The geographical effects of the NAFTA on Canadian provinces

Transcription

1 Ann Reg Sci (2009) 43: DOI /s ORIGINAL PAPER The geographical effects of the NAFTA on Canadian provinces Martin A. Andresen Received: 20 May 2007 / Accepted: 2 January 2008 / Published online: 19 January 2008 Springer-Verlag 2008 Abstract This paper investigates the effects of the North American Free Trade Agreement (NAFTA) on the Canadian provinces. A large body of research has emerged testing the effects of the Canada US Free Trade Agreement and the NAFTA, but the majority of that research has analyzed the effect of free trade at the national scale despite the fact that different provinces have different industrial compositions and levels of integration with the US. It is found that there is a geographical component to the effect of the NAFTA, and this geographical component varies from province to province. JEL Classification F130 F140 F150 1 Introduction Since 1989, the year the Canada US Free Trade Agreement (CUFTA) entered into force, Canada s exports of merchandise to the US have almost tripled and imports of merchandise from the US have almost doubled in real terms. Canada s share of trade with the US has risen from 69 to 73%; however, this increase is all due to exports (74 85%) given that Canada s import share from the US has fallen (65 60%). At the provincial level, all provinces except Newfoundland have more than doubled their international trade with the US, (Statistics Canada 2004). Curiously, despite these significant increases in international trade between Canada and the US I am grateful to Trevor J. Barnes for comments on an earlier draft of this paper. M. A. Andresen (B) School of Criminology, Simon Fraser University, 8888 University Drive, Burnaby, BC V5A 1S6, Canada andresen@sfu.ca

2 252 M. A. Andresen over the past 15 years, the American economist Paul Krugman has been cited as saying that the impact of the North American Free Trade Agreement (NAFTA) on Canada is zero (Contenta 1996). The CUFTA outlined a systematic tariff reduction schedule between Canada and the US that was completed by 1998 as well as instituting deeper levels of integration regarding institutional compatibility, the harmonization of rules for competition, and investment. In other words, the CUFTA is much more than a free trade agreement. There is little doubt that the CUFTA has significantly changed the Canada US trading relationship. As shown by Coulombe (2004) and Courchene (2003), Canada s interprovincial trade was greater than Canada s international trade in the early 1980s, but that trend had reversed by the early 1990s. But does the addition of Mexico have any implications for Canada because Canada trades so little with Mexico? The short answer is yes. Canada may trade relatively little with Mexico, but US Mexico trade has risen significantly in recent years. With large volumes of trade flowing between the US and Mexico, any alterations to the trading relationship between the US and Mexico could alter Canada s trading relationship with the US. However, the effects of the NAFTA on the Canadian economy have been elusive to identify. The purpose of this paper is to identify the effect of the NAFTA on the Canadian economy. This is done both at the national and provincial levels. 1 Now that 14 years have passed since the establishment of the NAFTA a number of studies have emerged testing its impact on its different countries. This literature, along with some of its methodological issues, is discussed in the following section. Section 3 outlines the data and methodology employed. The results are discussed in Sect. 4. Section 5 concludes that the impact of the NAFTA has been positive for Canada as a whole, but its effects vary geographically. 2 Past reseach of the effects of the NAFTA Gould (1998) is the first study to explicitly test the independent effect of the NAFTA. Using trade both originating and destined for the NAFTA countries and a 90% confidence interval, Gould (1998) finds statistical support for the NAFTA generating increases in US Mexico international trade, particularly imports from Mexico. However, he finds that the impact of the NAFTA has been positive though statistically insignificant for Canada with both the US and Mexico. Krueger (1999) comes to a similar conclusion for Canada using a gravity model of international trade. Though finding that the NAFTA has decreased NAFTA country imports from non-members, the NAFTA has an insignificant positive effect on the NAFTA countries. This is likely due to aggregation bias. The entire effect of the NAFTA is insignificant for Krueger (1999) because Canada US international trade dominates international trade within North America as a whole. Therefore, 1 Another literature that has investigated provincial-level trade is the work of McCallum (1995), Helliwell (1996), and others that investigate the border effect. This literature, however, does not investigate the impact of the NAFTA, per se.

