of international trade. <See Table 4.12, pages 263 - 265> Turning to Table 4.12, the relative industrial sector shares in Canada - United States' trade, the changes over the study period are much more varied indicating changes in the industrial structure of international trade flows. Paper Products, Printing and Publishing, Primary and Fabricated Metals, Non-electrical Machinery, and Motor Vehicles and Parts all exhibit declines in their shares of international trade flows, despite strong growth in the levels of that trade. The decrease in the share of Motor Vehicles and Parts from 28.6 percent (1989) to a low of 23.3 percent (2001) is indicative of the Canadian economy's move toward a more diversified international trade portfolio now that free trade is established in more than this one industrial sector of the economy. The most recent decrease in the share of Wood Products, particularly after 1999, is likely due to the softwood lumber dispute that escalated in recent years—the industrial sector's share had been increasing, albeit slowly, until that time. Notable expansions include Electrical Machinery (prior to 2001), Mining Quarrying and Petroleum and Plastics and Rubber Products industrial sectors, with Food, Chemicals, and Other exhibiting moderate relative expansion. Referring back to Tables 4.10 and 4.11, the industrial sectors that experienced above average increases in international trade flow volumes also generally experienced above average tariff rate reductions. For both Canada and the United States, the Beverages and Tobacco, Plastics and Rubber Products, Textiles, and Clothing industrial sectors all had above average increases in trade flows and average tariff rate reductions— 109 this is particularly true for Canadian exports from the Clothing and Textiles industrial sectors. A simple correlation analysis confirms this supposition to be correct. Those Canadian industrial sectors facing the highest initial tariff barriers to the United States' market has the highest increases in export volumes (r = 0.46, p-value = 0.037), whereas those industrial sectors that received the most protection from Canadian tariff rates had the greatest increases in import volumes, though the result is statistically insignificant (r = 0.218, p-value = 0.355). This result indirectly shows the greater importance of the U.S. market to Canada than the Canadian market to the United States, particularly because Canadian tariffs were on average greater than those of their U.S. counterparts. 4.4.4. Canada-United States International Trade by Trade Types At the aggregate country level, the Grubel-Lloyd (GL) and Two-Way Trade (TW) indices indicate gradual increases over time. However, at the individual sector level, changes in two-way trade are, in some cases, large in magnitude (see Table 4.13). Only the industrial sectors of Mining Quarrying and Petroleum, Clothing, Non-metallic Mineral Products, Motor Vehicles and Parts, and Other Industries exhibit declines in the level of two-way trade, with significant declines only in Clothing and Non-metallic Mineral Products. The significant difference in the share of two-way trade in Motor Vehicles and Parts between the two study periods is due to the significantly greater level of disaggregation in the latter time period—fewer commodities are grouped together using the Harmonized Tariff System. Industrial sectors that expanded their two-way trade significantly include Animal Agriculture, Chemicals, Plastics and Rubber Products, Paper Products, Printing and Publishing, Textiles, and Professional Goods. As shown in Table 4.11, the general expansion of two-way trade comes from Canada's increased exports to 110 the United States in almost every industrial sector. Aside from Plastics and Rubber Products and Paper Products, none of these industrial sectors exhibited significant change in their relative international trade flow shares or exceptional growth in the levels of international trade flows. Therefore, merely separating international trade flows into one-way and two-way trade provides significant insight into the changes in the level of industrial sector cross-border integration. Separating two-way trade into low-, medium-, and high-end markets provides further insights into the changing relationship of the Canadian and U.S. economies. At the aggregate national level, despite little change in two-way trade as a whole, Canada continues to move into the middle- and high-end market trade in international trade after the implementation of the C U F T A . The medium-quality market trade (HD), though volatile, shows an upward trend over the study period (0.133 to 0.157), with a similar trend, somewhat less volatile, in the high-end market (VDHQ) over the study period (0.079 to 0.112). Also worth noting is the timing of these changes. Both increases only occur after the implementation of the N A F T A , indicating (though not proving) that the N A F T A has an independent affect on the Canada - United States trading relationship, over and above that of the C U F T A . This point is discussed further, below. <See Table 4.13, pages 266 - 270> At the industrial sector level, sectors that show increases in two-way trade generally exhibit increases in high-end markets at the expense of low- and/or middle-end 2 9 In order to properly asses the effect of the NAFTA, many factors that have coincided with the NAFTA's timing must be controlled for in an inferential analysis. Such factors include, but are not limited to: the economic performance of the U.S. economy, financial markets, and the value of the Canadian dollar. I l l markets, though middle-end markets do commonly rise. This result is the opposite of that in the preceding period. Industrial sectors that exhibit notable changes in the composition of international trade flows are Animal Agriculture and Leather. Animal Agriculture has since increased its share of two-way trade, whereas Leather has continued to decrease two-way trade, but only moderately. In the previous period, Animal Agriculture increased its share of low-quality market trade at the expense of medium- and high-quality market trade, and similarly for Leather. However, the opposite is true in this latter study period. In the case of Animal Agriculture, both medium- and high-quality market trade have increased at the expense of low-quality market trade; for Leather, high-quality market trade has increased at the expense of low- and medium-quality market trade. This pattern of substituting into high- and medium-quality market trade at the expense of low-quality market trade is also present in those industrial sectors that exhibit growth in two-way trade, 1979 - 1988. Vegetable Agriculture, Food, Plastics and Rubber Products, Textiles, and Non-electrical Machinery all exhibit this trade pattern in the latter study period. This movement into high- and medium-quality market trade in the majority of industrial sectors under free trade, rather than only a few, shows promise for Canada's integration into the North American economy. Though similar data are not available for U.S.-Mexico international trade flows, it can be confidently assumed, due to the levels of skilled labour and capital, that the United States trades relatively high-quality goods for relatively low-quality goods with Mexico, if that trade can be classified as two-way trade—this is the general finding of Schott (2004), with respect to product quality and the wealth of nations. Given that Canada trades high-quality goods for low-112 quality goods on average with the United States and the majority of that trade is classified as two-way trade, the ranking of high-, medium-, and low-quality market trade appears to follow a north-to-south relationship. That is, it appears that Canada is specializing in the export of high-quality market goods, the United States is specialising in the export of medium-quality market goods, and Mexico is specialising in the export of low-quality market goods. As indicated at the national level, changes in the low-, medium- and high-end portions of the market dominantly occur at or after the time of the N A F T A entering into force, once again indicating that the N A F T A may have had an independent affect on the Canada - United States trading relationship. Consequently, there appears to be a new spatial division of labour in international trade patterns occurring in North America. Usually, it is thought that Canada specializes and trades in the low value-added goods with the United States, along the lines of the arguments made in the staples thesis (see Barnes et al. 2001). However, this result undermines that assertion. Most industrial sectors are moving away from inter-industry trade (the dominant form of trade that operates under the staples thesis) and into intra-industry trade in high- and medium-quality goods. The inclusion of Mexico into the North American free trade area, and/or the corresponding trade policies to deal with Mexico's inclusion, has more favourably placed Canada with regard to its most significant trading partner, the United States. Though this point is not discussed by Trefler (2004), this finding is consistent with Trefler's (2004) effect of the C U F T A . Resulting from the C U F T A , skill levels, wages, and productivity have all increased in Canada, providing the necessary conditions for increased quality production: a higher-paid, higher-skilled Canadian workforce. Though it may be premature to claim that 113 Canada is now beginning to work its way out of the staples trap (and which is beyond the scope of this dissertation) it certainly appears as though this may be the case. However, caution should be taken in any interpretation of changes in the Printing and Publishing, Other Transport, Professional Goods, and Other industrial sectors due to their small (non-existent for Printing and Publishing) samples of two-way trade commodity categories that have quantities reported for both imports and exports. As shown in Table 4.13, these industrial classifications have very low T W R values, relative to their respective T W values. As indicated above in the section on data and methodology, this occurs when quantities are not available for all of the commodity categories. Consequently, product quality can only be calculated for very few of the commodities in these industrial classifications making any inferences on the level of product quality in trade suspect. 4.5. SUMMARY AND CONCLUSIONS The Canada - United States Free Trade Agreement and the North American Free Trade Agreement have undoubtedly changed Canada's international trading relationship with the United States and the world in a far more comprehensive manner than previous trading agreements between these two countries. Both Agreements set the standard for the integration of regional trading partners not only on trade and goods, but services, investment, dispute resolution, and trade in agricultural industries. The N A F T A is also the first free trade agreement negotiated and implemented between industrialized and developing economies. 114 There has been a definite reorganization of Canada's trading relationships with regions of the world, but this reorganization does not come at the price of trade diversion. Not only has Canada's trade within North America increased, but so has its trade with the rest of the world. Canada - United States international trade flows have grown in all but a few industrial sectors in the Canadian economy, and there has been some significant restructuring in the relative shares of these industrial sectors pertaining particularly to United States' international trade flows. However, these more traditional measures of change in international trade relations tell part of the story resulting from the C U F T A and, in particular, the N A F T A . Separating international trade flows into one-way and two-way trade shows the changes in the level of industrial sector cross-border integration. Two-way trade has increased within most industrial sectors, indicating stronger linkages between Canadian and U.S. industries. And within the two-way trade category, Canada is moving into the higher-end product markets. On the whole, the free trade agreements appear to have impacted the Canadian economy positively. Though significant trade-induced industrial adjustment is likely present, the overall effect of free trade in North America has been an increase in trade that may allow the Canadian economy to better place itself in the North American and global economies for the benefit of all Canadians in the long run. Through international trade with the United States and Mexico, the Canadian economy is moving away from lower quality markets into higher quality markets, likely increasing the demand for skilled labour, providing better wages and working conditions. And, though this is speculation at this point, as a result Canada appears to be breaking free of the 115 staples trap, potentially releasing Canada from the inherent instabilities of staples production. Future chapters investigate the local effects associated with these apparently positive outcomes at the national aggregate level. As indicated in Britton (1996), the various industrial sectors of the Canadian economy are largely associated with particular provinces: forestry in British Columbia, oil and gas in Alberta, and automobiles in Ontario. Therefore, the changes occurring in the industrial sectors are likely representing changes occurring at a provincial scale of analysis. As such, after discussing issues regarding the integration of Canada and the United States involving the political geography of Canada - United States interregional trade and the effects of the NAFTA in the following chapter, the analysis proceeds to the provincial scale. Chapter 5 116 Canada - United States integration: the effects of the national border and the NAFTA 5.1. INTRODUCTION Before I turn to my own geographical analysis of Canada's international trade flows with the United States, I will examine two related issues that have emerged from the existing literature on Canada - United States economic integration, and which have been touched upon at least implicitly above. These are the border effect and the North American Free Trade Agreement ( N A F T A ) effect. The border effect refers to the impact that the Canada - United States border has on international trade flows, and the N A F T A effect refers to the inability of previous research to uncover the impact the N A F T A has had on Canada - United States international trade flows. Over the last 40 years, as I have shown, the level of integration between Canada and the United States has increased substantially. Today, as a result of that integration, approximately $2 billion of goods and services crosses the Canada - United States border every single day (Sydor 2003). Since 1989, the year the Canada - United States Free Trade Agreement (CUFTA) entered into force, Canada's exports of merchandise to the United States has almost tripled and imports of merchandise from the United States have almost doubled in real terms—almost an increase factor of 2.5 for total merchandise A version of this chapter has been submitted for publication. 