Trade creation and trade diversion in the Canada – United States Free Trade Agreement Kimberly A. Clausing Department of Economics, Reed College

Abstract. In this paper the changes in trade patterns introduced by the Canada-United States Free Trade Agreement are examined. Variation in the extent of tariff liberalization under the agreement is used to identify the impact of tariff liberalization on the growth of trade both with member countries and non-member countries. Data at the commodity level are used, and the results indicate that the Canada-United States Free Trade Agreement had substantial trade creation effects, with little evidence of trade diversion. JEL Classification: F13, F14 Création de commerce et diversion de commerce dans l’Accord de libre-échange CanadaU.S. Ce mémoire examine les changements dans les patterns de commerce international engendrés par l’Accord de libre-échange entre la Canada et les Etats-Unis. La variation dans l’intensité de libéralisation tarifaire selon les secteurs dans l’Accord est utilisée pour identifier l’impact de la libéralisation tarifaire sur la croissance du commerce à la fois entre les pays membres et avec les pays non-membres. A l’aide de données par produits, on montre que l’Accord a eu des effets substantiels de création de commerce mais qu’il n’y a pas lieu de croire qu’il y a eu beaucoup de diversion de commerce.

1. Introduction Preferential trading arrangements are reshaping the world trading system. An incomplete inventory of recent initiatives includes proposals to extend NAFTA to Chile, to create a Free Trade Area of the Americas, to continue the expansion of the European Union to other European countries, and to establish free trade among the APEC nations. These initiatives, as well as many others throughout the world, have This work has been supported in part by a National Science Foundation Graduate Fellowship. I am grateful to the referees of this paper for constructive comments that greatly improved the paper. In addition, I would like to thank James Hines, John Helliwell, Richard Caves, Donald Davis, Daniel Trefler, Janet Ceglowski, and participants at seminars at Harvard University, the University of Toronto, and the Federal Reserve Board of Governors for helpful comments on an earlier version. Email: [email protected] Canadian Journal of Economics 0 Revue canadienne d’Economique, Vol. 34, No. 3 August 0 août 2001. Printed in Canada 0 Imprimé au Canada

0008-4085 0 01 0 677–696 0 r Canadian Economics Association

678

K.A. Clausing

led to a fierce debate about the merits of regional trading agreements. While some have heralded such agreements as stepping stones towards worldwide free trade, others fear that these initiatives will be stumbling blocks, acting primarily to divert trade from other countries to those countries receiving preferential treatment. Although these issues are essential for the future of the world’s trading relationships, a number of obstacles have prevented economists from reaching any consensus on the effects of preferential trading agreements. The second-best nature of tariff liberalizations under preferential trading arrangements makes it very difficult to assess a priori whether the welfare effects from a preferential trading arrangement will be positive, even for the members of the arrangement. In addition, the empirical work has failed to reach firm conclusions on even the most basic issue regarding preferential trading agreements: whether trade creation outweighs trade diversion. In this paper I attempt to better isolate the effects of a preferential trading agreement, by analysing the Canada-United States Free Trade Agreement ~CUSFTA!. The CUSFTA provides a valuable natural policy experiment, since policy makers committed themselves to eliminate tariffs on all goods, creating variation in the degree of tariff liberalization because of the substantial variation in the initial, predetermined, tariff rates. This variation is used to identify the impact of tariff liberalization on the growth of trade with both member countries and non-member countries. Unlike the approaches of many previous studies of preferential trading agreements that have relied on aggregate data, disaggregate data ~at the commodity level! are used to analyze how actual tariff changes affect trade flows. Without utilizing the variation in the extent of liberalization across goods, it would be far more difficult to distinguish the effects of an agreement from other influences affecting trade flows. In this paper I find that the CUSFTA did have very noticeable effects on trade flows between Canada and the United States. There is substantial evidence of increased trade due to the agreement, particularly for those goods undergoing the largest tariff liberalizations. Further, the data do not reveal the evidence one would expect were trade diversion widespread. In the next section I review the previous work in this area. In section 3 I examine the theoretical framework for the hypotheses tested in this paper. The data are presented in section 4, and issues of specification are discussed. In section 5 I discuss the main results, which are qualified in section 6. Section 7 concludes.

2. The effects of preferential trading agreements 2.1. Previous theoretical and empirical work Before Viner, analysts often assumed that a customs union would be welfare improving, since some tariffs would fall and tariffs are, in general, welfare reducing. In 1950 Jacob Viner showed that a customs union will not necessarily improve welfare, since the tariff reductions occur in a world of second best. Whether or not the

Trade creation and trade diversion

679

increase in trade caused by the formation of a customs union would be welfare improving depends on the source of the increased trade. Trade creation occurs when the lowering of tariffs allows partner country imports to replace high-cost domestic production; this improves welfare. Trade diversion, on the other hand, occurs when the removal of tariffs causes trade to be diverted from a third country to the partner country despite the fact that, were the countries treated equally, the third country would be the low cost source of imports. In the Vinerian framework, welfare therefore depends on the extent of trade creation relative to trade diversion. Viner’s analysis illustrates the main result regarding the welfare effects of customs unions and free trade areas: no general statement can be made about the welfare effects of preferential trading agreements. Whether or not the partner countries, much less the world, will gain depends on the assumptions laid out in the model and the circumstances of the particular agreement. Further, the authors responsible for this literature disagree on what is likely to happen in practice. Some authors, such as Krugman ~1991, 1994!, believe that preferential arrangements between natural trading partners are likely to be positive developments. Others, such as Bhagwati ~1994!, fear that preferential trading arrangements may lead to trade diversion as the international mechanisms to discipline trade diversion, such as article XXIV of the GATT, are too weak. Empirical researchers have also had difficulty reaching firm conclusions regarding the effects of preferential trading agreements. In particular, past studies have been unable to convincingly disentangle the effects of preferential trading agreements from other effects occurring simultaneously. These studies generally fall into one of three categories. 2.1.1. Ex post studies examining the share of intra-agreement trade Many authors ~including Krueger 1999; Brada 1994; Cline 1978; de la Torre and Kelly 1992; Drysdale and Garnaut 1993; Robertson 1970; Saxonhouse 1994b! have examined trade shares before and after an agreement in order to assess what effect the agreement may have had on trade patterns. It is often implicitly assumed that the share of trade occurring with partner countries would not have changed in the absence of the agreement. A more sophisticated counterfactual is necessary, however, in order to assess the effect of an agreement and the extent of trade creation relative to trade diversion. 2.1.2. Ex post studies with more elaborate counterfactuals Developing an accurate counterfactual of how much trade would have increased in the absence of a given free trade agreement or customs union has proved difficult.1 1 For instance, Balassa ~1967, 1975! constructed a counterfactual of how trade would have changed in the absence of European integration by calculating pre-integration income elasticities that then were assumed to continue post-integration. It was later demonstrated, however, that income elasticities varied substantially pre- and post-integration, making these results sensitive to the sample period.