3 The geographical effects of the NAFTA on Canadian provinces 253 within Krueger s analysis, an insignificant effect of the NAFTA on Canada US likely dominates Krueger s (1999) total intra-nafta effect. The last national analysis of the effect of the NAFTA is the most comprehensive. Romalis (2005) uses commodity level data, a sophisticated theoretical model, and corresponding estimation procedure. He finds that the NAFTA has had a substantial impact on international trade, but only a modest impact on welfare in the NAFTA countries. This establishment of the positive impact on international trade is critical because as stated by Gould (1998), the NAFTA can only impact welfare if it alters international trade. As such, Romalis (2005) is the first national level analysis of the effect of the NAFTA to find an overall positive effect. Coughlin and Wall (2003) is the first paper to study the geography of the effects of the NAFTA, albeit on US states. 2 Employing a rather parsimonious gravity model of international trade, Coughlin and Wall (2003) find that the overall effect of the NAFTA on US exports to Canada is positive and statistically significant, a 15% increase in international trade. Geographically, 36 US states had a greater than 10% increase in international trade with Canada, while eleven US states exhibited little change ( 10 to 10%), and four US states had a decrease of more than 10%. When the individual US states are aggregated into the nine Bureau of Economic Analysis regions, all US regions indicate positive change for international trade with Canada. 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) with the nine Bureau of Economic Analysis US 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 US are up 29 and 14%, respectively. Eastern Canada decreased exports and imports to and from the US resulting from the NAFTA, down 9 and 13%, respectively. Central Canada exhibits large increases in both exports to and imports from the US, up 43 and 18%, respectively. Western Canada demonstrates a 0.9 percent increase in exports to the US and a 0.5% decrease in imports from the US leading Wall (2003) to claim that the overall impact of the NAFTA on western Canada is insignificant. The primary limitations of past research investigating the effects of the NAFTA on Canada relate to geographical and statistical specification. The geographical specification is with regard to the scale of analysis. The vast majority of research on the effects of the NAFTA is at the national scale of analysis. Such analyses ignore the fact that regions within a country will trade differently from each other just as two geographically close (contiguous) countries will trade differently (Hoare 1993). Within countries as geographically large as Canada and the US the effects of free trade agreements must be analyzed at the sub-national level or risk shrouding the effects of free trade agreements through aggregation bias and the modifiable areal unit problem. 2 There are some studies not discussed here that do investigate Canada s regional trade such as Baldwin and Brown (2004) andbrown and Anderson (1999), but these studies do not assess the impact of the NAFTA: the latter uses 1 year of data (1992) and the former is concerned with employment and industry specialization. As such, these other studies, though concerned with the regional dimension of Canada s trade, do not assess NAFTA s impact. Also not discussed here because of its scale of analysis is some interesting work on high-technology firms in Toronto, ON that finds firm level responses to free trade depend on pre-free trade export intensity and firm size Britton (2002a,b).