117 trade. Not surprisingly, Canada's share of international trade with the United States has risen from 69 to 73 percent. This increase is all due to exports (74 to 85 percent), however. In fact, the import share from the United States has fallen since the inception of the C U F T A (65 to 60 percent). At the provincial level, all provinces except Newfoundland have more than doubled their international trade with the United States (Newfoundland increased its international trade with the United States by a factor of 1.5). And separating provincial trade into exports and imports, all provinces except Prince Edward Island's imports increased substantially in real terms, 1989 - 2003 (Statistics Canada 2004b). Regardless of these increases, research undertaken by economic geographers and economists has found that the Canada - United States border remains a significant barrier to international trade. For example, McCallum (1995) estimated that in 1988 Canadian provinces traded 22 times more with other Canadian provinces than with U.S. states of similar size and distance away. Given that trade tariffs between the two countries are now zero, and were low (approximately 4 percent on average) in 1988, McCallum's (1995) result and others like it is viewed as a border puzzle: the "border effect." The second related issue is that a number of empirical studies have concluded that the implementation of the N A F T A seems to have had no effect on international trade between Canada and the United States: the N A F T A effect. In fact, the American economist Paul Krugman has said that the impact of the N A F T A on Canada is zero (cited in Contenta 1996). I will argue in this chapter that neither the border nor the N A F T A effects are real. That they are thought to be real is a consequence in both cases of improper geographical, economic, and statistical specification. Once proper specification 118 is undertaken both effects disappear. I first discuss geographical misspecification, which applies mainly to the N A F T A effect, and then economic and statistical misspecification which applies to both effects. Most past research has investigated N A F T A at the national scale, and at that scale finds almost no effect. The problem, though, is that the appropriate scale to examine the effects of Canada - United States free trade agreements is not at the level of the nation state, but at the provincial/state level. The provincial/state level is the appropriate level of geographical specification because of North America's highly variegated economic geographical landscape. This varied economic landscape means that the Canada - United States border represents a different barrier to trade for the different provinces and states. The automobile industry, for example, has had low or zero tariffs since the mid-1960s and comprises of a significantly large portion of Ontario's international trade with the United States, over 35 percent. Consequently, tariff and non-tariff reductions through the C U F T A and/or the N A F T A did not affect the significance of Ontario's barrier to trade with the United States. On the other hand, British Columbia (trade in forestry products) and Quebec (trade in clothing and apparel) faced much larger barriers to trade before either of the free trade agreements, and so their removal had significant trade effects. Therefore, the appropriate spatial scale of analysis to examine the effect of the N A F T A is the Canadian province and U.S. state. As I wil l show below, once analysis of N A F T A is specified at the appropriate geographical scale the effect of N A F T A is significantly positive, and certainly not zero. Let me now turn to the economic and statistical misspecification, that is common to both the border and N A F T A effect. By economic specification I mean the choice of 119 variables used in the statistical analysis. International and interregional trade are complex phenomena and cannot be explained using only a handful of economic variables. But in many studies of both the border and NAFTA effects only two variables are employed: economic size and distance between trading partners. Two variables are simply not enough, however. Other variables must also be included such as the wealth of the trading partners, their endowment of capital and labour, tariffs, and currency volatility. Only if all of these variables are included will the border and NAFTA effects be isolated. If they are not included, and in most studies they are not because only size and distance are variables, there is misspecification, and as a result there is statistical bias and incorrect inference. Finally, the appropriate statistical specification (estimation procedure) is just as critical (perhaps more critical) than the appropriate economic specification. The research investigating the border and NAFTA effects use data that vary across both space and time, panel data. These type of data must be analyzed using the appropriate statistical estimation procedure (panel data estimation), otherwise statistical bias becomes an issue here as well. As should always be the case, the context of data analysis matters and dictates what must be done in order to avoid incorrect inference. My intention in this chapter, then, is to show that once Canada - United States interregional trade is correctly analyzed with respect to the scale of analysis, the set of economic variables, and the correct statistical estimation procedure, neither the border nor NAFTA effects are real. They are artefacts of geographical, economic, and statistical misspecification. The result is that first, the Canada - United States international border does not now, or in its recent history, impede international trade flows between Canada 120 and the United States, and second, that the NAFTA has a positive impact on Canada -United States international trade flows, but the effect varies across the Canadian provinces. ' , • The chapter is organized as follows. The literature regarding the border effect and its difficulties are reviewed in the next section. Section 3 presents the data used to investigate the border effect and the NAFTA, as well as the different methodologies that correspond to their respective literatures. The statistical results from the border effect investigation are discussed in section 4. The effect of the NAFTA is presented in section 5. Section 6, using all of Canada's international trading partners, discusses the appropriate direction of future research on Canadian trade. And section 7 concludes that Canada and the United States are indeed two integrated economies that have benefited (in terms of trade volumes) from their recent international trading agreements. 5.2. T H E BORDER E F F E C T The border effect entered the vocabulary of most economists and economic geographers who investigated Canada - United States international trade flows in 1995 when McCallum (1995) investigated the degree of integration between these two countries using a new international trade data set that measured international trade flows at the Canadian province and U.S. state levels in 1988. Combining these data with interprovincial trade data, McCallum (1995) was able to investigate whether Canada traded more with itself or with the United States after controlling for the economic sizes of the provinces and states as well as the distances between them using a gravity model of international trade flows. The gravity model of international trade flows states that trade 121 flows between larger economies will be larger than between smaller economies and that the greater the distance between the two economies the less they wil l trade. The intuition for these two prior expectations is straightforward: larger economies have the potential for more interaction because their economies produce greater volumes of more goods, and distance decreases interaction because the greater, the distance travelled, the greater the transportation costs. Therefore, in a statistical framework, the estimated parameters on the economic sizes of the trading economies (usually gross domestic product) are expected to be positive, and the estimated parameter on distance is expected to be negative. The gravity equation McCallum (1995) estimated is as follows: Inx^B.+P^nyf+B.Xnyj+fi.Dist + P.BE, (5.1) where x,j is the natural logarithm of trade between region i and region j, y, is the gross domestic product of region i, yj is the gross domestic product of region j, Dist is the distance between region i and region j, and BE represents the effect of the Canada -United States border on international trade flows. The border effect variable is a dummy variable equal to one i f the trade flows cross the Canada - United States border and zero if the trade flows are interprovincial. If Canada is more integrated with the United States than itself, the estimated parameter for the border effect wil l be positive. The interpretation of this border effect would then be that Canadian provinces trade more with the United States than other Canadian provinces after controlling for economic size and distance. However, i f the estimated parameter is negative, then Canadian provinces are more integrated with other Canadian provinces than the United States after controlling for economic size and distance. The expectation of the parameter was positive given the 122 - 5 1 high degree of integration between Canada and the United States. However, this was not the case. Rather than Canada having a positive estimated parameter indicating a high degree of integration with the Untied States' economy, the estimated parameter was negative, statistically significant, and large in magnitude. McCallum's (1995) border effect stated that Canadian provinces are more inclined to trade with themselves than they are with the U.S. states by 2100 percent, a factor of 21." Therefore, despite the large volumes of international trade flowing between Canada and the United States and their trading agreements either in force or about to enter into force, Canada in 1988 does not appear to be at all integrated with the United States. Further work by Helliwell (1996, 1998) confirmed McCallum's (1995) original finding, but had also showed that the magnitude of the border effect decreased to a factor of 12 by 1996, likely due to the further integration between Canada and the United States. This decreased negative magnitude itself is still unexpectedly high because of free trade between Canada and the United States was established seven years before. The explanation provided by Helliwell (1998) is grounded in the relative integration between Canada's provinces and between Canada and the United States. 3 1 In later research, Helliwell (1998) conducted a survey of economists and their students, finding that this view was dominantly held. 3 2 McCallum (1995) actually made the statement that it was a factor of 22, exp(p5) = 22, but that is the result of incorrect statistical inference in a semilogarithmic equation. As shown by Halvorsen and Palmquist (1980) and Kennedy (1981), the correct interpretation of a dummy variable in a semilogarithmic specification is exp(P5) -1=21. 123 Helliwell (1998) argues that there is a greater ease of economic operations within one country than between two or more countries because of factors such as a single nation having common institutions, a common currency, cultures, and information networks. And it is because of these commonalities that Canada is more integrated with itself than with the United States, leading to a negative estimated parameter for the border effect. However, a nation such as Canada is not without its own internal barriers (Doern and MacDonald 1999). In fact, the barriers to trade within Canada were so great that the federal and provincial governments of Canada negotiated an Agreement on Internal Trade that entered into force 01 July 1995. How could internal integration explain the Canada - United States border effect i f Canada needed to negotiate a free trade agreement with itself? Additionally, aside from the non-existence of tariff barriers between the Canadian provinces, the types of barriers to trade within Canada are quite similar to those between Canada and the United States. The division of responsibilities between federal, provincial, and local governments as well as preferences based on provincial residency and different product/safety/shipping standards have created interprovincial trade barriers within Canada (Leidy 1998; Schwanen 1992).3 3 These barriers, and their deleterious effects, have long been recognized by the different levels of Canadian government, but the push for policies to deal with these barriers did not begin until the mid-1980s. These policies took many years to come to fruition because of governmental focus on international free trade agreements over that same time period (Statistics Canada 1996). In 1994, the estimated cost of these barriers was $6.5 billion per year, enough to raise average income These authors do list other barriers, but these are the ones that are common for most goods and services. 124 in Canada by 1.5 - 6.5 percent depending upon province of residence (Palda 1994). Needless to say, any statements claiming a high degree of integration within Canada to explain the border effect must be made with caution. In fact, the barriers to interprovincial trade may actually exceed the barriers between Canada and the rest of the world. This is specifically the case for the United States due to the decreased trade barriers resulting from the establishment of two international free trade agreements (Loizides and Grant 1992). As a result of expectations regarding Canada - United States integration, the statistical evidence to the contrary, and a lack of adequate explanation for the border effect, those studying Canada - United States international trade flows have been puzzled. Consequently, a significant body of research has emerged investigating the border effect without any resolution, leading Obstfeld and Rogoff (2001) to call the border effect one of the six major puzzles of international economics. The work on border effects immediately following McCallum (1995) was dominated by extending the time frame of analysis to see i f the border effect has reduced over time and using different data sets to see i f the border effect is an artefact of Canadian trade data. More recent research on the border effects has focussed on the proper interpretation of the border effect coefficient, and the proper specification of the gravity equation in terms of economics, statistics, and geography. 5.2.1. Interpretation of the Gravity Equation Proper interpretation of the border effect has been somewhat problematic in a portion of the research using the gravity equation. Consequently, part of the puzzle of the border effect revolves around the misinterpretation of the border effect itself. 125 The first misinterpretation comes from incorrect statistical inference. As pointed out by Anderson and van Wincoop (2003), the gravity equation modelled by McCallum (1995), and those that followed, does not lend itself to standard statistical inference. For example, Brown and Anderson (1999) have claimed that Canada would have to export its entire gross domestic product to the United States i f the border effect were to be removed because export volumes to the United States would have to increase by a factor of 21. Rather, the magnitude of the border effect, as modelled, is better interpreted as the magnitude of change in the spatial distribution of existing trade flows favouring U.S. states (as opposed to Canadian provinces) i f the border effect were removed: interprovincial trade would fall and international trade to the United States would rise. Total provincial exports (interprovincial plus international) would not have to rise i f the border effect disappeared, though they likely would—Anderson and van Wincoop (2003) estimate that international trade flows would increase by 44 percent. The second misinterpretation of the border effect also results from a misunderstanding of the border effect in the context of an economically large and an economically small country. Suppose that the 10 Canadian provinces and 48 continental U.S. states all have the same economic size, and are trading with each other freely. Now, suppose a border separates the Canadian provinces and U.S. states using a policy instrument such as a tariff. The Canadian provinces now have a barrier to 83 percent of their potential trading partners, whereas the U.S. states only have a barrier to 17 percent of their potential trading partners—if actual economic sizes are used, the relative loss of potential markets for Canada is even larger. With such a large difference in potential market loss, we should expect that the border effect variable is significantly larger for 126 Canada (Anderson and van Wincoop 2003; Feenstra 2002). Therefore, the border plays the same role for U.S. states as it does for Canadian provinces, but it affects a larger proportion of the Canadian provinces' market, explaining why the border effect appears stronger simply because of which side of the border you are on. 5.2.2. Economic Specification of the Gravity Equation In addition to the improper interpretation of the border effect variable within the gravity equation, it is quite clear that the original border effect estimates suffer from specification error within the