680

K.A. Clausing

Some ~including Frankel and Wei 1995; Frankel and Kahler 1993; Frankel 1997; Krueger 1999; Aitken 1973; Aitken and Obutelewicz 1976; George et al. 1977; Willmore 1976! have used gravity equations to assess the impact of preferential arrangements on trade flows. These equations have the advantage of including several variables that are affecting trade flows, such as income changes and exchange rate variables. A dummy variable is typically used to assess the impact of various preferential trading agreements on trade flows. Still, there are problems with the gravity equation approach. First, the dummy variables used to distinguish the effects of preferential trading agreements are blunt instruments that have difficulty capturing only the effects of preferential trade liberalization.2 Second, this type of analysis does not indicate the extent of trade creation relative to trade diversion.3 Third, these studies examine trade flows at a very aggregate level. Since these studies are therefore unable to exploit variations in the extent of trade liberalization across goods or industries, it is more difficult to distinguish the effects of liberalization from other influences that are acting on trade flows.4 2.1.3. Ex ante studies of trade agreements In addition to the ex post studies above, there have been a large number of ex ante CGE studies that examine what effects can be expected from preferential trading arrangements ~e.g., Brown, Deardorff, and Stern 1992; Brown and Stern 1989a,b; Haaland and Norman 1992!. One flaw of CGE studies is that their results are very sensitive to the assumptions, parameters, and data used in the model, and have to be interpreted accordingly. While CGE models are useful for speculating what the effects of a particular agreement might be, they do not allow an investigation of the questions we are concerned with here. 2.2. Work on the Canada – United States Free Trade Agreement Since CUSFTA went into effect in 1989, authors of several studies have considered the effects of the agreement, including Statistics Canada ~1993!, Schwanen ~1997!, 2 One difficulty is that such dummy variables may also capture other effects that are difficult to distinguish from preferential trade liberalization. For instance, Frankel and Wei ~1995! use a gravity equation to estimate the effects of European integration. They find that the only years that the EC dummy is statistically significant and positive are 1985 and ~perhaps! 1990. Since the trade liberalization that occurred within the EC does not particularly coincide with the years in which this dummy variable is larger, the dummy variable may be indicating that trade within the EC is becoming more ‘regionalized.’ This need not imply, however, that trade creation has occurred because of preferential liberalization. 3 If, for example, the coefficient on a dummy variable for partner trade is statistically significant and positive, this does not tell us whether the excess trade among member countries was a result of trade creation or trade diversion. 4 As Frankel ~1997, 93! notes, ‘some effects of regional trading agreements are lost in tests like ours on highly aggregated data. The problem is particularly severe in a context, such as the Canada-US FTA, where the small number of observations in the case of aggregated data is a constraint. But Clausing’s finding that the effects differ widely @depending on the extent of liberalization# . . . also illustrates the deeper perils of aggregation.’

Trade creation and trade diversion

681

Thompson ~1994!, Helliwell, Lee, and Messinger ~1998!, Gaston and Trefler ~1997!, and Head and Ries ~1999!. The question of this paper is not systematically addressed in any of these papers, but they do provide insights into the effects of CUSFTA. Statistics Canada ~1993! provides an overview of the changes in trade patterns during the early years of the agreement, between 1989 and 1991. Schwanen ~1997! undertakes a comprehensive study of changes in Canadian trade patterns, considering the effects of both CUSFTA and NAFTA between 1989 and 1995. Schwanen compares trade in sectors that have been liberalized by these agreements to trade in other sectors, finding that trade growth with the United States was much faster in liberalized sectors.5 Helliwell, Lee, and Messinger ~1998! consider the effect of CUSFTA on trade between Canadian provinces, using both aggregate data and data at the industry level ~67 industries included!. They find that CUSFTA increased international ~province-state! trade, while trade between the Canadian provinces decreased. Head and Ries ~1999! analyse a panel of 230 Canadian manufacturing industries, examining the impact of trade liberalization on efficiency and scale. Head and Ries find that during the years following CUSFTA the Canadian manufacturing sector experienced substantial rationalization. Rationalization is defined as a decline in the number of plants accompanied by an increase in output per plant. Gaston and Trefler ~1997! address the labour market effects of CUSFTA, trying to account for the employment contraction in Canada during 1989–93. Although Gaston and Trefler find that the net employment effects of the CUSFTA are negative, they estimate that tariff cuts under CUSFTA account for no more than 15 per cent of the Canadian job losses.

3. The theory of trade creation and trade diversion The unilateral removal of a tariff generally increases imports of the good in question, increasing domestic consumption and reducing domestic production. The gains to consumers outweigh the loss of tariff revenue and producer surplus, leading to overall welfare gains. As Viner pointed out, however, the analysis is more complex if the tariff is only reduced on partner imports. An examination of figure 1 makes this ambiguity clearer. Figure 1 shows an analysis of a good in Canada that is initially protected by a tariff. Imports are equal to the quantity AB, the difference between domestic demand and domestic supply at the tariff-inclusive price. Consider, first, the case where S U.S. is the U.S. supply curve. U.S. suppliers are not competitive prior to liberaliza5 Sectors are classified either as liberalized or as not liberalized and then are compared. This does not allow a consideration of the extent of liberalization or of disaggregate data, such as those employed in this study. Trade diversion is not explicitly considered, although data on trade with the rest of the world are also provided. Nonetheless, Schwanen ~1997! addresses many important effects of CUSFTA and NAFTA, including not only trade effects but also investment and employment effects.