4 254 M. A. Andresen The statistical specification issue is one of misspecification. As indicated above, parsimonious statistical models are used to assess the effects of the NAFTA on Canadian international trade. Most of these models include measures of GDP and NAFTA variables, but the inclusion of other important variables such as time trends, exchange rates, and other structural variables are not always present. This potentially leads to statistical bias and incorrect inference. Additionally, not all of the statistical models use panel data estimation procedures when their data are spatial cross-sections measured repeatedly over time. The lack of the implementation of these estimation procedures also necessarily imposes statistical bias potentially leading to incorrect inference. These issues are addressed in the present analysis. 3 Data and methodology 3.1 Statistical specification issues The effects of the NAFTA on Canada as a whole and its provinces are estimated using both international and interprovincial trade data only considering other Canadian provinces and US states as export destinations and import origins. This circumscription of trading partners is chosen because the vast majority of trade flowing to and from Canadian provinces is interprovincial and to the US. The addition of Mexico to the CUFTA could change US trading patterns with Canada. This impact on Canada is a secondary effect, whereas measuring any impact on Canada from changes in Canada Mexico trade is a primary effect. Though the primary effect may be of interest, any significant economic (or welfare) impacts of the NAFTA will occur through the secondary effect. This is simply because of the volume of trade (particularly Canadian exports) between Canada and the US. Even if the NAFTA were to increase Canada Mexico international trade by 100%, the economic impact on Canada would be small because Canada Mexico trade only accounts for 2% of Canada s total international trade in % of Canada s exports. However, even a 5 or 10% increase in Canada US international trade would have a significant economic impact on Canada s economy. Consequently, the focus of the current analysis is restricted to the trading relationship that may have a significant impact on Canada s economic welfare. 3 The gravity equation of international trade, used often to asses the NAFTA, has been found to empirically capture both inter-industry trade and intra-industry trade models. This equation can be theoretically derived for both models (see Anderson 1979; Anderson and van Wincoop 2003; Bergstrand 1985, 1989; Deardorff 1998; Helpman and Krugman 1985), having both theoretical foundations and empirical support. Therefore, the determinants of both inter- and intra-industry trade must be included in any proper specification. This is not the case in the studies on the effects of the NAFTA. 3 There is also the complication that Mexico s trade data are only available at the national level and the current study has a regional focus.

5 The geographical effects of the NAFTA on Canadian provinces 255 The variables that need to be included in a proper gravity equation model specification are the economic sizes (GDP) of the trading economies and the distances between them (Tinbergen 1962; Pöyhönen 1963; Linneman 1966). The level of economic development (GDP per capita) measures tendency toward intra-industry trade (Frankel et al. 1995). Trade barriers (both tariff and non-tariff barriers) need to be included because of their role in stimulating trade (Feenstra 2002). Relative factor endowments (capital labor and land labor ratios) capture the degree of difference between trading economies indicating inter-industry trade (Hummels and Levinsohn 1995). Exchange rate and linear trend variables must be included to account for changes in the relative cost of Canadian goods and the general trend of increased trade, respectively. Lastly, variables are needed to capture the presence of free trade agreements. The final issue with regard to statistical specification is the statistical estimation procedure. Cheng and Wall (2005) and Wall (2000) both show that gravity models of international trade that not only measure trade across space but across time are improperly specified when estimated using OLS. The result of this misspecification is biased statistical estimates and, therefore, potentially improper inference. A fixed effect panel estimation procedure provides each bilateral trading relationship with its own dummy variable that controls for this misspecification. This form of fixed effect also controls for historical, social, political, and cultural factors, as well as distance and contiguity measures (Cheng and Wall 2005). The variables discussed above are estimated in the following panel gravity equation specification spanning : ln x ijt = α+α ij +β 1 ln y Exporter t +β 2 ln y Importer t +β 3 ln yc Exporter t +β 4 ln yc Importer t +β 5 kl Exporter t + β 6 kl Importer t + β 7 tl Exporter t + β 8 tl Importer t +β 9 ln er t + β 10 tariff t + β 11 t + β 12 NAFT A + ε, (1) where x ijt is the bilateral trade between trading economies i and j at time t,α is the common intercept, α ij represents the trading-partner-specific fixed effects, y Exporter t and y Importer t are GDPs at time t, yc Exporter t and yc Importer t are GDP per capita at time t, kl Exporter t and kl Importer t are the capital labor ratios at time t, tl Exporter t and tl Importer t are the land-labor ratios at time t, er t is the exchange rate between Canada and the US at time t with interprovincial trade given an exchange rate value of one, tariff t is the average Canada US tariff rate at time t, t is the linear trend, NAFTA is a dummy variable that is one after 1993 and zero otherwise, and ε ijt is the remaining iid error term. 4 Because the CUFTA is in effect for the entire time span of these data, , the dummy variable for the NAFTA captures the effect of the NAFTA independent of the CUFTA. The NAFTA variable takes on a number of forms in the analysis below depending on the regression context. Equation 1 is the general form of all the models estimated below: four regression models are undertaken in this analysis, two using Canada US international trade 4 All reported p values reflect White standard errors.