gravity equation itself—important variables are missing from McCallum's (1995) analysis as well as many researchers who followed in his stead. Anderson and van Wincoop (2003), as part of their solution to the border puzzle, state that in the border effect "estimation results are biased due to omitted variables" (Anderson and van Wincoop 2003, 170). They resolve the omitted variable bias through the inclusion of three variables measuring trade resistance derived through a theoretical framework. Their estimation procedure is far more difficult to implement than McCallum's (1995) specification, and Feenstra (2002) finds that fixed effects (dummy variables representing each bilateral trading pair) can account for these trade resistance variables while maintaining the ease of implementation used by McCallum (1995). However, there is a greater difficulty with their solution to the border puzzle. The theoretical framework used by Anderson and van Wincoop (2003) is based on the trade diversion hypothesis in its purest sense: any increase in cross-border trade flows implies a one-for-one decrease in Canadian interprovincial trade flows. Both Helliwell et al. (1999) and Couiombe (2003; 2004) strongly reject this assertion, finding 127 that interprovincial trade and cross-border trade are complements.3 4 The reason is clear: a dynamic efficiency argument. Once a province trades in those goods in which it possesses a comparative advantage with U.S. states, it wil l likely become more efficient thus fostering interprovincial trade (Trefler 2004). Regardless of this aspect of Anderson and van Wincoop's (2003) theoretical framework, or how one wishes to interpret the complementarity of interprovincial and international trade flows, Anderson and van Wincoop (2003) are on the right track with omitted variable bias and the border effect. Anderson and Smith (1999) added gross domestic product per capita and contiguity variables to the original McCallum (1995) gravity equation reducing the border effect from 22 to 15, a 31 percent decrease. This quite clearly shows the amount of bias an omitted variable can cause. Additionally, both of these variables are commonly employed in gravity models of international trade flows, and their inclusion theoretically justified in the intra-industry trade literature (Frankel et al. 1995). The effect of omitted variable bias found by Anderson and Smith (1999) demonstrates the importance of proper economic specification in the gravity equation. The two broad classifications of international trade flows (inter-industry and intra-industry trade)3 5 each have factors that determine (at least partially) the quantity of trade that flows between two regions. One of the common criticisms of the gravity equation is that it cannot differentiate between inter-industry and intra-industry trade—see Feenstra et al. (2001) that offers a method to differentiate between these trade types within the gravity equation—but this weakness in the gravity equation is likely the reason for its 3 4 The data used in this paper also show that interprovincial and cross-border Canada - United States trade are complements. See Table 5.4 below. 3 5 See chapter 4 for a discussion of the differences between these two types of trade. 128 continued success in explaining trade flows. And because the gravity equation can be theoretically derived for both trade types (see Anderson 1979; Anderson and van Wincoop 2003; Bergstrand 1985; Bergstrand 1989; Deardorff 1998; and Helpman and Krugman 1985), the determinants of both types of trade must be included in any proper specification. This includes the economic sizes of the trading economies (GDP), the geographic distance between them, the level of economic development of the trading economies (GDP per capita), the differences in both the economic sizes and development of the trading economies i f these variables are not entered separately, contiguity, trade barriers (tariffs and non-tariff barriers), relative factor endowments (capital-labour and land-labour ratios and their differences if these variables are not entered separately), currency volatility i f a time series or panel data are used in estimation, and variables capturing the presence of any free trade agreements. The economic sizes and geographic distances between trading partners have been part of the gravity equation for over 40 years (see Tinbergen 1962, Poyhbnen 1963, and Linneman 1966) and require no further discussion; economic development captures the degree of industrialization and the tendency toward intra-industry trade (Frankel et al. 1995; Greenaway and Milner 1986); the differences in economic size and economic development are also related to intra-industry trade, as similar countries have high degrees of intra-industry trade (Helpman 1987; Hummels and Levinsohn 1995); and relative factor endowments are another measure of differences between trading partners with increases in relative factor endowment differences decreasing intra-industry trade in favour of inter-industry trade (Hummels and Levinsohn 1995). 129 Trade barriers, particularly tariffs in the context of recent free trade agreements between Canada and the United States, are critical in any study of international trade flows over time. Feenstra (2002) finds that in O E C D countries the reduction in tariffs accounts for 38 percent of the increases in international trade flows. Tariff are clearly important in the explanation of the border effect because decreases in Canada - United States tariffs over the past 20 years have been significant. However, their inclusion along with the border effect variable can take place only within a particular estimation procedure because of perfect collinearity. Tariffs are. implicated in the research using sectoral trade data (see Brown 1998; Brown 2003; Brown and Anderson 2002; Hillberry 2002) to investigate the border effect of various industries in the Canadian economy, but are never employed in estimation despite their importance in analyzing trade flows. Finally, currency volatility should be included in any temporal context, particularly in a North American context because of the perceived loss of trade from currency value uncertainty (Grubel 1999). Exporters, facing uncertainty from currency volatility or profit-squeeze from the currency value, may substitute international trade for interprovincial trade. Given the recent volatility of the Canadian and U.S. dollar exchange rate, such effects cannot be ignored. However, as with tariffs measured at the national level, currency variables cannot be used in conjunction with the border effect in statistical estimation due to collinearity issues. The inclusion of variables representing free trade agreements capture the reduction of non-tariff barriers contained within the agreements. 5.2.3. Geographic Specification of the Gravity Equation 130 Recent research by economists, regional economists, and geographers has also revealed that there is a geography to the border effect itself. Anderson and Smith (1999) and Helliwell (1998), among other questions, ask whether or not the border effect is uniform across Canadian provinces. Though Helliwell 's (1998) analysis includes a temporal component—estimating the provincial border effects, .1988 - 1996—their results are qualitatively similar. Overall, the lowest border effects are found in western Canada, becoming moderate in eastern Canada, with the highest provincial border effects in the Maritime provinces and Newfoundland. To illustrate the magnitude of provincial variation, using the same data and estimation model as McCallum (1995), Anderson and Smith (1999) find that there is substantial differentiation: British Columbia has the lowest border effect, " o n l y " being biased towards interprovincial trade by a factor of 10, whereas Prince Edward Island has the largest in magnitude border effect with a factor of 49. The time trend of the provincial border effects are negative for most provinces, indicating a decrease in the bias toward interprovincial trade. Only Newfoundland and Prince Edward Island have increased the magnitudes of bias toward interprovincial trade over Helliwell 's (1998) time period, with Nova Scotia and New Brunswick exhibiting small decreases in their bias. Though this high degree of bias toward interprovincial trade by the Maritime provinces and Newfoundland is not surprising given their strong economic ties to central Canada, the relative rankings of western and central Canada is curious. Central Canada has historically maintained strong economic ties to the United States through international trade, particularly Ontario, whereas western Canada has not had its economic ties to the United States as strong. The western provinces do, however, have 131 significant trade volumes with the United States (oil and lumber), just not to the same degree as Ontario. To explain this unexpected result, Helliwell (1998) argues that the border effect for manufacturing-based provinces (central Canada) should be higher than for resource-based provinces (western Canada). Curiously, no justification is given for this claim and it does not appear to be related to tariff rates as the tariff rates in the resource industries are substantially higher than those for manufacturing industries. This is particularly true for automotive products, implying a lesser friction for manufacturing trade movements across the Canada - United States border. Anderson and Smith (1999) and Helliwell (1998) further investigate provincial border effects to find if exports and imports have the same border effect—the spatial distribution of import origins wil l not necessarily be the same for export destinations. With this separation, in 1988, British Columbia's propensity to export to Canadian provinces drops significantly to a factor of 5, with its propensity to import from Canadian provinces increasing to a factor of 20. This phenomenon leads Anderson and Smith (1999) to refer to British Columbia as an export platform, importing from Canadian provinces and exporting to U.S. states. But there is no way to know if there is any overlap in the goods that are imported and then exported. Because of British Columbia's relative lack of manufacturing (Barnes et al. 2001), it is rather doubtful that Anderson and Smith's (1999) reference to British Columbia is correct. The re-export of imports from other Canadian provinces is not likely occurring after any value-added production. Ontario, curiously because of the high degree of trade flows in automotive products with the United States, has the second highest border effect for exports making it appear to be more of an import platform—in 1996 Ontario's border effect for exports 132 was 18, whereas its border effect for imports was only 8. Though there has been significant restructuring in the automotive industry over the past 20 years (Holmes 1996), the close proximity of Ontario to the automotive industry's core in the United States makes the import platform argument for Ontario much more viable then the export platform argument for British Columbia. Though British Columbia, and the other western provinces, export many raw materials and manufactured goods imported from other Canadian provinces, relatively speaking Ontario better exemplifies itself as an export platform according to Helliwell 's (1998) findings. Overall, the rank of the provinces with respect to their export and import provincial border effects is similar to the total trade provincial border effects that increase moving east across the country— Helliwell 's (1998) temporal analysis essentially finds the same results. Therefore, for brevity's sake, in the analysis that follows, only the total trade provincial border effects are investigated. 5.2.4. Statistical Specification of the Gravity Equation Finally, there is the proper statistical specification. Matyas (1997; 1998) shows that the majority of research employing the gravity equation, including research investigating the border effect, is improperly specified. With this misspecification comes biased statistical estimates and, therefore, potentially improper inference. Specifically, Matyas (1997; 1998) calls for the use of as many as three fixed effect variables depending on the type of data used (cross-section, temporal, or panel): an origin country effect, a destination country effect, and a business cycle effect. Although they believe that heterogeneity (origin and destination country effects) must be modelled in a correct gravity equation specification, Cheng and Wall (2003) and Wall (2000) show that the Matyas (1997; 1998) specification leads to difficulties with residuals and in-sample predictions. Cheng and Wall (2003) and Wall (2000) propose trading-partner-pair specific effects to account for the heterogeneity. This specification has the advantage of not imposing the restriction that the intercept for the United States, for example, must be the same for all of its trading partners. With these fixed effects being specific to each bilateral trading relationship, this form of fixed effect controls for political, social, historical, and cultural factors that are all too difficult to measure. Also, because these fixed effects do not change over time, the problematic measurement of geographic distance does not need to be incorporated into the model because of perfect linear collinearity (Cheng and Wall 2003) (see Head and Mayer (2002) for a discussion of the difficulties with distance measurements). Therefore, coupled with the business cycle measurement proposed by Matyas (1997; 1998), the fixed effects of Cheng and Wall (2003) and Wall (2000) can easily be accommodated within most statistical and econometric software programs without their explicit inclusion in the data set (see 36 Kennedy (2003) for a good discussion of issues with panel data estimation). 5.3. D A T A AND ESTIMATION M E T H O D O L O G Y The data on Canada - United States trade flows are obtained from Statistics Canada (2004b) Canadian International Trade, 1988 - 2003. These data measure trade flows at 3 6 There is the additional complication of the presence of spatial autoconelation within border effect studies. Spatial autocorrelation does not bias statistical estimates, but it does artificially reduce variance when the spatial autocorrelation is positive. However, controlling for spatial autocorrelation with panel data is not possible within current standard statistical or econometric software and must be undertaken within programs that require user programming and is beyond the scope of my thesis. 134 the 8- and 10-digit Harmonized Tariff Schedule (HTS) level of aggregation for exports and imports, with Canadian provinces and U.S. states the smallest geographic units.37 All HTS classifications are aggregated based on their origin and destination. Interprovincial trade flow data has three sources: Statistics Canada (1998) Interprovincial Trade in Canada, 1984 - 1996, Statistics Canada (2000) Interprovincial and International Trade in Canada, 1992 -1998, and Statistics Canada's (2005a) Canadian Socio-Economic Information Management System (CANSIM) for the most recent interprovincial trade data. Provincial GDP, GDP per capita, and labour forces are also obtained through CANSIM. Provincial tariff rates are calculated using the duty collected variables within the Statistics Canada (2004b) Canadian International Trade, 1988 - 2003 publication. Provincial capital stocks are obtained through the Centre for the Study of Living Standards (2005) NAICS based Capital, Labour and Total Factor Productivity Tables by Province for Canada, available from the CSLS on-line. Arable land is obtained from Statistics Canada's (1986, 1992, 1999a, 2002, 2005b) censuses of agriculture. These data are available at the provincial level, but only at 5-year intervals—intervening years are obtained using linear interpolation. U.S. state GDP and GDP per capita are obtained from the Bureau of Economic Analysis and U.S. state labour forces are obtained from the Bureau of Labor Statistics, all available on-line. Arable land estimates for the U.S. states are obtained from the United 3 7 One potential problem with using provincial level data for measuring trade origins and destinations is locational bias: a good may leave Canada from British Columbia and be recorded as such, but originated from Alberta. Statistics Canada has mitigated this potential for error by recording the actual province/state of origin. 