682

K.A. Clausing

FIGURE 1

Trade creation and trade diversion

tion. Once the tariff is eliminated on U.S. goods, imports from the United States replace those from the rest of the world. Since the U.S. duty-free price is lower than the tariff-inclusive world price, demand increases and Canadian domestic production is reduced. New imports are higher, at quantity CD. Domestic consumers gain areas FGHI, domestic producers lose area F, tariff revenue falls by HL, and the overall welfare effects are ambiguous. Trade creation leads to a gain of GI, but trade diversion leads to a loss of L, as the U.S. imports replace lower cost ROW imports. In practice, there are several cases when the outcome would be less ambiguous. For example, if the United States were already the low cost producer before the FTA, trade creation would result in welfare gains equal to areas GIKM, without any trade diversion losses. If U.S. supply were instead uncompetitive before the tariff reduction, however, and just epsilon less than the ROW tariff inclusive supply after the FTA, only trade diversion would take place, with a loss in tariff revenue of HL but no noticeable gains.6,7 The CUSFTA lowered tariffs on thousands of goods, so the aggregate effect will depend on the extent of trade creation and diversion in many different markets. Further, there may be a host of other more indirect effects as changes in one market 6 With upward-sloping supply curves for the partner country or the world, the same ambiguities hold. One can construct examples with only trade creation, only trade diversion, or intermediate cases. 7 Some authors, such as Kowalczyk ~2000!, Harrison, Rutherford, and Wooton ~1993!, and Ethier and Horn ~1984!, have criticized the literature’s excessive reliance on the notions of trade creation and trade diversion and have suggested alternatives. Kowalczyk, for example, develops a general equilibrium model that instead considers volume of trade and terms of trade effects. This model, while presenting the economic effects differently, raises many of the same questions. In general, while the alternative frameworks have merits, none has gained widespread acceptance.

Trade creation and trade diversion

683

affect others. Nonetheless, this simple framework prompts two essential areas of enquiry to consider as we examine data on how trade patterns changed in the years following CUSFTA. First, to what extent did trade between Canada and the United States increase as a result of tariff liberalization within the agreement? What is the magnitude of trade response to tariff reductions? The graphical analysis above predicts that larger tariff reductions should lead to greater increases in trade with the partner country. Is that pattern observed in the data? Second, to the extent that trade did increase, how much of this increase was a result of trade diversion? In particular, do imports from the rest of the world fall more ~or increase less! for trade that was more highly liberalized between partners? Are the same goods that are experiencing disproportionate gains in trade with the partner country experiencing disproportionately slower growth in trade with the rest of the world? 4. Data and methodology 4.1. The data If one considers the most commonly used methods for studying the effects of trade agreements on trade, it is difficult to parse out any effects of the Canada-United States Free Trade Agreement. For example, many ex post analyses of trade agreements have focused on how the share of trade between members has changed in the wake of the agreement. Here, though, trade shares are not particularly instructive.8 In addition, a gravity equation analysis of U.S. and Canadian trade flows fails to discern an effect of the agreement.9 8 Figures showing trade shares are available upon request. The share of exports and imports occurring with the partner country ~the United States or Canada! does not increase in most cases, and when it does ~the share of Canadian exports destined for the United States!, it is highly correlated with the simultaneous depreciation of the Canadian dollar vis-à-vis the U.S. dollar. Still, there are several difficulties with using simple trade shares, as noted in the text above. Further, even more sophisticated measures of trade bias ~such as those discussed in Drysdale 1988, 86–8! lead to ambiguous conclusions regarding the effects of CUSFTA on trade flows. Tables of these calculations are available upon request. 9 An appendix of tables containing the gravity equation analysis is available upon request. The typical gravity equation specification is estimated with dummy variables included to capture the effects of CUSFTA. The specification is: ln ~Tradeij ! 5 a 1 b1 ln~GDPi * GDPj ! 1 b2 ln ~GDPper capitai * GDPper capitaj ! 1 b3 ln~distance ij ! 1 b4 ln~Ex.Rate ij ! 1 b5 Z. All tested approaches fail to find effects of the Canada – United States Free Trade Agreement on trade flows. In the first approach, a dummy variable is included for the partner country ~Canada or the United States! in all years, as well as a separate dummy variable for the years following the agreement. The latter variable is never statistically significant. The second approach includes dummy variables for the partner country for each year following the agreement. These yearly dummy variables are also not statistically significant. A third approach estimated regressions on subsets of data before and after the agreement. A dummy for partner trade was included in each in an attempt to discern differences before and after the agreement. In all cases, the 95 per cent confidence interval for the value of the partner dummy coefficient in these regressions substantially overlapped, implying that one can not distinguish a difference in excess trade between the two countries before and after the agreement.

684

K.A. Clausing

TABLE 1 Summary statistics, disaggregate United States import data

Imports from Canada ~thousands of U.S.$! ROW imports ~thousands of U.S.$! Candian tariff rate Rest of world tariff rate Change in Canadian tariff rate

Mean

SDev

Number of Observations

12,100

179,000

50,240

50,400

523,000

50,240

3.5 5.8 0.5

5.1 16.2 2.6

40,429 49,528 30,660

U.S. data is for the years 1989–94. Rest of world ~ROW! imports are imports from all countries except Canada. The tariff rate is the duties paid relative to the total value of imports for a given commodity in a given year. The change in the tariff rate is the percentage point decrease in a given commodity’s tariff in a given year. Data are from the U.S. Census, reported in Feenstra ~1996!.