6 256 M. A. Andresen (n = 6,500) and two using both interprovincial and international trade (n = 7,085). In the former analysis, the NAFTA s effects are investigated only for changes in the Canadian trading patterns with the US, similar to most previous analyses. In the latter analysis the NAFTA s effects are investigated both on changes in the Canadian patterns with the US and with other Canadian provinces. Past research has not made such an inquiry. Given the use of dummy variables in a semilogarithmic statistical equation, the appropriate interpretation of that variable is critical. As shown by Halvorsen and Palmquist (1980), the correct interpretation of a dummy variable in a semilogarithmic specification is that exp(β) 1 represents the percentage change in the dependent variable from the dummy variable changing its value from zero to one see Kennedy (1981) for a small sample equivalent. All Canadian provinces (no territories due to a lack of explanatory variables) and all US continental states are included in the data set. All zero values of bilateral trade are replaced with one dollar to facilitate the natural logarithm. 5 This allows for all possible trading relationships to be modelled without any serious concerns for bias (Greene 2000). All GDP and GDP per capita variables are measured in natural logarithms to ease interpretations: the coefficients for these variables are then elasticities. All estimation is performed using NLogit 3.0. One final methodological consideration is the issue of spatial dependence. However, there are two complications that prevent the appropriate testing of spatial dependence in this context. First, the data used in this analysis are spatial interaction data. As a result, each province has 60 trading partners for each year. The nature of shapefiles (when used with ESRI products or GeoDa) does not allow each spatial unit to have multiple values for the same variable. This situation arises because the 60 values for each year are not intrinsically a value associated with a single spatial unit (province), but a value that represents a relationship between two spatial units. Secondly, the data employed here are a panel. Therefore, each trading relationship has 13 observations, one for each year each province has 780 observations for the dependent variables in the analysis. 6 The end result of these two complications is that the current data set cannot be incorporated into a GIS format in order to create spatial weights matrices and properly test for spatial dependence. This is problematic not only for the current analysis but also for other studies that employ either spatial interaction and/or panel data, because in the presence of spatial dependence the variances for parameter estimates are biased downwards, potentially indicating statistical significance when results are not statistically significant. Future research clearly needs to develop a test for spatial dependence in the presence of either spatial interaction and/or panel data. However, previous (geographical) research on the effects of the NAFTA has not been able to test for spatial dependence so the results presented below can be interpreted in comparison to that previous research. 5 The number of zero values is small, only 17. This number equates to 0.26% of the observations and is mostly for pre-nafta observations. 6 There is some double-counting here when considering to total size of the data set used in the analysis: Ontario Quebec interprovincial trade is only counted once in the data, for example.