135 States Department of Agriculture (2000) Summary Report: 1997 Natural Resources Inventory (revised December 2000). As with the Canadian data, arable land for the United States is only available at 5-year intervals, with the intervening years obtained through linear interpolation. U.S. state capital stock estimates are not available, nor are U.S. state total factor productivity estimates which can be used to calculate state-level capital stocks. In order to calculate the capital stocks for each U.S. states, the annual United States capital stock is distributed based on each state's share of U.S. GDP. This same technique was used to calculate capital stocks for Canadian provinces to test the similarity of this estimation to actual data. For the Canadian data, the two measures of provincial capital stock have an extremely high and statistically significant correlation coefficient, r = 0.969. Therefore, the estimates for U.S. state capital stock cause little concern for bias in any estimation. The geographic distances between trading regions are measured in the number of kilometres between the principal centres in each region using Google Earth and Microsoft MapPoint. 5.3.1. The Border Effect Specification The variables listed above are estimated in the following gravity equation specification to estimate the border effect: In xtj =BX+B2 In y. + B3 In y. + B4 In Dij + B5 In yct + B6 In yCj + B, In AX. Z2001 ' k-m9a><BE* where xtj is the bilateral trade flow between region i and region j, y, and y7 are the G D P of regions i and j, Dy is the geographic distance between regions / and j (used only in the sensitivity analysis, below), yc, and yq are the G D P per capita of regions / and j, KL, and 136 KLj are the capital-labour ratios of regions i and j, TLj and TL) are the land-labour ratios of regions i and j, Tarifffs the provincial level tariff measuring the level of protection for each province, NAFTA is a dummy variable that is one after 1993 and zero otherwise, and BEk is the border effect for the years 1989 - 2001. The border effect is not modelled for 1989 in the panel estimations and the currency volatility and contiguity variables are not in any of the models due to collinearity issues. 5.3.2. The NAFTA Effect Specification The primary difference between the border effect specification and the N A F T A effect specification is in regard to the way the independent variables are measured. This difference is an attempt to remain as consistent as possible with the past literatures within each topic, while correcting for their limitations. In the border effect specification, the variables representing each trading economy are measured as consistently representing the same country. For example, in the context of trade between British Columbia and Washington, if British Columbia was region / for the G D P variable, then British Columbia was also region / for G D P per capita. This is not necessarily the case for the N A F T A effect. Rather, to be consistent with Hummels and Levinsohn (1995), for example, the two G D P variables wil l represent their economies depending on their relative sizes, large versus small, GDPLarge and GDPsmai i -For example, suppose that Washington has a larger G D P than British Columbia, but British Columbia has a larger G D P per capita than Washington. 3 8 In this example, Washington would be represented by G D P L a r g e and G D P per capitasm aii, whereas British Columbia was be represented by GDPsmai i and G D P per capitaLarge-Washington's GDP and GDP per capita are in fact both larger than for British Columbia. 137 Regarding statistical interpretations, differences in G D P decreases intra-industry trade, so GDPi_arge is expected to have a negative coefficient because it represents an increasing difference between the two trading economies; consequently, GDPsmai i is expected to have a positive coefficient because, holding GDP L a r ge constant, increasing GDPsmai i decreases the difference in economic size between the two trading economies. In this context, Ontario's G D P wil l always be measured as GDP L a r ge for interprovincial trade flows, but wil l be G D P s m a i i for some trading relationships with the United States. The same representation is used for G D P per capita and the relative factor endowment variables. This leads to the following gravity equation specification: In x..t =px + 82 In y M f i G £ , + B3 In ySMALLl + B4 In y c M S G £ , + B5 In ycSMALLt + m KLLARGEt + B7 In KLSMALLl + Bs I nTL L A R G E t + B9 InTL S M A L L t , (5.3) + Bw In ER, + BnTariff, + Bnt + B13NAFTA where xy, is the bilateral trade flow between trading economies i and j at time t, yiARCEt and ysMALu are the GDPs of the large and small trading economies at time t, yciARCEt and ycsMALu are the G D P per capita measures of the large and small trading economies at time t, KLiARGEt and KLSMALU are the capital-labour ratios for the large and small trading economies at time t, TLLARGEI and TLSMALU are the land-labour ratios for the large and small trading economies at time t, ER, is the exchange rate between Canada and the United States at time t with interprovincial trade given an exchange rate value of one, Tariff, is the average Canada - United States tariff rate at time t, t is the linear trend, and NAFTA is a dummy variable that is one after 1993 and zero otherwise. The N A F T A variable takes on a number of forms in the analysis below depending on the regression context. Four regression models are undertaken in this analysis, two 138 using Canada - United States international trade flows and two using both interprovincial and international trade flows. In the former analysis, the N A F T A ' s effects are investigated only for changes in the Canadian trading patterns with the United States. This is similar to most previous analyses, less non-NAFTA member countries. In the latter analysis the N A F T A ' s effects are investigated both on changes in the Canadian patterns with the United States and with other Canadian provinces. Any significant change in the trading patterns, positive or negative, with the United States may impact interprovincial trade flows. The first analysis is to have a baseline investigation similar to that of past research in the effects of the N A F T A and the second analysis is to extend that investigation to uncover whether or not the N A F T A has impacted interprovincial trade flows. The effects of the N A F T A are measured at the national level and at the provincial level to investigate the geography of the effects of the N A F T A in both analyses. Equations 5.2 and 5.3 are estimated using a fixed effect panel procedure. A l l Canadian provinces (no Territories due to a lack of explanatory variables) and all U.S. continental states are included in the data set—the District of Columbia is treated as a separate spatial unit of analysis and is therefore considered a U.S. state in the analysis similar to that of Coughlin and Wall (2003). A l l zero values of bilateral trade are replaced with one dollar to facilitate the natural logarithm. This allows for all possible trading relationships to be modelled without any serious concerns for bias (Eichengreen and Irwin 1995, Greene 2000). A l l variables except the average tariff rate, linear trend, and N A F T A variables are measured in natural logarithms to ease interpretations: all of the coefficients on these variables are then elasticities, representing the percentage 139 change in the dependent variable given a one percentage change in the independent variable. A l l estimation is performed using NLogit 3.0. 5.4. B O R D E R E F F E C T R E S U L T S 5.4.1. Standard Gravi ty Equation Results The results of estimation are shown in Table 5.1 (standard gravity equation variables) and Table 5.2 (border effect variables). The adjusted R2 for the OLS and random effects models are on par with previous research on the border effect, with a large jump in the adjusted R2 for the fixed effects model. The estimated parameters for GDP, G D P per capita, and the factor endowment variables, when significant, are generally similar in sign and magnitude across the different models, and the N A F T A dummy variable is significant and positive in the fixed and random effects models but negative and significant in the OLS model. The provincial tariff parameters are always positive and significant, which is a curious result because tariffs typically have a strong negative impact on trade flows. <See Table 5.1, page 271> With regard to expectations, the signs and magnitudes of the estimated G D P parameters are consistent with previous border effect results. G D P per capita, however, is not consistent across the three models. In the panel regressions, G D P per capita is similar to that found by Anderson and Smith (1999) with one region's G D P per capita having a positive effect and the other region's G D P per capita having a negative effect on the level of trade. In the OLS regression, the estimated parameters for both region i and j are positive and significant. The negative estimated parameters for the capital-labour 140 ratios are consistent with intra-industry trade expectations given that differences in relative factor endowments decrease intra-industry trade, and the positive estimated parameters for the land-labour ratios are consistent with inter-industry trade expectations because those differences in relative factor endowments increase inter-industry trade. The coefficients for the capital-labour and land-labour relative factor endowment variables are expected to be consistent with each other, capturing the dominant form of trade. In the case of Canada and the United States, all of the coefficients are expected to be negative given that intra-industry trade dominates Canada - United States trade flows at the national level. However, the degree of intra-industry trade varies substantially across the Canadian provinces. Therefore, the inconsistency is likely the result of intra-industry trade dominating in central Canada and inter-industry trade dominating in other areas of the country. Capital-labour ratios are capturing the intra-industry trade in central Canada and the land-labour ratios are capturing the inter-industry trade elsewhere in Canada. With regard to the N A F T A variable, its negative coefficient within the OLS framework is not expected. However, the positive and significant N A F T A coefficients in the panel data models do conform to expectations: only a moderate positive impact on Canada - United States international trade, because the decreases in trade barriers between the two countries were implemented through the Canada - United States Free Trade Agreement in 1989. Though these structural variables are not the primary concern of this paper, overall the fixed and random effects models fare better than the classical OLS model with respect to the expectations on the estimated parameters. The F-test for a panel regression strongly rejecting (p-value < 0.001) the classical OLS model in addition to the better fit 141 with expectations only strengthens the case for panel estimation made by Cheng and Wall (2003) and Wall (2000). 5.4.2. Border Effect Results The border effects for the three estimated models are reported in Table 5.2. For ease of explanation, the notation of the border effect is presented differently than previous research. When the border effect reduces Canada - United States trade it is represented as negative (the same sign as its estimated parameter) and i f the border effect increases Canada - United States trade it is represented as positive. This notation has not been necessary in previous research because the border effect had always reduced Canada -United States trade. The results from the OLS model are qualitatively similar to the findings of Helliwell (1998), with the border effect decreasing over time. The border effect reaches a high of 99 in 1991 and falls to 13 by 2001. Though in line with previously estimated border effects by 2001, the 1991 magnitude is incredibly high after controlling for GDP, economic development, factor endowments, tariffs, and the N A F T A . However, the border effects in the OLS model may be higher than past research due to the inclusion of all bilateral trading partners between Canada and the United States, rather than only the economically large U.S. states and the U.S. border states. Nevertheless, the OLS results presented here are qualitatively similar to that of past border effect research, favouring interprovincial trade and decreasing over time, showing that the following results are not an artefact of particular data. <See Table 5.2, page 272> 142 Turning to the fixed and random effects models, the results are radically different. In the fixed effects model, all of the estimated parameters are positive implying that after controlling for the standard gravity equation variables Canadian provinces trade more with U.S. states than other Canadian provinces. The magnitude of this positive border effect steadily increases over the study period, becoming statistically significant in 1992 and reaching a maximum value of 1.94 in 2001. Therefore, both at and shortly after the establishment of the Canada - United States Free Trade Agreement (CUFTA) in 1989, there was no positive or deleterious effect of the Canada - United States border on trade flows relative to Canadian interprovincial trade. Only in 1992 after tariff reductions and institutional changes had taken place did the C U F T A have a positive impact on Canadian trade flows to the United States. By 2001, Canadian provinces traded approximately 194 percent more with U.S. states than other Canadian provinces after controlling for other structural variables. Not only are these results in magnitudes that are reasonable, but they are consistent with expectations regarding the establishment of free trade agreements. The random effects model has slightly different results than the fixed effects model. In 1990, the border effect does start out negative implying that Canadian provinces trade 43 percent less with U.S. states than other Canadian provinces, a number comparable to that found by Anderson and van Wincoop (2003). The border effect becomes statistically insignificant in 1992, and positive in the following years reaching a maximum of 0.871 in 2001 implying that Canadian provinces trade 87.1 percent more with U.S. states than other Canadian provinces after controlling for other structural variables. The interpretation of these results is similar to that of the fixed effects model. Before and shortly after the establishment of the C U F T A , the trade barriers between 143 Canada and the United States posed a modest resistance to Canada - United States trade flows. Once those trade barriers were reduced during the first few years of the C U F T A that resistance became insignificant. Eventually, the tariff barriers became low enough to increase provincial trade with U.S. states relative to other Canadian provinces. 