However, aggregate data could mask changes that may be occurring at a disaggregate level. A large percentage of trade between Canada and the United States was free before the agreement, and it could be the case that the most liberalized trade flows were too small to be picked up by the aggregate analysis. Also, the use of more disaggregate data allows one to exploit the variation in the extent of tariff liberalization under the agreement. Without utilizing such variation, it is difficult to identify the effects of tariff liberalization. Data exist from the U.S. Census ~reported in Feenstra ~1996!! that describe imports and exports broken down by commodity and country, according to the Harmonized Classification system. This system was adopted by Canada in 1988 and the United States in 1989, and the tariff liberalizations taking place within the CUSFTA were also scheduled according to the Harmonized system. This analysis uses the harmonized data, covering the period 1989 to 1994 for the United States. Import values are provided in the data, and tariff rates are calculated as the duties collected relative to the total value of trade in a given category. Tariff liberalization under the agreement began in 1989. While some tariffs were eliminated immediately, tariffs were typically brought down in equal increments over five or ten years. The Harmonized Classification System, in its fullest detail, separates trade into over 8,000 ten-digit categories. Table 1 shows summary statistics for the data set. There is a substantial variation in the level of initial protection as well as the degree of tariff liberalization. For U.S. imports from Canada, table 2 shows the percentage of commodities in different categories of tariff protection, both by the number of observations and by the amount of trade. Even at the onset of the agreement in 1989, a large percentage of imports were freely traded. Still, there are many observations where tariffs are high. By 1994 Canadian imports are subject to much lower tariffs than they were previously.

Trade creation and trade diversion

685

TABLE 2 The pattern of protection in 1989 and 1994 ~for U.S. imports from Canada! 1989

Free trade Tariffs under 5 per cent Tariffs between 5 and 10 per cent Tariffs between 10 & 25 per cent Tariffs over 25 per cent

1994

Per cent of obs.

Per cent of imports

Per cent of obs.

Per cent of imports

22.3 45.0 20.8 10.2 1.6

62.6 33.8 3.1 0.5 0.0

26.7 60.0 9.3 3.7 0.2

62.7 36.6 0.6 0.1 0.0

Table 3 shows the same information for U.S. imports from countries other than Canada. As one can see from this table, in both 1989 and 1994 there is a substantial amount of trade in the United States that is covered by significant tariff barriers. 4.2. Specification In order to assess the effects of the tariff liberalization measures within CUSFTA, this analysis focuses on the two main questions raised at the end of section 3: 1. To what extent did trade between Canada and the United States increase as a result of tariff liberalization within the agreement? 2. To the extent that trade did increase, how much of this increase was a result of trade diversion? By examining how the growth of trade between the two countries depends on the extent of tariff liberalization, one can assess how much trade between the two countries increased as a result of the agreement. To assess the extent of trade diversion, it is necessary to assume that tariff rates are distributed across goods in a manner that is independent of the natural rate of import growth. In section 6, the implications of removing this assumption will be discussed in greater detail. The specification of the regression equations emerges from a simple supply and demand framework. Take the case of linear-in-logs import demand and export supply curves. In particular, import demand schedules for various goods ~indexed by i ! take the form: ln Di 5 g d 1 dpd ln Pi 1 dzd Z.

~1!

dpd is negative, so that a higher price implies a lower import demand; Z are various terms that vary over time, such as real exchange rates and income. Export supply schedules are ln Si 5 g s 1 dps ln Pi 1 dzs Z.

~2!

686

K.A. Clausing

TABLE 3 The pattern of protection in 1989 and 1994 ~for U.S. imports from countries other than Canada! 1989

Free trade Tariffs under 5 per cent Tariffs between 5 and 10 per cent Tariffs between 10 & 25 per cent Tariffs over 25 per cent

1994

Per cent of obs.

Per cent of imports

Per cent of obs.

Per cent of imports

15.8 42.4 25.1 14.5 2.1

13.7 64.1 13.9 6.3 1.9

16.4 44.6 23.5 13.9 1.5

22.2 57.8 12.9 6.0 1.1

Here dps is positive, so that export supply increases with higher prices. The marketdetermined price generates an equilibrium level of trade where import demand and export supply are equal. Changes in the equilibrium level of trade will result from factors that change prices ~P ! or factors that vary over time ~Z!, such as national income. Tariffs drive a wedge between the import demand curve and the export supply curve, acting to increase the prices paid by consumers of imports while reducing the prices received by producers of exports. With an ad valorem tariff, the price becomes Pi ~1 1 Ti !, and the import demand function is ln Di 5 g d 1 dpd ln~Pi ! 1 dpd ln~1 1 Ti ! 1 dzd Z.

~3!

Setting the supply and the tariff-inclusive demand functions equal, one can find an equation determining the equilibrium price, which in turn can be employed to find the quantity traded in the presence of tariffs. The quantity transacted depends on both the tariff level, T, and time-specific factors, Z. ln Di 5 Constant Term 1 B ln~1 1 Ti ! 1 CZ.

~4!

Here, B is a term that depends on the price elasticities of import demand and export supply, dpd dps 0~dps 2 dpd !. This term is negative as long as dpd is negative and dps is positive; C also depends on these elasticities, as well as dzd and dzs. As tariffs are eliminated, holding other factors constant, trade increases. The change in trade over a given time period is given by ln Di, t 2 ln Di, t21 5 B~ln~1 1 Ti, t ! 2 ln~1 1 Ti, t21 !! 1 C~Z i, t 2 Z i, t21 !.

~5!

When tariffs are zero, the free trade level of trade occurs; as tariffs are higher, trade decreases, eventually falling to zero at the prohibitive tariff level. As tariffs are reduced between Canada and the United States, prices change so that the new values of import demand and export supply are equal. The resulting change in trade can be estimated by the following simple equation:

Trade creation and trade diversion %DImportsit 5 a 1 b1 DTariffit 1 bT Year Effects.

687 ~6!