7 The geographical effects of the NAFTA on Canadian provinces Data The data on Canada US international trade are obtained from Statistics Canada (2004). These data measure trade at the 8- and 10-digit Harmonized Tariff Schedule (HTS) level of aggregation for exports and imports, respectively, having Canadian provinces and US states as the smallest geographic units. All HTS classifications are aggregated to these geographical units based on their origin and destination. Interprovincial trade data have three sources: Statistics Canada (1998), Statistics Canada (2000), and Statistics Canada (2005a) (CANSIM) for the most recent interprovincial trade data. Provincial GDP, GDP per capita, and labor forces are also obtained through CAN- SIM. Provincial capital stocks are obtained through the Centre for the Study of Living Standards (2005), available from the CSLS online. Arable land is obtained from Statistics Canada (1986, 1992, 1999, 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. US state GDP and GDP per capita are obtained from the Bureau of Economic Analysis and US state labor forces are obtained from the Bureau of Labor Statistics, all available online. Arable land estimates for the US states are obtained from the United States Department of Agriculture (2000). As with the Canadian data, arable land for the US is only available at 5-year intervals, with the intervening years obtained through linear interpolation. US state capital stock estimates were not found to be available, nor were US state total factor productivity estimates to use growth accounting techniques to calculate state-level capital stocks. 7 In order to calculate the capital stocks for each US states, the annual US capital stock is distributed based on each state s share of US 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 = Therefore, the estimates for US state capital stock cause little concern for bias in any estimation. Lastly, Canadian tariff rates are obtained through External Affairs Canada (1687) and the US tariff rates are obtained through the The Center for International Data at the University of California, Davis. All tariff rates are based at the commodity level with the national Canadian and US tariff rates being the unweighted averages of the commodity level tariff rates. 4 Results 4.1 Canada US international trade and the NAFTA The statistical results from the four regression models are presented in Tables 1, 2, 3, 4. Table 1 shows the results from the Canada US international trade data only measuring 7 Economists at the United States Federal Reserve Board were contacted regarding the availability of US state capital stocks. They were not aware of any available data, free or proprietary, but did state that proprietary data are likely to be available. However, no such data were found.

8 258 M. A. Andresen Table 1 Canada United States trade Adjusted R 2 = F = , p value <0.01 n = 6500 All p values are based on White standard errors Coefficient Percent p value impact GDP Exporter GDP Importer GDP per capita Exporter 1.63 <0.01 GDP per capita Importer 2.74 <0.01 Capital labour ratio Exporter Capital labour ratio Importer Land labour ratio Exporter Land labour ratio Importer Exchange rate 1.31 <0.01 Average tariff rate 0.09 <0.01 Linear trend NAFTA Table 2 Canada United States trade, provincial NAFTA effects Adjusted R 2 = F = , P value <0.01 n = 6500 All p values are based on White standard errors Coefficient Percent p value impact GDP Exporter GDP Importer GDP per capita Exporter GDP per capita Importer 2.76 <0.01 Capital labour ratio Exporter Capital labour ratio Importer Land labour ratio Exporter Land labour ratio Importer Exchange rate 1.40 <0.01 Average tariff Rate Linear trend NAFTA British Columbia US NAFTA Alberta US NAFTA Saskatchewan US NAFTA Manitoba US NAFTA Ontario US NAFTA Quebec US NAFTA New Brunswick US NAFTA Nova Scotia US NAFTA Prince Edward Island US <0.01 NAFTA Newfoundland US the effect of the NAFTA at the national level for a more direct comparison to past research. The estimated coefficients for GDP, GDP per capita, and the relative factor endowment variables are generally consistent with past research with the exception