5.4.3. Sensitivity Analysis The incorporation of panel estimation techniques have not eliminated a border effect for Canada - United States international trade flows, but reduced its magnitude to a reasonable level and changed the sign of the estimated parameter to what would be expected for Canada and the United States with their long trading history. Clearly, the existence of a large and negative border effect in other studies for Canadian province and U.S. state trade is a result of some sort of statistical bias. However, the source of this statistical bias, as yet, is unknown. Is the initial specification of McCallum (1995) the source of statistical bias through omitted variables? Or is it the estimation procedure itself? As shown below, the answer to both of these questions is yes. <See Table 5.3, page 273> McCallum's (1995) estimate of the border effect, and those that followed his lead, does suffer from omitted variable bias, but that bias is not the primary source of the error. In fact, the bias induced from the omitted variables reduced the negative magnitude of these border effect parameters. This is shown in Table 5.3: the McCallum specification is y„ yj, and Dy, whereas the full specification is the set of variables in Table 5.1 plus Dy. The estimated border effects in the OLS McCallum specification are similar to those found in previous research with a border effect of 30 in 1989 falling to 14 by 2001, somewhat higher than those found by McCallum (1995) and Helliwell (1998) due to the 144 inclusion of all U.S. states in the present specification. However, in the OLS full specification, the border effect jumps to over 200 in 1989 falling only to 26 by 2001. The inclusion of theoretically-informed variables to explain the pattern of trade between Canada and the United States has not only changed the value of the estimated border effects, but made their magnitudes ridiculous. Surely i f these values of the estimated border effect had been reported the issue would not have garnered such a high degree of attention. Instead, the border effect would have been dismissed as problematic specification. That specification problem, moreover, is statistical not economic. In a panel data set, if any cross-section heterogeneity is not accounted for, the remaining estimated parameters suffer from omitted variable bias (Kennedy 2003). Consequently, in the presence of cross-section heterogeneity, the appropriate statistical specification must be used to avoid potentially misleading estimation results. This is the primary source of the bias in previously estimated border effect parameters. As shown in Table 5.3, the fixed effect models for both the McCallum and full specifications have estimated border effect parameters that are insignificantly different from zero until 1991, then positive and significantly greater than zero afterward (the random effects model results are qualitatively similar, but lower in magnitude as in Table 5.2, so they are not presented here for the sensitivity analysis). Therefore, though proper economic specification is an issue with the border effect, the primary difficulty with past estimates is the statistical specification. To be fair, McCallum (1995) did not have the ability to control for cross-section heterogeneity because he had only one year of trade data. Subsequent work did not suffer from this limitation, however. As a result, all of the cross-section heterogeneity was 145 forced into the estimated border effect parameter, causing it to be large in magnitude and negative. 5.4.4. The Geography of the Border Effect The results for the provincial level border effects are qualitatively similar to that of the national border effects (see Table 5.4). Prior to 1992 for most provinces the border effects are negative, significant, or both. They become positive and significant by 1993, except in Newfoundland. As stated above, in past research the western provinces exhibited significantly less bias toward interprovincial trade than the provinces in central Canada. This pattern is similar in the present analysis with the western provinces now exhibiting a stronger bias toward interregional trade with the United States than central Canada. The Maritime provinces, however, exhibit a stronger bias toward interregional trade than do both Ontario and Quebec. This result for the Maritime provinces is qualitatively different (aside from the positive provincial border effect that is already qualitatively different) from past research. For Helliwell (1998), the bias toward interprovincial trade increased eastward across Canada. M y figures show, in contrast, that the "b ia s" moves from the western provinces to the Maritime provinces and then back to central Canada. <See Table 5.4, pages 274 - 275> Newfoundland and Prince Edward Island are the anomalies in the present analysis. Newfoundland's provincial border effect, aside from 1991 and 1992, is always insignificant and is dominantly negative. Therefore, neither the Canada - United States Free Trade Agreement nor the North American Free Trade Agreement have significantly altered Newfoundland's bias toward interprovincial trade. 146 Prince Edward Island's results are opposite to that of Newfoundland. It now has the strongest bias toward interregional trade with the United States. This is the complete opposite to past research that found Prince Edward Island exhibited the strongest bias toward interprovincial trade. Though Prince Edward Island's provincial border effect was of similar magnitude with the other provinces until 1994 and 1995, the magnitude of its provincial border effect has since more than tripled indicating that Prince Edward Island now trades with the U.S. states 367 percent more than the average for other Canadian provinces after controlling for economic size, economic development, relative factor endowments, tariff rate changes, and distance. However, given the nature of Prince Edward Island's international trade flows with the U.S. states (relatively low trade volumes compared to other Canadian provinces) this change should be interpreted with caution. 5.5. NAFTA EFFECT RESULTS 5.5.1. Canada - United States International Trade and the NAFTA Turning to the effects of the N A F T A , the statistical results from the four regression models are presented in Tables 5.5 - 5.8. Table 5.5 shows the results from the Canada -United States international trade flow data only measuring the effect of the N A F T A at the national level for a more direct comparison to past research. The estimated elasticities for GDP, G D P per capita, and the relative factor endowment variables are generally consistent with expectations. When the "Targe" and ""small" variables are significant the large variable is negative and the small variable is positive (except the land-labour ratio for the small trading economy), generally being consistent with the dominant 147 portion of intra-industry trade for Canada. None of these elasticities are particularly high or low in magnitude. The exchange rate and linear trend coefficients have their expected positive signs with the exchange rate elasticity being particularly high in magnitude indicating the gains in trade from a weaker Canadian dollar. The average tariff rate, however, has a positive and significant coefficient, a curious result similar to that found above with regard to the border effect. <See Table 5.5, page 276> The national effect of the N A F T A is positive and statistically significant, contrary to most of the past research. However, it should be noted that the key difference between this national effect and most of the past research on the effects of the N A F T A is that the current result is based on province-state trade flows rather than Canada - United States aggregate national trade flows. Nevertheless, the impact of the N A F T A on Canada United States international trade flows has been a 16.32 percent increase over and above any increases from the C U F T A . This result is more modest than that found by Wall (2003), but his analysis separates exports and imports for Canadian and U.S. regions rather than dealing with total trade flows between provinces and states—neither type of analysis is superior to the other, they merely address different questions. Turning to the geography of the effects of the N A F T A on Canada's trade flows to the United States, Table 5.6 exhibits significantly different results from Wall (2003)—the N A F T A variable in this regression model measures the impact of the N A F T A on each provinces' trade flows with the United States as a whole. The G D P , G D P per capita, relative factor endowment, exchange rate, average tariff rate, and linear trend variable 148 coefficients are all similar to that shown in Table 1, but the effect of the N A F T A varies significantly across the Canadian economic landscape. <See Table 5.6, page 277> Except for British Columbia, all of the provincial N A F T A effects are positive, contrary to that found by Wall (2003). Even the Maritime provinces and Newfoundland, all exhibiting negative effects from the N A F T A in Wall 's (2003) analysis, have positive coefficients with Prince Edward Island's coefficient being statistically significant and high in magnitude. As a result of the N A F T A , Prince Edward Island's international trade flows with the United States have increased by more than 76 percent. Though Prince Edward Island's international trade flows with the United States are a small share of Canada's total trade flows with the United States, this result exemplifies the impact on results from changing the spatial scale of analysis. Turning to central Canada, the overall results presented in this analysis are qualitatively similar to that found by Wall (2003), but only Ontario has experienced a positive effect from the N A F T A , an increase of trade flows of almost 16 percent, whereas Quebec's positive effect from the N A F T A is highly insignificant. The effect of the N A F T A for Ontario is also significantly more modest than that found by Wall (2003). The 16 percent effect found here is less than both of the effects found by Wall (2003) for exports and imports, 43 and 18 percent, respectively. The national effect of the N A F T A found in this analysis is expected to be within Wall 's (2003) export and import effects given that aggregate trade effects should be a weighted average of the independent effects on imports and exports. Western Canada also differs significantly from Wall (2003) in that the overall impact of the N A F T A on western provinces is positive. British Columbia and Alberta exhibit insignificant effects from the 149 N A F T A , but both Saskatchewan and Manitoba exhibit significant positive effects on international trade flows to the United States, 39.34 and 14.46 percent, respectively. The overall result presented here is that the N A F T A has had a positive impact on trade for the individual Canadian provinces. British Columbia's N A F T A effect is negative but in such a low magnitude that it is inconsequential. Not all of the provinces with significant and positive N A F T A effects are significantly impacted by the non-tariff changes in the N A F T A . Ontario, for example, has been affected because of changes in the rules of origin in the automotive sector. A natural question to ask then, is what is the driving force behind these changes? As discussed above (chapter 3) in the context of trade creation and diversion, the new economic geography literature provides insight here. The N A F T A altered the mean centre of the consumer market for Canada and the United States moving it further south through the inclusion of Mexico in the N A F T A . Because of the change in the mean centre and the corresponding possibility of firms changing locations within the same country or to another country in response to this new mean centre, post-NAFTA trade flows are altered. Rather than a province trading with another province, that province now trades with the United States because of firms moving closer to the mean centre of their market. In order to better understand this positive effect of the N A F T A on the provinces' trade flows, the incorporation of interprovincial trade flows is needed to investigate whether or not the N A F T A has negatively impacted Canada's interprovincial trade. In general, the present analysis again exemplifies the impact of aggregating provinces and states into larger sub-national regions to assess the N A F T A , or any other policy change. As found by Coughlin and Wall (2003) regarding the effects of the 150 N A F T A on U.S. states, aggregating U.S. states to larger sub-national regions conceals the distinctive effects of the N A F T A on the spatial units within each of those sub-regions. 5.4.2. International and Interprovincial Trade and the NAFTA Turning to the international-interprovincial results, presented in Tables 5.7 and 5.8, all of the non-NAFTA variables in both regression results are quite similar to the two regression models discussed above, so they will not be discussed. The results for the effects of the N A F T A on interprovincial and international trade flows to the United States, however, are worthy of independent comment. The effects of the N A F T A on international trade flows to the United States are positive, statistically significant, and very close in magnitude to the effect of the N A F T A shown in Table 5.5—the effect is now a 14.65 percent increase in international trade flows to the United States in addition to the effects from the C U F T A . The effect of the N A F T A on interprovincial trade is negative but not statistically significant. Therefore, at the national level, the effect of the N A F T A has been to increase international trade flows to the United States without significantly decreasing the interprovincial trade flows. <See Table 5.7, page 278> The geography of the effects of the N A F T A on international trade flows to the United States and Canadian interprovincial trade flows are shown in Table 5.8. There are both similarities and differences between the results using only international trade flow data and using international and interprovincial trade flow data (comparing Tables 5.6 and 5.8), but where the variables in each of the models overlap the results are strikingly similar. The primary difference is that now the effect of the N A F T A on Manitoba's international trade flows with the United States, though still positive, is statistically 151 insignificant. The interesting result that emerges out of the use of both international and interprovincial trade data is the effect of the N A F T A on interprovincial trade. <See Table 5.8, page 279> Of the ten effects of the N A F T A on interprovincial trade flows, nine of them are statistically insignificant, some positive and some negative, generally confirming the national statistically insignificant result. British Columbia, however, is negative, statistically significant, and high in absolute magnitude. The effect of the N A F T A has been to decrease British Columbia's interprovincial trade flows by almost 28 percent. This is a curious result because British Columbia has increased its interprovincial trade over this study period by 27 percent, significantly more than Ontario's and Quebec's meagre increases of 7 and 8 percent, respectively—it should be noted that Ontario and Quebec do have quantitatively large estimated decreases in interprovincial trade flows resulting from the N A F T A , but those decreases are not statistically significant. British Columbia has also decreased its share of total trade to and from other Canadian provinces, but those decreases are common for all Canadian provinces and British Columbia's decreases are not exceptionally high: Alberta, Saskatchewan, Ontario, Quebec, and Nova Scotia all have greater relative decreases in their shares of interprovincial trade flows (Statistics Canada 1998; 2000; 2005a)—and the changing pattern of British Columbia's interprovincial trade flows now focusing more on geographically closer provinces found in previous chapters is no different than any of the other Canadian provinces. Once again, the utility of using individual provinces and states is manifest, rather than using aggregated sub-national regions for the analysis. Overall, the effects of the N A F T A are positive for the Canadian provinces with regard to 152 international trade flows to the United States, without any significant loss of interprovincial trade aside from the loss for British Columbia. Though the geographical pattern of the effects of the N A F T A on both international trade flows to the United States and interprovincial trade flows is interesting in itself, the causes of this geographical effect is also of particular interest. However, the causes behind the geographical effects are more difficult to ascertain than the effects themselves. However, with the majority of provinces decreasing their interregional trade and increasing their international trade (most of which are statistically insignificant, however) the possible explanation of firms changing locations due to the new mean centre of the consumer market for Canada and the United States has gained merit. If large Canadian firms locate further south (even only just across the Canada - United States border), former interprovincial trade becomes international trade. It is certainly the case that the mean centre of the consumer market shifted further south with the C