This is the first estimating equation.10 Tariff changes are observed. Year effects ~dummy variables for all but one year! are included to control for circumstances that vary over time, such as exchange rates, cyclical factors, and so on. One option would be to include each of these variables directly. Owing to the limited time dimension of the panel, however, I have employed time-specific dummy variables instead. An additional variable that may affect the responsiveness of imports to tariff changes is also considered: the share of imports that originated in Canada prior to the liberalization. Recall from figure 1 that the degree of trade creation due to tariff liberalization will depend on the prior competitiveness of the partner country. For example, if Canada is already the most competitive source of imports, tariff reductions on Canadian goods will likely lead to trade creation. On the other hand, if Canada is not competitive prior to liberalization, it is possible it may not be competitive after liberalization, and imports may not increase as a result. For instance, free trade on U.S. coffee or mango imports coming from Canada is unlikely to lead to increases in U.S. imports of these goods from Canada. Thus, in some specifications below, I also include a variable indicating the original market share of Canada in the total U.S. imports of commodity i.11 %DImportsit 5 a 1 b1 DTariffit 1 b2 Canada Share i, t21 1 bT Year.

~7!

5. Results Before turning to the regressions, consider one simple graphical illustration of the disaggregate data. I divide the sample into several groups based on the extent of liberalization during the time period 1989–94. Using these categories, one can show how trade for each category changed in the wake of the agreement. Figure 2 shows the growth of U.S. imports from the partner country, Canada, as well as from the rest of the world.12 Consider, first, the lighter bars, which show U.S. imports from Canada. The trade that was already freely flowing from Canada to the United States increased approximately 40 per cent over the five years following the agreement. However, the more extensive the liberalization, the faster trade grew. For example, 10 The %DImports is approximately equal to the term ln Di, t 2 ln Di, t21 . The DTariff variable is approximately equal to the term ~ln~1 1 Ti, t ! 2 ln~1 1 Ti, t21 !! when the tariff rate is small, which is generally the case in this data set. 11 This variable may also be a useful test of the theory ~discussed in Frankel 1997, chap. 8! that natural trading blocs ~i.e., preferential trading arrangements between countries that already trade a great deal with each other! should be less likely to lead to trade diversion and hence more likely to increase welfare. 12 Figure 2 shows the growth in trade for the average commodity in each category. If one instead examined the growth in trade for the total value of trade in each category, a similar pattern would be seen.

FIGURE 2

Growth in U.S. imports from Canada and the rest of the world, 1989–94

Trade creation and trade diversion

689

the imports that saw tariff reductions in excess of 5 per cent saw trade increase dramatically, to approximately twice its 1989 level by 1994. This figure suggests increases in trade resulting from tariff liberalization. But was the increase in trade due to trade creation or trade diversion? If trade diversion were present, then one would expect, all other things equal, trade with non-partner countries in goods highly liberalized with respect to Canada to increase more slowly than trade in goods less liberalized, or not liberalized at all. Figure 2 also shows import growth from the rest of the world. Import growth from the rest of the world does not appear to be systematically related to the extent of Canadian liberalization. How should one interpret this finding? One possibility is that trade diversion did not occur. Still, it could also be the case that trade diversion did exist, if the goods where imports were naturally growing rapidly were also those previously protected. However, as discussed further in section 6 below, there is little reason to suspect that is the case. 5.1. Regression analysis The first regression equations attempt to relate the variation among goods in the percentage change in imports from year to year to the extent of tariff liberalization and year dummy variables. The year dummy variables are intended to capture the effects of income changes and variations in the exchange rate.13 The data cover five years of changes; initial results for equations explaining the percentage change in U.S. imports from Canada are presented in table 4. The elimination of tariffs did have a statistically significant, positive, and large effect on imports. Equation ~1! indicates that each one percentage point reduction in tariffs is associated with a 9.6 per cent increase in imports from Canada.14 Equation ~2! adds the share of imports from Canada in the previous year as an independent variable. This variable is negative and statistically significant, indicating that a 1 per cent increase in the original share of imports from Canada is associated with a 0.5 per cent reduction in the increase in U.S. imports from Canada, when changes in other variables are controlled for. This is contrary to the above, where it was hypothesized that there would be greater trade creation gains where Canadian export firms were initially more competitive.15 13 I used year dummy variables, rather than including these variables directly, since the time dimension of the panel data is very short. 14 In many of these specifications, the adjusted R-2 are not very satisfying. The low explanatory power results largely from the very disaggregate data. A host of unmeasurable influences, such as varying supply conditions or changes in tastes, likely accounts for the bulk of the variation apparent in the data. If one aggregates the data to progressively fewer numbers of commodities, one finds that the explanatory power of the regressions increases correspondingly. However, this greater explanatory power comes at a cost: the year dummies explain more and more of the variation, while we steadily lose variation in tariff liberalization and hence are less able to discern the impact of tariff liberalization on trade flows. 15 A possible explanation for this finding may be that the tariff liberalization gave Canadian producers preferential treatment relative to those in other countries, and thus Canadian firms were able to increase their exports most where their initial share of the market was smallest.

690

K.A. Clausing

TABLE 4 The change in U.S. imports from Canada, 1989–94 ~Dependent variable: per cent change in imports from Canada, year over year! Equation ~1! Tariff change

0.0964 ~0.0065! 20.2441 ~0.0238! 20.2165 ~0.0237! 20.1089 ~0.0237! 20.1078 ~0.0237! 0.4519 ~0.020!

0.0886 ~0.0065! 20.4835 ~0.0254! 20.2558 ~0.0237! 20.2246 ~0.0236! 20.1121 ~0.0236! 20.1077 ~0.0236! 0.5652 ~0.0178!

28,551 71.6 0.00 0.012

28,551 120.7 0.00 0.025

Share from Canada Year dummy 1990 Year dummy 1991 Year dummy 1992 Year dummy 1993 Constant Number of obs. F Prob ~ F! Adjusted R-2

Equation ~2!