9 The geographical effects of the NAFTA on Canadian provinces 259 Table 3 Canada United States and interprovincial trade Adjusted R 2 = F = , p value <0.01 n = 7085 All p values are based on White standard errors Coefficient Percent p value impact GDP Exporter GDP Importer GDP per capita Exporter 1.70 <0.01 GDP per capita Importer 2.29 <0.01 Capital labour ratio Exporter Capital labour ratio Importer Land labour ratio Exporter Land labour ratio Importer Exchange rate 1.30 <0.01 Average tariff rate 0.07 <0.01 Linear trend NAFTA UnitedStates NAFTA Canada of the GDP variables being statistically insignificant. The GDP per capita variables, however, are statistically significant and large in magnitude. This latter result coupled with the presence of the relative factor endowment variables may be the cause for the insignificant GDP results these estimated coefficients, however, are not the primary concern of this analysis. 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 has its expected negative and statistically significant coefficient. The national effect of the NAFTA 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 NAFTA is that the current result is based on province state trade rather than Canada US aggregate national trade. Nevertheless, the impact of the NAFTA on Canada US international trade has been a 10.47% increase over and above any increases from the CUFTA. This result is more modest than that found by Wall (2003), but his analysis separates exports and imports for Canadian and US regions rather than dealing with total trade 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 NAFTA on Canada s trade to the US, Table 2 exhibits significantly different results from Wall (2003). Except for British Columbia, Alberta, and New Brunswick, all of the provincial NAFTA effects are positive, contrary to that found by Wall (2003). Even the remaining Maritime provinces and Newfoundland, all exhibiting negative effects from the NAFTA in Wall (2003) analysis, have positive coefficients with Prince Edward Island s coefficient being statistically significant and relatively high in magnitude, an increase of more than 68%. Though Prince Edward Island s international trade with the US are a small share of Canada s total trade with the US, this result exemplifies the impact on results from changing the spatial scale of analysis. It is possible that this result for

10 260 M. A. Andresen Table 4 Canada United States and interprovincial trade, provincial NAFTA effects Adjusted R 2 = F = , p value <0.01 n = 7085 All p values are based on White standard errors Coefficient Percent p value impact GDP Exporter GDP Importer GDP per capita Exporter GDP per capita Importer 2.45 <0.01 Capital labour ratio Exporter Capital labour ratio Importer Land labour ratio Exporter Land labour ratio Importer Exchange rate 1.35 <0.01 Average tariff rate Linear trend NAFTA BritishColumbia US NAFTA Alberta US NAFTA Saskatchewan US NAFTA Manitoba US NAFTA Ontario US NAFTA Quebec US NAFTA New Brunswick US NAFTA Nova Scotia US NAFTA Prince Edward Island US <0.01 NAFTA Newfoundland US NAFTA British Columbia Canada <0.01 NAFTA Alberta Canada <0.01 NAFTA Saskatchewan Canada NAFTA Manitoba Canada <0.01 NAFTA Ontario Canada NAFTA Quebec Canada NAFTA New Brunswick Canada NAFTA Nova Scotia Canada NAFTA Prince Edward Island Canada <0.01 NAFTA Newfoundland Canada Prince Edward Island stems from the opening of the Confederation Bridge in June 1997 as this significantly decreased trade costs for the province. However, if this were in fact the case a significant increase in interprovincial trade would also be expected. As discussed below, this did occur but the effect on international trade with the US is much greater. Turning to central Canada, the overall results are qualitatively similar (in terms of coefficient sign) to that found by Wall (2003), with Ontario and Quebec experiencing a positive effect of approximately 5%. Western Canada also differs significantly from Wall (2003) in that the overall impact of the NAFTA on western provinces is positive. British Columbia and Alberta exhibit highly insignificant effects