U F T A , but the N A F T A has shifted the mean centre even further south, providing more incentive (and likely more firms) to move into the United States, i f not Mexico. Most of the statistical results are not statistically significant, so this spatial shift is not likely incredibly strong, but appears to be present nonetheless. Unfortunately, in order to authentically test this hypothesis, firm-level data that includes their (changing) locations over time are required but not available. However, some other hypotheses can be investigated with the available data. <See Table 5.9, page 280> As shown in Table 5.9, most of the correlations between the provincial effects of the N A F T A on international and interprovincial trade flows and the economic 153 relationships between Canada's provinces and their trading partners are statistically insignificant. Despite this insignificance, the signs of the correlation coefficients are instructive. Those provinces with the lowest export shares, import shares, and export-G D P ratios in 1989 had the largest impact from the N A F T A on trade, as well as those provinces that had the highest export shares, import shares, and export-GDP ratios in 2001 had the largest impact from the N A F T A on trade. Similarly, those provinces with the highest 1989 tariff rates had the greatest positive impacts from the N A F T A on trade. Therefore, those provinces with the least amount of economic ties in 1989 and the most protection through tariffs gained the most in trade flows once barriers to trade were reduced, and those provinces that had the highest amount of economic ties in 2001 gained the most in trade flows. These relationships, though insignificant, do make sense because the provinces that were able to increase their trade (1989 correlations) and did increase their trade (2001 correlations) with the United States experienced the greatest gains in trade from the N A F T A . The difficulty in this interpretation is that many of the changes in export shares, import shares, export-GDP ratios, and tariff rates are due to the CUFTA—there is no control group for this analysis to investigate independently the effect of the N A F T A . Consequently, with the limitations of available data, the most conclusive evidence for the effect of the N A F T A is that it has had a strong geographical impact on the Canadian economic landscape. Though circumstantial, those geographical effects appear to be related to the degree of the province's economic ties with the United States and the rest of Canada before and after the N A F T A entered into force. 154 5.6. T H E DIRECTION OF F U T U R E R E S E A R C H ON CANADIAN T R A D E The natural question to ask at this point is: what is the appropriate direction for future research on Canada - United States international trading relations? As shown in Table 5.10 using an increase factor,39 total international trade flows have increased in real terms to all regions of the world for most provinces—the increase factors for imports and exports are also provided. Aside from the Territories that have seen substantial decreases in international trade flows to Mexico, East Asia plus Australia and New Zealand (EA-ANZ) , 4 0 and the rest of the world (ROW), decreases in aggregate international trade flows are few and of low magnitude: British Columbia and Prince Edward Island have decreased their total international trade flows with the European Union (EU15),41 with those decreases both being from a decrease in exports—imports from the European Union to both British Columbia and Prince Edward Island have increased over the study period. New Brunswick has decreased its total international trade flows with East Asia, also from a decrease in exports to that region. And Nova Scotia has decreased its international trade flows with Mexico and East Asia, both from decreases in imports from those regions. <See Table 5.10, pages 281 - 283> 3 9 The increase factor is measured as the ratio of 2001 and 1989 trade levels in constant dollars. 4 0 East Asia is defined as the ASEAN 10: Brunei, Cambodia, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Thailand, Vietnam, plus China, Hong Kong, Taiwan, South Korea, and Japan plus Australia and New Zealand. 4 1 European Union-15: Austria, Belgium-Luxembourg, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, Spain, Sweden, United Kingdom. 155 Turning to international trade flows between and within Canada and the United States, all of the Canadian provinces have increased their interprovincial trade levels in real terms, by an unweighted average of 23 percent. There are some decreases in the level of interprovincial trade in some cases, but those decreases are small and aside from Ontario's small decrease in interprovincial exports are in the smaller east coast provinces and the Territories. In contrast, Canadian provinces have increased their levels of international trade with U.S. states by an unweighted average of 153 percent, with all of the provinces except Newfoundland and the Territories more than doubling (in some cases almost quadrupling) their international trade flows with the United States. Thus far, it appears as though the most fruitful analyses of Canadian international trade flows are with the United States. <See Table 5.11, pages 284 - 285> Analysing the shares of international and interprovincial trade flows only supports the above statement (see Table 5.11). The shares of international trade to East Asia fell for all provinces except Manitoba, Ontario, and Newfoundland. 4 2 These three provinces did exhibit increases in their trade share to these regions, but those increases are all very modest. This is somewhat of a curious result given the attention East Asia has garnered as a force in the international economy. Even British Columbia has decreased its share of international trade flows with East Asia. Only having international trade flows with the United States have risen. British Columbia has increased its share of imports from East Asia, but its share and level of exports have fallen. On the international front, the rest of the world has lost total trade shares with all provinces except Quebec, Nova Scotia, and It should be noted that this time period includes the 1997 Asian Crisis. 156 Newfoundland, and the European Union has lost total trade shares from British Columbia, Ontario, the Maritimes, and Newfoundland. At a more local level, there is a clear winner and loser when it comes to interprovincial and international trade shares. For all of the provinces, total interprovincial trade shares have fallen. On the import side, Prince Edward Island very modestly increased its share of interprovincial trade, and for exports only British Columbia and Newfoundland increased their interprovincial trade shares, with Newfoundland's increase in interprovincial trade shares being substantial, from 21 to 40 percent of all of its trade. Turning to the trade shares with the United States, all of the provinces have increased their total trade shares with the United States. Prince Edward Island slightly decreased its already low share of imports from the United States, and Newfoundland and the Territories both decreased their exports to the United States. Additionally, in 1989 only Ontario had a larger trade share with the United States than with Canadian provinces, but by 2001 only Saskatchewan, Nova Scotia, Prince Edward Island, Newfoundland, and the Territories had a greater share of their trade with Canadian provinces, with almost all provinces having significant increases in their shares of trade with the United States. Despite that increases in international trade outside of North America and interprovincial trade may lead to increases in the economic welfare of Canadian provinces, of increasing importance for possible provincial economic welfare gains is their trade flows with the United States. Accordingly, less attention should be focussed on the international trading patterns outside of North America with more attention being focussed on province-state trading patterns. The trading relationship for each Canadian 157 province has its own unique geography with the United States. Only through a detailed analysis of each of those trading relationships wil l a complete understanding of the complex geographical trading relationship between Canada and the United States become manifest. 5.7. CONCLUSIONS For over ten years how the finding of a border effect between Canada and the United States has been a major puzzle in international economics and the N A F T A has not been shown to definitely have had an impact on Canada - United States international trade flows. I've argued in this chapter that research showing both such effects was wrong. Because of improper geographic, economic, and statistical specification, previous estimates of the border effect and N A F T A effect were biased, and which produced incorrect inferences. In the case of the border effect, the solution does not stem from imposing a new theoretical framework. It involves taking the gravity equation and then modelling with theoretically-informed variables commonly used in most empirical international trade research along with an appropriate statistical procedure and proper statistical inference for a semilogarithmic equation. Using this approach I have uncovered two facts about the integration of Canada and the United States. First, the border effect is insignificant or negative in low magnitude following the establishment of the C U F T A and then moderately positive at or after 1992. So, in 1990, interprovincial trade is favoured over province-state trade by approximately 7 - 4 3 percent. But by 2001 province-state trade is favoured over interprovincial trade by approximately 100 - 200 percent, depending on whether the fixed or random effects 158 model is used. These results are in a direction and at a magnitude that are not surprising within the context of the free trade agreements between Canada and the United States. And second, the N A F T A has indeed had a positive effect on Canada's international trade flows with the United States. Additionally, the N A F T A has been shown to have had a statistically insignificant impact on overall interprovincial trade flows and that the geographical effects of the N A F T A are best viewed at the provincial level. Using the Canadian provinces and U.S. states as the spatial units of analysis, it is found that the geographical effects of the N A F T A are overwhelmingly positive for international trade flows to the United States and that only British Columbia has experienced a negative change in interprovincial trade flows resulting from the N A F T A . With these two issues regarding the integration of Canada and the United States and the appropriate direction of research regarding Canada - United States international trade flows clarified, my analysis now turns to the geographical dimensions of Canada's trade with the United States. This analysis begins with an account of the trading patterns of the Canadian provinces with other Canadian provinces and the United States. This is then followed with an investigation into the degree of spatial change in the Canadian provinces' trading patterns in the twelve years since the inception of free trade between Canada and the United States. Chapter 6 159 Canada - United States interregional trade flows, 1989 - 2001 6.1. I N T R O D U C T I O N Now that the appropriate direction of research on Canadian trade has been established in the previous chapter, the analysis now moves into a regional account of the trade flows of Canadian provinces. This chapter presents and examines the trading patterns between Canadian provinces and other Canadian provinces as well as between Canadian provinces and U.S. states.43 Using interregional trade volumes (measured in constant dollars), interregional trade shares, and economic dependence through interregional trade, the general trading patterns of the Canadian provinces are presented. As stated in the introduction, most of the attention geographers have paid to international trade flows has been at the nation-nation scale (see Dicken 2003; Gaile and Grant 1989; Hanick 1987, 1988, 1989; Michalak and Gibb 1997; Nierop and De Vos 1988; and Poon et al. 2000), despite the fact that regional production systems are at the heart of economic geography (Walker 2000). Additionally, Hoare (1993) notes that " a little-explored facet of the global economy is the way regions within nations and different parts of the international community interact through trade flows.... Given the well-established tendency for any one country to trade more with some overseas nations than with others we should expect at least as much and probably more trading-partner specialization on the part of that country's constituent localities" (Hoare 1993: 701). 4 3 For simplicity in explication, the District of Columbia is referred to as a U.S. state. 160 This poses the question of why the nation is the unit of analysis in most North American studies of international trade flows. Canada and the United States undoubtedly have great variation in the destinations of their interregional exports, but each country is most often treated as a single unit of analysis. The 15 countries of the European Union (EU-15: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, and the United Kingdom) are each used as separate units of analysis when assessing the effects of the European Union on international trade flows (see Fontagne et al. 1997, for example). <See Figure 6.1, page 333> As shown in Table 6.1 and Figure 6.1, the EU-15 has a geographic area of 3.2 million kilometers squared. Therefore, the EU-15 fits inside the United States three times and Canada three times. Granted, though most of the United States can be considered economically active, the same cannot be said for Canada. However, if we only consider the area within 500 kilometers of the Canada - United States border, a conservative estimate of the dominant area of economic activity within Canada, Canada and United States would be equivalent to four times the area of the EU-15. From this perspective based on area, Canada and the United States consist of 60 spatial units of analysis, the number of continental U.S. states (50) plus the 10 Canadian provinces. The point here is that based on the arguments of Hoare (1993) and the large geographic size of both Canada and the United States, analyzing Canada - United States interregional trade flows must be undertaken at a sub-national scale. <See Table 6.1, page 286> 161 A body of research has emerged that investigates the exports of U.S. states and/or their regional groupings—see Coughlin and Wall (2003), Erickson and Haywood (1991), Gazel and Schwen (1998), Hayward and Erickson (1995), Sawchuk and Sydor (2003), and Warf and Cox (1993) for representative studies. However, even these studies concerned with the sub-national regions of the United States tend to examine the trade flows of those regions with other countries and with the "rest of the wor ld" rather than with other sub-national regions. There are also a small number of region-specific studies: Calzonetti (1991), Hayter and Holmes (1999), McConnell and MacPherson (1991), Melvin (1988), Warf and Cox (1990), but these articles only consider the effects of the C U F T A and/or the N A F T A on regions, cities, and/or specific industries within Canada and the United States rather than their relationship with the rest of the two countries. In short, within the geographical trade literature, the sub-national region is a relatively unexplored research area, particularly with respect to Canadian regional (provincial level) trade flows. 