NOTES The dependent variable is year-to-year per cent changes in imports between 1989 and 1994 for the United States. The tariff change is the percentage point decrease in a given commodity’s tariff since the previous year ~typically a positive number!. The share from Canada variable is the share of total imports of that product that originated in Canada in the previous year. Standard errors are in parentheses.

Table 4 uses a panel of year-to-year changes in imports to study the effects of tariff liberalization on imports. This has the advantage of allowing an investigation of several trade changes as tariffs are slowly lowered. I also examined regressions explaining the change in imports over the entire time period ~1989–94! as a function of tariff liberalization. The results are given in table 5 and are consistent with the previous results. The tariff change variable is of a very similar magnitude, indicating that a 1 percentage point reduction in tariffs is associated with a 10 or 11 per cent increase in imports from Canada. The share variable indicates that a 1 per cent increase in the original share of imports from Canada is associated with a 0.69 per cent reduction in the increase in U.S. imports from Canada, when the other variables are controlled for. 5.2. Estimates of increased trade due to CUSFTA One can use the regressions in table 5 to estimate an elasticity that measures the effect of changes in tariffs over this time period on import growth. In equation ~2!,

Trade creation and trade diversion

691

TABLE 5 The growth in U.S. imports from Canada, 1989–94 ~Dependent variable: per cent change in imports from Canada, from 1989 to 1994!

Tariff change

Equation ~1!

Equation ~2!

0.1107 ~0.0119!

Constant

0.4325 ~0.0361!

0.1041 ~0.0119! 20.6858 ~0.0960! 0.5976 ~0.0426!

Number of obs. F Prob ~ F! Adjusted R-2

3,692 85.7 0.00 0.022

3,692 68.9 0.00 0.036

Share from Canada

NOTES The data cover the total percentage change in U.S. imports from Canada between 1989 and 1994. The tariff change is the percentage point decrease in a given commodity’s tariff brought about by the agreement between 1989 and 1994. The share from Canada variable is the share of total imports of that product that originated in Canada in 1989. Standard errors are in parentheses.

each 1 percentage point cumulative reduction in tariffs is associated with imports 10 per cent higher in 1994. Using this elasticity and taking into account the average change in tariffs, one can estimate the effects of CUSFTA on trade. For the United States, this estimate suggests that imports from Canada were 26 per cent higher, owing to CUSFTA. Comparing this figure to the actual growth in U.S. imports from Canada by 1994, I estimate that over half ~54 per cent! of the $42 billion increase in U.S. imports from Canada was due to CUSFTA.16 5.3. Trade diversion? In order to examine the possibility of trade diversion, a series of regressions was examined in which the dependent variable was the percentage change in imports of a particular commodity from the rest of the world. If trade diversion were present, one would expect the percentage change in imports from the rest of the world to be negatively related to the extent of tariff liberalization with Canada. Table 6 shows regressions for U.S. data. The specifications shown are parallel to those of table 4, with the exception of an additional independent variable indicating the degree of liberalization of imports from Canada. Equation ~2! also includes the change in imports from Canada as an explanatory variable. 16 One can make similar calculations, of course, using coefficients from other specifications. These estimates are quite similar to those reported here.

692

K.A. Clausing

TABLE 6 Regressions examining the change in imports from the ROW ~Dependent variable: per cent change in imports from the rest of the world, year over year! Equation ~1! ROW tariff change Canadian tariff change

0.0215 ~0.0034! 0.0016 ~0.0022!

Per cent change Can Im

Equation ~2! 0.0214 ~0.0034! 0.0016 ~0.0022! 20.0000 ~0.0000!

Share from Canada Year dummy 1990 Year dummy 1991 Year dummy 1992 Year dummy 1993 Constant Number of obs. F Prob ~ F! Adjusted R-2

Equation ~3! 0.0252 ~0.0040! 0.0030 ~0.0022!

20.1357 ~0.0100! 20.1622 ~0.0100! 20.0710 ~0.0100! 20.0380 ~0.0100! 0.1951 ~0.0071!

20.1357 ~0.0100! 20.1622 ~0.0100! 20.0710 ~0.0100! 20.0380 ~0.0100! 0.1951 ~0.0071!

0.1604 ~0.0121! 20.1339 ~0.0100! 20.1604 ~0.0100! 20.0703 ~0.0100! 20.0378 ~0.0100! 0.1644 ~0.0074!

28,881 67.1 0.00 0.014

28,881 57.5 0.00 0.014

28,881 83.1 0.00 0.020

NOTES The data cover year to year percentage changes in U.S. imports from countries other than Canada between 1989 and 1994 for the United States. The ROW tariff change is the percentage point decrease in a given commodity’s tariff in a given year, for countries other than Canada. The Canadian tariff change is the percentage point decrease in a given commodity’s tariff in a given year, for Canada. The % Change Can Im variable is the percent change in imports from Canada, year over year. The share from Canada variable is the share of total imports of that product that originated in Canada in the previous year. Standard errors are in parentheses.

Turning to the results, coefficients on the ‘ROW tariff change’ variable indicate that to the extent that tariffs on goods from non-Canadian countries decreased, imports from these countries increased. A 1 percentage point reduction in tariffs was associated with a 2 per cent increase in imports. In all cases, the coefficients on the variables indicating tariff liberalization on Canadian goods are statistically indistinguishable from zero. Thus, there is no discernible relationship between the extent of Canadian tariff liberalization and import growth from countries in the rest of the world. One might expect the percentage change of imports from the rest of the world to be lower as tariff reductions on Canadian goods are higher and increases in Canadian imports are larger. For example, if the United States lowers tariffs on mango