11 The geographical effects of the NAFTA on Canadian provinces 261 from the NAFTA, Saskatchewan exhibits a statistically significant positive effect of 21.23%, while Manitoba exhibits a statistically significant positive effect on international trade to the US of 17.01%. The overall result presented here is that the NAFTA has had a positive impact on trade for the individual Canadian provinces. British Columbia s, Alberta s, and New Brunswick s NAFTA effects are negative but in such a low magnitude that it is inconsequential. Not all of the provinces with significant and positive NAFTA effects are significantly impacted by the non-tariff changes in the NAFTA. The insignificant impact for many Canadian provinces may be a result of multicollinearity between the provincial NAFTA variables. However, the correlations between the provincial NAFTA variables are never greater than 0.102, indicating that multicollinearity is not problematic. 8 In order to better assess this issue, and provide results comparable to previous research, the provincial NAFTA variables are aggregated to the traditional regional grouping of western, central, and eastern Canada. The results of this aggregation (not shown for brevity) is that a great deal of variation in NAFTA s impact is lost. For example, the significant results for Saskatchewan and Manitoba become statistically insignificant when aggregated to western Canada. Little changes for central Canada because Ontario and Quebec have similar parameter estimates. Lastly, eastern Canada has the only positive and statistically significant result, indicating a 17.54% increase in international trade; this last result is clearly driven by Prince Edward Island s effect that is reduced because of the inclusion of other eastern provinces. Consequently, the present analysis exemplifies the impact of aggregating provinces and states into larger sub-national regions to assess the NAFTA, or any other policy change: aggregating provinces into larger sub-national regions conceals the distinctive effects of the NAFTA on those provinces. 4.2 International and interprovincial trade and the NAFTA Turning to the international interprovincial results, presented in Tables 3 and 4, all of the non-nafta variables in both regression results are quite similar to the two regression models discussed above; so they are not discussed. The results for the effects of the NAFTA on interprovincial and international trade to the US, however, are worthy of independent comment. The effect of the NAFTA on international trade to the US is positive, statistically significant, and very close in magnitude to the effect of thenaftashownintable1. The effect of the NAFTA on interprovincial trade is also positive and statistically significant, increasing interprovincial trade by 15.29%. Therefore, at the national level, the effect of the NAFTA has been to increase both international and interprovincial trade. The geography of the effects of the NAFTA on international trade to the US and Canadian interprovincial trade are shown in Table 4. Of the ten effects of the NAFTA on interprovincial trade, seven of them are statistically significant, some positive and 8 The correlations between the provincial NAFTA variables and the other independent variables are most often below 0.25 and never greater than Therefore, none of the correlations are above 0.80, a common threshold for a concern regarding multicollinearity.

12 262 M. A. Andresen some negative, generally confirming the national statistically significant result. Two provinces exhibit both statistically significant and negative effects from the NAFTA on interprovincial trade: British Columbia and Ontario. British Columbia s effect is not only negative and statistically significant, but high in absolute magnitude (a decrease in interprovincial trade of 30.39%). The negative coefficient for British Columbia is a curious result because British Columbia has increased its interprovincial trade over this study period by 27%, significantly more than Ontario s and Quebec s meagre increases of 7 and 8%, respectively. British Columbia has decreased its share of total trade to and from other Canadian provinces, but those decreases are common for all Canadian provinces. In fact, Alberta, Saskatchewan, Ontario, Quebec, and Nova Scotia all have greater relative decreases in their shares of interprovincial trade than British Columbia (Statistics Canada 1998, 2000, 2005a). Ontario s negative and statistically significant interprovincial NAFTA effect is not of the same magnitude as British Columbia, but a noteworthy 13.01%. Similar to the NAFTA effects discussed above, when the provincial NAFTA variables are aggregated into the three traditional Canadian regions much of the interesting results are lost. The impact of the NAFTA on international trade is statistically insignificant for western and central Canada and statistically significant and positive for eastern Canada, similar to the results reported above. However, the effect of the NAFTA on interprovincial trade is statistically significant in all three regions: positive in western and eastern Canada, negative in central Canada. The result for central Canada is representative of the provincial NAFTA results, but not for western and eastern Canada. New Brunswick has an insignificant but negative interprovincial NAFTA effect and Nova Scotia s interprovincial NAFTA effect is statistically insignificant. More importantly, however, is that Alberta, Saskatchewan, and Manitoba s positive interprovincial NAFTA effects dominate British Columbia s negative effect. Clearly valuable information has been lost. The utility of using individual provinces and states is manifest, rather than aggregated sub-national regions. Overall, the effects of the NAFTA are positive for the Canadian provinces with regard to international trade to the US, without any significant loss of interprovincial trade aside from the loss for British Columbia that is more than compensated by Alberta s increase in interprovincial trade. 4.3 Implications of these results The implications of these results are twofold. First, and quite clear from the discussion of the regression results, is that any aggregated analysis of the effects of the NAFTA shrouds the effects on smaller geographical units. As such, when investigating the impacts of a national policy, the spatial unit of analysis must be chosen with care. It may be true that the NAFTA has not had an impact on some provinces, but that is not the case for all provinces. Second, the NAFTA has been found to affect Canada s trade with itself and the US, despite the fact that the CUFTA had already been in force. Because of these (geographical) effects of the NAFTA in the presence of another free trade agreement, any future national policy formation regardless of how significant it is considered to be