6.2. PAST RESEARCH ON CANADA - U.S. INTERREGIONAL TRADE Not surprisingly, especially given the spatial distribution of U.S. states' international trade flows found by Erickson and Hayward (1991), Canada - United States trade is very regionally oriented. Breaking Canada - United States trade flows into only three regions (Atlantic Canada - New England, Ontario - Great Lakes, and Cascadia), Brown (1998) finds that the composition of Canada - United States trade varies geographically: Atlantic Canada is dominantly involved in natural resource-based trade flows; Ontario has a broad-based composition of trade flows dominated by manufacturing, particularly the 162 automotive industry; and Cascadia's trade flows are in between with both manufactured goods and natural resources. Norcliffe (1996), though, measuring the destinations and origins of Canadian regions' exports and imports at the national level, finds that Ontario and the Prairies import the greatest proportions of trade from the United States (72.2 and 84.5 percent of imports in 1993, respectively), whereas Ontario exports the greatest proportion of trade to the United States (89.5 percent), largely due to the automotive industry (Quebec and the Prairies are also high at respectively 78.9 and 75.3 percent). Turning to Canadian regional trade flows outside of the United States, Quebec and the Atlantic provinces have the strongest ties to Europe, and British Columbia has the strongest ties to Japan (British Columbia and the Territories have the weakest ties to the United States).44 Quite clearly, geographical proximity, colonial ties back to Europe, and the regional importance of the automotive industry play important roles in determining the spatial distribution of Canadian regional trade flows. Brown and Anderson (1999) separate Canada into five regions and the United States into nine regions, finding similar regional ties as Brown (1998) and Norcliffe (1996). Building on this analysis, Brown and Anderson (1999) calculate the regional economic dependence of the Canadian and U.S. regions on each other. Due to the United States' greater diversification in international trade flows the economic dependence of U.S. regions on Canada is significantly lower than for Canadian regions on the United States. Only 1.32 percent of the United States' economy, in 1992, was tied to Canada, 4 4 Though not explained by Norcliffe (1996), British Columbia's weak tie to the United States is from a low level of imports, relative to other provinces because British Columbia's export shares to the United States are high. 163 whereas 17.45 percent of the Canadian economy was tied to the United States in that same year. The greatest interregional ties are between the Great Lakes region in the United States (3.15 percent of GDP) and Ontario (23.91 percent of GDP). The remainder of the Canadian regions are below the Canadian average, with the prairies being the lowest—this result is curious given that the Prairies overwhelmingly import and export from the United States. Using export trade flows as a percentage of G D P to measure economic dependence, Brown and Anderson (1999) reinforce the finding that geographically close regions are the most integrated. A l l Canadian regions exported more than 10 percent of their G D P to the United States in 1992. The largest proportions of that economic dependence is tied to their respective U.S. border regions—the automotive industry in Ontario and the Great Lakes, for example. More insightful, however, is their calculation of intra-industry trade using the Grubel and Lloyd (1975) Index. Intra-industry trade is found to vary significantly across the Canadian regions. Ontario has the highest level of intra-industry trade, particularly with the Great Lakes region (automotive industry trade flows), followed by Quebec, Alberta/British Columbia, the Atlantic provinces, and the Prairies. Although Brown and Anderson (1999) mistakenly associate intra-industry trade as capturing similarity in industrial structure, their analysis is instructive in showing the geographical nature of Canada - United States in intra-industry trade. Those regions that have the greatest degrees of intra-industry trade also have the largest manufacturing bases. And because a greater degree of manufacturing leads to greater product differentiation (both horizontal and vertical) intra-industry trade is positively associated with the degree of the manufacturing base. 164 The primary limitation of the above studies is their lack of a temporal component to investigate whether or not the spatial distribution of Canada - United States trade flows have changed over time, in particular, as a result of the C U F T A or N A F T A . Polese (2000) is the first paper to fill this gap in the literature using Canada - United States interregional trade flows. Asking whether or not Quebec is special with regard to North American integration, Polese (2000) finds that Quebec is more integrated with the rest of Canada than Ontario. And Ontario is the least integrated with Canada, but the most integrated with the United States. Curiously, Ontario appears to be the most dedicated to the Canadian confederation and Quebec appears to be the least dedicated, at least at the political level, although it is Quebec that potentially stands to lose the most from separation from Canada. Furthermore, since 1993 the percentage of Canadian regional G D P exported to the United States essentially doubled by 1997, likely a result of the N A F T A . The addition of the temporal component by Polese (2000) only exemplifies the importance of the U.S. economy to Canada found by Brown and Anderson (1999). And finally, Acharya et al. (2003), using four U.S. and five Canadian regions, analyze Canada - United States interregional trade flows 1980 - 2000. Overall they find that Canadian exports to New England and the Midwest have fallen while the Northwest has remained constant; and imports for all of three regions have remained relatively constant in terms of the interregional distribution of trade flows. The most striking result is the large increases for both imports and exports with the U.S. South. This phenomenon is common across all five Canadian regions ranging from small increases (Ontario increased its share of exports to the South from 1 0 - 1 3 percent) to rather large increases (Ontario increased its share of imports from the South from 13 - 23 percent). Other 165 Canadian regions also had increases in their share of trade flows as high as 10 percent, but the sheer volume of Ontario's trade flows makes any increases in Ontario's trade flows the most significant of all the provinces. Additionally, British Columbia, the Prairies, and Atlantic provinces dominated trade in natural resources, Quebec dominated trade in labour intensive products, and Ontario is more specialized in manufactured goods (Acharya et al. 2003). This literature on Canada - United States interregional trade flows, though small, is particularly instructive. The static analyses of interregional trade flows show that there are definite spatial patterns relating to geography and history, and the panel studies show that these spatial patterns are changing over time, likely due to the establishment of free trade agreements. However, the current scope of analyses on Canada - United States interregional trade flows leaves much to be desired for an understanding of the geography of Canada - United States interregional trade. Therefore, in this and the following chapters an extensive analysis, both descriptive and inferential, is undertaken to fill this gap in the literature. 6.3. DATA AND METHODOLOGY The data used for the interregional trade analysis of Canada and the United States mainly derives from Statistics Canada. Statistics Canada (1998; 2000; 2005a) provides both interprovincial and international trade flow data at the province45 and state level, indicating both the origin and destination of trade flows. These are the same data used in 4 5 The use of the term province is used in the present analysis to represent all Canadian regions, including the Territories. Though the Territories are not provinces, this terminology is used to simplify explication. 166 Chapter 4 that presents a national analysis of Canada - United States international trade flows. But rather than being aggregated by industry to the national level, the data are aggregated to total exports and imports for Canadian provinces and U.S. states. There are no industrial classifications presented in the present analysis for all provinces and industrial sectors, but Ontario's international trade at the industrial sector level is presented for the automotive industrial sector to explicate the results found below. Data for provincial GDP is obtained through Statistics Canada (2005a), and data for U.S. state GSP (gross state product) is obtained through the website maintained by the U.S. Bureau of Economic Analysis <http://www.bea.doc.gov/>. Using these data, this chapter presents a descriptive analysis of Canada - United States interregional trade. First, the volume of provincial exports and imports are calculated for each province with each of the Canadian provinces and U.S. states. This includes the use of increase factors (the ratio of 2001 trade levels to 1989 trade levels) that measure of the degree of increased or decreased interregional trade flows to summarize the changing patterns of province-state trade flows. Second, the shares of trade for exports and imports at the province-province and province-state levels are calculated. The purpose of this analysis is to show the changing provincial focus (if any) of interregional trade flows since the establishment of the CUFTA and the NAFTA. Third, the economic dependence of each province on each U.S. state and Canadian province is calculated to investigate whether or not economic dependence has changed after the establishment of the CUFTA and the NAFTA. And fourth, an analysis of the patterns found is undertaken using Ontario's international trade flows with the U.S. states in the automotive industrial sector. Essentially, the descriptive analyses presented in this 167 chapter complements the descriptive analyses in Chapter 4 at the national level, but is not undertaken to the same degree because of the number of spatial units to analyze. Calculating these various indices at the regional level shows the geography of Canada -United States international trade flows in reference to the analyses at the national level. The regional descriptive analyses, therefore, provide a much richer and comprehensive account of the world's largest bilateral trade flow from an explicitly geographical perspective. 6.4. C A N A D A - U.S. I N T E R R E G I O N A L T R A D E FLOWS, 1989 - 2001 6.4.1. The Volumes of Interregional Trade The export and import trade volumes representing all exports and imports to Canada and the United States are shown in Tables 6.2 and 6.3. The interprovincial and international export trade flows for all provinces except the Territories increased faster than G D P from 1989 to 2001—the Territories increased their real levels of exports by 21 percent, but their economy grew by 61 percent. The interprovincial and international import trade flows, however, increased slower than G D P for all provinces except Alberta, Saskatchewan, Manitoba, and Ontario. None of Canada's provinces decreased their real levels of imports from 1989 to 2001, but the export orientation of Canadian provinces has increased since the establishment of the free trade agreements. Export trade volumes for the Canadian provinces have the general pattern of increased trade with most U.S. states. For most provinces, the vast majority of U.S. states have undergone above average increases in exports from Canadian provinces, with the largest increases typically related to geographic proximity and the economic size of 168 the export destination (the issue of trade volumes and economic size is dealt with in the following chapter). However, many of the increase factors for the United States must be interpreted with caution due to the small number effect. For example, British Columbia's greatest magnitude increase factor for exports is for South Dakota a result of an increase in export volumes from 3.96 to 33.77 million 1997 dollars, but British Columbia's export volumes to Washington increased by more than 4 billion 1997 dollars over the same time period (Washington also has a high increase factor). Therefore, export and import destinations that have relatively small 1989 export or import trade volumes must be interpreted with caution with regard to the increase factor. More interesting, however, is the changing export trade volumes to the Canadian provinces. As shown in Chapter 5, interprovincial trade has increased since the establishment of a free trade agreement in 1989, but those increases have been much lower than the increases in trade between Canadian provinces and U.S. states during the same period. The geography of the increases in interprovincial trade varies significantly from province to province and is vastly different from the experience of U.S. states. <See Tables 6.2 and 6.3, pages 287 - 298> Aside from Alberta, British Columbia's export values to the Canadian provinces are less than average, with decreases to New Brunswick and the Territories. Alberta had no interprovincial exports increasing greater than average and also decreasing its export trade volumes to the Territories. Saskatchewan is far more Canadian focused, only decreasing its export values to Quebec, but increasing export values to Alberta, Nova Scotia, Prince Edward Island, and Newfoundland. Manitoba resembles Alberta, with below average increases for all provinces except Saskatchewan and New Brunswick, that 169 both experienced decreases. Ontario decreased its export trade volumes to most other provinces, and Quebec reduced its export trade volumes to four provinces—neither Ontario nor Quebec had above average increases to any Canadian province. New Brunswick, Nova Scotia, and Prince Edward Island appear to be reorganizing their interprovincial export trade volumes, with some provinces having decreases in those volumes while others have above average increases. Newfoundland, however, exhibits significant growth in its interprovincial export trade volumes to all of central and eastern Canada—British Columbia, Alberta, Saskatchewan, and the Territories decreased their export trade volumes from Newfoundland. Incidentally, Newfoundland also has the largest number of decreases in export trade volumes to the United States. And the Territories, aside from British Columbia and Ontario, exhibit sharp decreases in its interprovincial export trade volumes. Therefore, with the exception of Manitoba and the spatial redistribution of interprovincial trade in the Maritimes and Newfoundland, interprovincial export trade volumes have suffered, particularly with respect to the economically largest provinces. Import trade volumes have seen a somewhat different pattern of change. As with export trade volumes, most Canadian provinces increased their imports from the U.S. states, with the largest increases dominantly being related to geographic proximity and economic size. However, Quebec, the Maritime provinces, Newfoundland, and the Territories all decreased their import trade volumes from many U.S. states, particularly those of Nova Scotia, Prince Edward Island, Newfoundland, and the Territories. Additionally, a large number of these decreases in import trade volumes do not appear to be motivated by geographical proximity. 