Trade creation and trade diversion

693

imports from Canada, this should not result in any trade diversion, since Canada will not export mangos to the United States under most imaginable circumstances. On the other hand, if a reduction in the tariff on wooden furniture from Canada results in an increase in imports of such furniture, this may lead to fewer imports of furniture from other countries, thus causing trade diversion. One way to investigate this sort of relationship is to include variables that indicated the percentage change in imports with the partner country; this specification is reported in equation ~2!. The estimated coefficient on this variable is 0. I also tested regressions included interaction terms between the change in Canadian imports and tariff liberalization ~not reported here!. In all cases, the interaction terms were statistically indistinguishable from zero. One can also test specifications similar to those in table 5 for imports from the rest of the world. Though not reported here, the results were very similar to those of table 6. The one noticeable change was that changes in imports from Canada were sometimes positively related to changes in imports from the rest of the world.17 6. The variation in initial tariff rates One useful aspect of studying the CUSFTA is that policy makers committed themselves to lowering tariffs on all goods. The initial tariff rates and schedules were predetermined, thus making the variation in tariff rates useful for identifying the impact of tariff changes on trade flows. If tariffs are distributed among the goods in a way that is correlated with import increases, however, this makes it more difficult to discern whether increases in trade between the two countries are due to trade creation or trade diversion. For example, if tariffs were for some reason higher for goods that are likely to have larger increases in imports, this could bias upward the size of the import responses measured above and bias downward the estimates of trade diversion due to the agreement. However, while this is a noteworthy consideration, there are several reasons to suspect that this concern should not invalidate the results presented here. First, work by Gaston and Trefler ~1997! uses industry-level data to address the exogeneity of tariff cuts under the CUSFTA for Canada. After constructing tariffreduction instruments, they conclude that they are unable to reject the hypothesis that tariff cuts under the CUSFTA are exogenous. Second, in my own analysis using the data set of this study, all of the variables that I am able to examine show no clear relationship with initial tariff levels. For instance, the level of imports in the first year of the data is not positively correlated with tariff levels, as one might suspect if the above concerns are important.18

17 This result is not too surprising, since imports of a particular good may rise simultaneously from several countries, owing to an increased demand for the product, bad supply conditions at home, or other factors. 18 Results are available from the author upon request.

694

K.A. Clausing

Finally, there is no clear theoretical rationale common in the literature that would lead one to suspect that goods with large tariffs would also be those goods where import growth was naturally higher. There is a large literature in which the political economy of trade policy has been considered, more specifically, the factors that influence tariffs.19 The authors attempt to reconcile the economic arguments in favour of free trade with the reality that trade typically is not free. If policy makers did maximize an economist’s perception of social welfare, it is unlikely that they would use tariffs at all.20 In practice, though, policy makers may be inclined to allow tariffs on commodities that are deemed special because of high profits, high wages, suspected spillovers, or simply particularly aggressive lobbying. 7. Conclusions The CUSFTA had substantial effects on trade between Canada and the United States. The tariff liberalization brought about by CUSFTA was responsible for over half of the $42 billion increase in U.S. imports from Canada between 1989 and 1994. The commodities that experienced the largest reductions in tariffs had the largest increases in trade. In addition, there was little evidence of trade diversion from non-member countries. The use of disaggregate data makes it possible to distinguish effects that are more difficult to discern using aggregate data. Since initial tariff rates varied significantly, it is possible to compare the growth of imports that were extensively liberalized, less extensively liberalized, or not liberalized at all. This variation in tariff liberalization allows an identification of the impact of the agreement on trade flows with both member and non-member countries. These results suggest an encouraging assessment of the CUSFTA. Since the gains due to the agreement were not at the expense of other countries, it is less likely that the CUSFTA will discourage future efforts towards free trade worldwide. More likely, the CUSFTA increased the constituencies with an interest in free trade. However, this analysis certainly does not resolve all questions concerning the desirability of preferential trading agreements. Individual circumstances associated with such arrangements vary, and each arrangement must be judged on its own merits. Nonetheless, the importance of using disaggregate data to assess the effects of a given trade agreement is indicated by this paper. Ex post analyses that rely on aggregate data may be less informative, since it is difficult to separate the effects of liberalization from other influences affecting the data. References Aitken, N.D. ~1973! ‘The effect of the EEC and EFTA on European trade: a temporal cross-section analysis,’ American Economic Review 63, 881–92 19 There is an excellent and comprehensive review of this work in Rodrik ~1995!. 20 One exception might be an optimal tariff, to exploit market power. In practice, however, there is no evidence that policy makers implement optimal tariffs.

Trade creation and trade diversion

695

Aitken, N.D., and R.S. Obutelewicz ~1976! ‘A cross sectional study of EEC trade with the Association of African Countries,’ Review of Economics and Statistics 58, 425–33 Anderson, J.E. ~1979! ‘A theoretical foundation for the gravity equation,’ American Economic Review 69, 106–16 Balassa, B. ~1967! ‘Trade creation and trade diversion in the European Common Market,’ in Comparative Advantage, Trade Policy, and Economic Development, ed. B. Belassa ~New York: New York University Press! ––. ~1975! European Economic Integration ~Amsterdam: North-Holland! Bergstrand, J.H. ~1985! ‘The gravity equation in international trade: some microeconomic foundations and empirical evidence,’ Review of Economics and Statistics 67, 474–81 ––. ~1989! ‘The generalized gravity equation, monopolistic competition, and the factor proportions theory in international trade,’ Review of Economics and Statistics 71~1!, 143–53 ––. ~1990! ‘The Heckscher-Ohlin-Samuelson model, the Linder hypothesis, and the determinants of bilateral intra-industry trade,’ Economic Journal 100, 1216–29 Bhagwati, J. ~1994! ‘Regionalism and multilateralism: an overview,’ in New Dimensions in Regional Integration, ed. J. De Melo and A. Panagariya ~Cambridge: Cambridge University Press! Brada, J.C. ~1994! ‘Regional integration in Eastern Europe: prospects for integration within the region and with the European Community,’ in New Dimensions in Regional Integration, ed. J. De Melo and A. Panagariya ~Cambridge: Cambridge University Press! Brown, D.K., and R.M. Stern ~1989a! ‘U.S.-Canada bilateral tariff elimination: the role of product differentiation and market structure,’ in Trade Policies for International Competitiveness, ed. R. Feenstra ~Chicago: University of Chicago Press! –– ~1989b! ‘Computable general equilibrium estimates of the gains from U.S.-Canadian trade liberalization,’ in Economic Aspects of Regional Trading Agreements, ed. D. Greenaway et al. ~Brighton: Harvester Wheatsheaf! Brown, D.K., A.V. Deardorff, and R.M. Stern ~1992! ‘A North American Free Trade Agreement: analytical issues and a computational assessment,’ World Economy 15, 11–29 Cline, W.R. ~1978! ‘Benefits and costs of economic integration in Central America,’ in Economic Integration in Central America: A Study, ed. W.R. Cline and C. Delagado ~Washington, DC: Brookings Institution! Deardorff, A.V. ~1998! ‘Determinants of bilateral trade: does gravity work in a neoclassical world?’ in The Regionalization of the World Economy, ed. J.A. Frankel ~Chicago: University of Chicago Press! de la Torre, A., and M.R. Kelly ~1992! ‘Regional trading arrangements,’ IMF Occasional Paper No. 93. Drysdale, P. ~1988! International Economic Pluralism ~New York: Columbia University Press! Drysdale, P., and R. Garnaut ~1993! ‘The Pacific: an application of a general theory of economic integration,’ in Pacific Dynamism and the International Economic System, ed. C.F. Bergsten and M. Noland ~Washington, DC: Institute for International Economics! European Free Trade Association ~1969! The Effects of EFTA on the Economies of Member States ~Geneva: EFTA Secretariat! Ethier, W., and H. Horn ~1984! ‘A new look at economic integration,’ in Monopolistic Competition and International Trade, ed. H. Kierzkowski ~Oxford: Clarendon Press! Feenstra, R.C. ~1996! ‘U.S. imports, 1972–1994: data and concordences,’ NBER Working Paper No. 5515 and accompanying CD

696

K.A. Clausing

Frankel, J.A. ~1997! Regional Trading Blocs in the World Economic System ~Washington, DC: Institute for International Economics! Frankel, J.A., and M. Kahler ~1993! Regionalism and Rivalry: Japan and the U.S. in Pacific Asia ~Chicago: University of Chicago Press! Frankel, J.A., and S. Wei ~1995! ‘European integration and the regionalization of world trade and currencies: economics and politics,’ in Monetary and Fiscal Policy in an Integrated Europe, ed. B. Eichengreen et al. ~New York: Springer! Gaston, N., and D. Trefler ~1997! ‘The labour market consequences of the Canada-United States Free Trade Agreement,’ Canadian Journal of Economics 30, 18–41 George, R., et al. ~1977! ‘Short-run trade effects of the LAFTA,’ Kyklos 30, 618–36 Haaland, J., and V. Norman ~1992! ‘Global production effects of European integration,’ in Trade Flows and Trade Policy after ‘1992,’ ed. L.A. Winters ~Cambridge: Cambridge University Press! Harrison, G.W., T.F. Rutherford, and I. Wooton ~1993! ‘An alternative welfare decomposition for customs unions,’ Canadian Journal of Economics 26, 961–68 Head, K., and J. Reis ~1999! ‘Rationalization effects of tariff reductions,’ Journal of International Economics 47, 295–320 Helliwell, J.F., F.C. Lee, and H. Messinger ~1998! ‘Effects of the Canada – U.S. FTA on interprovincial trade,’ photocopy, University of British Columbia Kowalczyk, C. ~2000! ‘Welfare and integration,’ International Economic Review 41, 483–94 Kowalczyk, C., and R. Wonnacott ~1992! ‘Hubs and spokes, and free trade in the Americas,’ NBER Working Paper No. 4198 Krueger, A. ~1999! ‘Trade creation and trade diversion under NAFTA,’ NBER Working Paper No. 7429 Krugman, P. ~1991! ‘The move toward free trade zones,’ in Policy Implications of Trade and Currency Zones: A Symposium Sponsored by the Federal Reserve Bank of Kansas City ~Kansas City: Federal Reserve Bank of Kansas City! –– ~1994! ‘Regionalism vs. multilateralism: analytical notes,’ in New Dimensions in Regional Integration, ed. J. De Melo and A. Panagariya ~Cambridge: Cambridge University Press! Markusen, J.R. ~1986! ‘Explaining the volume of trade: an eclectic approach,’ American Economic Review 76, 1002–11 Robertson, D. ~1970! ‘Effects of EFTA on member countries,’ in Europe’s Free Trade Area Experiment, ed. H. Corbet and D. Robertson ~Oxford: Pergamon Press! Rodrik, D. ~1995! ‘Political economy of trade policy,’ in Handbook of International Economics, Vol. III, ed. G. Grossman and K. Rogoff ~New York: Elsevier Science! Saxonhouse, G.R. ~1994a! ‘Comment on “Multilateral trade negotiations and preferential trading arrangements,”’ in Analytical and Negotiating Issues in the Global Trading System, ed. A.V. Deardorff and R.M. Stern ~Ann Arbor: University of Michigan Press! –– ~1994b! ‘Trading blocs and East Asia,’ in New Dimensions in Regional Integration, ed. J. De Melo and A. Panagariya ~Cambridge: Cambridge University Press! Schwanen, D. ~1997! ‘Trading up: the impact of increased continental integration on trade, investment, and jobs in Canada,’ C.D. Howe Institute Commentary Statistics Canada ~1993! Trade Patterns: Canada – U.S., 1981–1991 ~Otttawa: Statistics Canada! Thompson, A. ~1994! ‘Trade liberalization, comparative advantage, and scale economies: stock market evidence from Canada,’ Journal of International Economics 37, 1–27 Viner, J. ~1950! The Customs Union Issue ~New York: Carnegie Endowment for International Peace! Willmore, L. ~1976! ‘Trade creation, trade diversion, and the effective protection in the Central American Common Market,’ Journal of Development Studies 12, 396–414