170 Curiously, the opposite pattern manifests itself when considering interprovincial import trade volumes. Though there are some notable increases in interprovincial import trade volumes dominantly based on geographical proximity, the western provinces and Ontario undergo many significant decreases in interprovincial import trade volumes and relatively few above average increases. Conversely, Quebec, the Maritime provinces, Newfoundland, and the Territories have many above average increases in interprovincial import trade volumes and relatively few decreases in interprovincial trade volumes. Furthermore, the majority of those interprovincial import trade volume decreases are related to the western provinces and Ontario, the same provinces that are decreasing their import trade volumes to Quebec and the provinces to its east. Overall, aside from Newfoundland, the export trade volumes of the Canadian provinces have shifted toward the U.S. states. Most provincial destinations of export trade volumes have experienced either decreases or below average increases. Import trade volumes, however, have undergone a change that is much more geographical than export trade volumes. The western provinces and Ontario are reorientating their import trade volumes away from other provinces and toward the U.S. states. In contrast, Quebec, the Maritime provinces, Newfoundland, and the Territories, though still increasing their import trade volumes from many U.S. states, appear to be reorientating the sourcing of their imports to originate from Canadian provinces. Much of this finding is consistent with Wall (2003). Wall (2003) found that the N A F T A has had a negative impact on trade in eastern Canada, a positive impact on trade in central Canada, and an insignificant impact on trade in western Canada. The weaker spatial reorientation of eastern Canada's export trade volumes to the U.S. states and the 171 spatial reorientation of eastern Canada's import trade volumes to Canadian provinces is consistent with Wall 's (2003) findings with respect to the N A F T A . Additionally, if Ontario's spatial reorientation of import and export trade volumes dominate those of Quebec's due to their relative sizes, central Canada's effect on trade from the N A F T A found in the present analysis is also consistent with that found by Wall (2003). Western Canada, however, is more problematic. With the exceptions of Saskatchewan in terms of export trade volumes and Manitoba in terms of import trade volumes, the western provinces appear to be undergoing a spatial reorientation toward the U.S. states. That said, after Wall (2003), there may be enough variation within the western provinces to make one believe that the N A F T A has had an insignificant impact on trade—the conflicting effects of the N A F T A may be washing themselves out. 6.4.2. The Shares of Interregional Trade The analysis of trade shares provides a much clearer representation of the changing spatial distribution of provincial exports and imports. Tables 6.4 and 6.5 present the export and import trade shares representing all exports and imports to Canada and the United States—just as the case of trade volumes, these shares are not to be confused with those presented in Chapter 5 that encompass provincial exports and imports with all countries in the world. A l l of the Canadian provinces expect Newfoundland and the Territories increased their share of exports to the United States by approximately 50 percent from 1989 to 2001—provinces that already exported large shares of their trade to the U.S. states in 1989, such as British Columbia and Ontario, did not exhibit such high magnitudes of change. The Territories maintained their very low share (4 percent) of exports to the 172 United States, whereas Newfoundland's export share fell by more than 30 percent, from 70 to 48 percent. Despite the fact that most Canadian provinces send 50 - 70 percent of their exports to the United States, these numbers are still disproportionately low. Of the 61 spatial units (Canadian provinces and U.S. states) in the present analysis, over 80 percent are in the United States. Even i f one does not consider the economic size of the U.S. states Canadian provinces still export disproportionately less with the United States. The same relationship also holds for imports, aside from Ontario. The import shares of the Canadian provinces from U.S. states is lower, on average, that that of exports (Ontario is the primary exception). However, the majority of provinces increased their shares of imports from the U.S. states by approximately 50 percent, similar to that of export shares. Prince Edward Island, not Newfoundland, was the exception, in this case importing a decreasing import share from the United States. However, Prince Edward Island, Newfoundland, and the Territories obtain a very small share of their imports from the United States. <See Tables 6.4 and 6.5, pages 299 - 306> Aside from these noted changes, most of which are not radical in magnitude, the majority of changes in export and import shares appear moderate. There may have been significant changes in export and import trade volumes to the United States, but the large increases in most of the provinces' export and import shares to the United States appear to be a result of the cumulative effect of small changes favouring that country. In other words, the relative spatial distribution of provincial exports and imports is changing in favour of the U.S. states, but that change is gradual. 173 Given that the United States has increased both its export and import shares to and from Canadian provinces, interprovincial export and import shares must have fallen—the shares presented in this chapter reflect the shares of exports and imports only within Canada and the United States. However, given the variation in the export and import trade volumes, it was expected that there be a similar variation in the export and import trade shares. If one ignores Newfoundland and the Territories, the overwhelming change in interprovincial export and import shares is negative. There are a few moderate increases, dominantly related to geographic proximity, and a number of export and import trade shares that have essentially remained the same, but the dominant direction of change is negative. Even Ontario, maintaining the largest export and import trade shares for almost all provinces, exhibits large magnitude changes in those shares in all provinces except Newfoundland and the Territories. The direction of change is clear. Canadian provinces are decreasing their exports and imports with other Canadian provinces and increasing their imports and exports with U.S. states. In the case of changes for the U.S. states, exports and imports are generally changing gradually. But, much of the change in exports and imports involving other Canadian provinces is abrupt. Therefore, Canadian provinces are substituting their interprovincial trade for interregional trade with the U.S. states. This substitution involves sharp decreases for a small number of Canadian provinces coupled with moderate increases to many U.S. states. In other words, the sharp decreases in interprovincial trade did not correspond with sharp increases in international trade. Early in the study period, the Canada - United States border posed a barrier to trade flows except for those province-state combinations that were already significantly integrated. 174 A consequence of this barrier is much lower export and import shares than would be expected given the geographic and economic size of the United States. However, as the tariff barriers decreased—all tariffs were "officially" zero by 1998—Canadian provinces began to export relatively more to the United States and relatively less to other Canadian provinces. At this point, it is likely that interprovincial barriers began to supersede national barriers fostering international trade at the expense of interprovincial trade. So by 2001, there had been dramatic increases in not only the export and import trade volumes between provinces and states but also in export and import trade shares. The trade shares of Canadian provinces with U.S. states are still not near what would be expected, particularly for imports, the eastern provinces, and the Territories. However, because of the decreases in trade barriers, those trade shares are now much closer to any a priori expectations given market size. As a result of increased volumes and increased shares of trade with U.S. states the economic dependence of Canadian provinces on U.S. states has increased. 6.4.3. Economic Dependence and Interregional Trade Following Brown and Anderson (1999) and Polese (2000), the economic dependence of a Canadian province either on another Canadian province or a U.S. state is measured using the export value to GDP ratio (see Table 6.6). Though not a perfect measure of economic dependence, the share of a province's GDP that is exported to another province or U.S. states does provide a good indicator of the importance of that export destination to the local economy. From 1989 to 2001 all Canadian provinces except the Territories increased their economic dependence on the United States. This result is expected given that exports are 175 growing faster than G D P over this study period. In fact, increasing economic dependence is a well-established fact for most countries of the world (Dicken 2003; International Monetary Fund 2005a, 2005b). And similar to export and import trade shares, most Canadian provinces have increased their economic dependence on the United States through trade by approximately 50 percent. It is because of the greater market access resulting from the establishment of free trade agreements, Canadian provinces have increased their economic dependence with the United States. Furthermore, the geography of this economic dependence is quite apparent. Most provinces, in 1989, have an economic dependence with the United States ranging from 1 0 - 1 5 percent—the Territories' economic dependence is notably low, and has markedly decreased with its closest U.S. neighbour, Alaska. Ontario is the exception in this year with over 20 percent of its G D P exported to the United States. Given the long established relationship between Ontario and Michigan of free trade in automotive products, this is expected. By 2001, most Canadian provinces doubled their economic dependence through trade with the United States. Some provinces more than doubled their economic dependence, with Prince Edward Island tripling its economic dependence. Consequently, most provinces now have more than 20 percent of their economy tied to the United States, with some close to 40 percent. Somewhat unexpectedly, Ontario no longer has the highest degree of economic dependence. Though Ontario's economic dependence is still high, 35.94 percent, Alberta tops the list at 38.43 percent, with New Brunswick following closely behind Ontario at 35.24 percent. <See Table 6.6, pages 307 - 310> 176 Once again, the results for interprovincial trade are significantly different. In 1989, most provinces had a greater degree of economic dependence through trade with other Canadian provinces than with the U.S. states—British Columbia, Ontario, and Newfoundland were the exceptions. This pattern reversed by 2001 for all provinces except Saskatchewan, Prince Edward Island, Newfoundland, and the Territories (the case of the former three provinces, economic dependence on Canada and the United States through trade is approximately equal). In the western (except Saskatchewan) and Maritime provinces, national economic dependence has either increased or decreased marginally, exhibiting very little change over this 12 year study period. Ontario, Quebec, and the Territories, however, exhibit significant drops in their economic dependence on Canada through trade. Only Saskatchewan and Newfoundland have experienced notable increases in their economic dependence—Saskatchewan's economic dependence increased by 39 percent and Newfoundland's economic dependence more than tripled. The overall pattern of change with regard to economic dependence through trade is that western and central Canada increased its economic dependence with the U.S. states at the expense of Canadian provinces. The Maritime provinces and Newfoundland increased their economic dependence with the U.S. states but also increased their economic dependence with geographically close Canadian provinces. And the Territories are decreasing their economic dependence with all regions in Canada and the United States except with British Columbia and Ontario. The analysis on the economic dependence is the most geographically related with the changes in economic dependence following much more of a geographical proximity relationship than the analysis of trade volumes in particular, but also trade shares. And similar to that of trade shares, the 177 increases in provincial economic dependence on U.S. states, aside from the noted moderate increases, dominantly come from the cumulative effect of small changes across the entire study region. Therefore, economic dependence is a dimension that changes slowly as economies adjust to their new economic geographical environment with free trade agreements. 6.4.4. Proximity and Economic Size In the previous sub-sections, I have alluded to proximity and economic size as the potential drivers of change in the provincial interregional trade patterns. With regard to proximity, I have stated that geographically close trading partners (particularly in the United States) have experienced greater increases in their trade flows to and from Canadian provinces than geographically distant trading partners, and that the large economic size of the United States has distributed the increases in interregional trade flows such that Canadian provinces have had sharp decreases in their trade with each other, but the U.S. states have generally had much more modest change over the study period because there are more U.S. states to absorb the corresponding decreases in interprovincial trade. However, at this point of the analysis, these claims have not been substantiated. Table 6.7 shows the trading partners with each of the Canadian provinces that have undergone substantial increases in economic dependence. Economic dependence is used for this analysis because its geographical pattern is the most pronounced. Substantial increases are considered those increases that are more than two standard deviations greater than the average increase in economic dependence, and bolded states and provinces are those trading partners that are geographically close. Two conclusions 178 can be drawn from the information in this table: first, geographically close trading partners are highly represented in this set of substantial increases in economic dependence; and second, there are some trading partners that are not geographically close that have also undergone substantial increases in economic dependence. Therefore, proximity alone, though clearly an important factor in the process of change taking place within Canada and the United States, cannot be considered the driving force of change now that barriers to trade between Canada and the United States have been officially eliminated. <See Table 6.7, page 311> When considering the shares of interregional trade, access to the larger United States' market is important for understanding the degree of change in interregional trade: sharp decreases for interprovincial trade and moderate increases in U.S. interregional trade. However, the individual market sizes of the U.S. states also appear to be important in identifying those trading partners that experienced significant increases in economic dependence. The prime example shown in Table 6.7 is Texas. Though Texas is not geographically close to any Canadian province, Texas has had significant increases in economic dependence for all of the western and central provinces, as well as New Brunswick. A similar result is found with California. In order to assess this possible relationship to explain the pattern of spatial change in Canada - United States interregional trade, a regression analysis is undertaken. Using the economic dependence of a Canadian province on each trading partner as the dependent variable, the natural logarithm of the geographic distance between trading partners and the natural logarithm of the economic size (gross domestic product) of 179 trading partners are estimated to test the hypothesis of proximity and economic size in determining the pattern of change in economic dependence. If this hypothesis is correct, the estimated coefficient for geographic distance will be negative and the estimated coefficient for economic size will be positive. However, the Adjusted-R2 values are of particular interest, not the signs of the estimated coefficients. The magnitudes of the Adjusted-R2 values will show how much of the variation in economic dependence is explained by these two variables.46 The results are presented in Table 6.8. <See Table 6.8, page 312> Almost all of the estimated coefficients correspond to their expectations, with all statistically significant coefficients corresponding to their expectations. Curiously, only the statistical models for British Columbia and Alberta exhibit the expected signs on the estimated parameters and have reasonably high Adjusted-R2 values. Though the results for economic size do fare better than those for geographic distance, there is still much to be explained in understanding the changing spatial pattern of economic dependence. Even in the cases of British Columbia and Alberta, the Adjus