13 The geographical effects of the NAFTA on Canadian provinces 263 must consider its varied impacts across the Canadian economic landscape. As with the NAFTA, further integration with the US may be considered insignificant with regard to its impact on Canada. However, as discussed above, if an agreement alters the US relationship with Mexico, for example, that agreement may have implications for Canada. 5 Conclusions and directions for future research This paper has investigated the effects of the NAFTA in Canada as a whole and its provinces. Past research on the effects of the NAFTA have had difficulty finding any effect because of a lack of post-nafta data, but that difficulty is not present in this analysis having nine post-nafta years of international and interprovincial trade data. Only two studies have investigated the geographical effects of the NAFTA with one study focusing on the US and the other focusing on Canada. Though the latter study used aggregated sub-national regions for its analysis in both Canada and the US, an overall positive effect of the NAFTA is found with that effect varying across Canada s economic landscape. The results presented in this paper both complement and extend this past research: the NAFTA has indeed had a significant positive effect on Canada s international trade with the US, a statistically significant and positive impact on interprovincial trade, and the geographical effects of the NAFTA are best viewed at the provincial level. Using the Canadian provinces and US states as the spatial units of analysis, it is found that the geographical effects of the NAFTA are overwhelmingly positive for international trade to the US and that only British Columbia has experienced a significant negative change in interprovincial trade resulting from the NAFTA. Natural directions for future research are to investigate the driving forces behind these changes. For example, the NAFTA altered the mean center of the consumer market in North America. Because of this change, firms may be changing locations within the same country or to another country, altering trading patterns. Alternatively, provincial industrial level analyses can be undertaken to further refine the spatial scale of NAFTA effects. References Anderson JE (1979) A theoretical foundation of the gravity equation. Am Econ Rev 69(1): Anderson JE, van Wincoop E (2003) Gravity with gravitas: a solution to the border puzzle. Am Econ Rev 93(1): Baldwin JR, Brown WM (2004) Regional manufacturing employment volatility in Canada: the effects of specialisation and trade. Papers Region Sci 83(3): Bergstrand JH (1985) The gravity equation in international trade: some microeconomic foundations and empirical evidence. Rev Econ Stat 67(3): Bergstrand JH (1989) The generalized gravity equation, monopolistic competition, and the factor-proportions theory in international trade. Rev Econ Stat 71(1): Britton JNH (2002) Regional implications of North American integration: a Canadian perspective on high technology manufacturing. Region Stud 36(4):

14 264 M. A. Andresen Britton JNH (2002b) Free trade and the high-technology response: a regional innovation system perspective on Toronto. In: Holbrook JA, Wolfe DA (eds) Knowledge, clusters and regional innovation: economic development in Canada. McGill-Queen s University Press, Montréal Brown WM, Anderson WP (1999) The influence of industrial and spatial structure on Canada US regional trade. Growth Change 30(1):23 47 Centre for the Study of Living Standards (2005) NAICS based Capital, labour and total factor productivity tables by province for Canada. Accessed: 21 July 2005 Cheng I-H, Wall HJ (2005) Controlling for heterogeneity in gravity models of trade and integration. Federal Reserve Bank St. Louis Rev 87(1):49 63 Contenta S (1996) Economist says world priorities misplaced. The Toronto Star, June 13 Coughlin CC, Wall HJ (2003) NAFTA and the changing pattern of state exports. Papers Region Sci 82: Coulombe S (2004) International trade diversion, the Canada United States free trade agreement and the L curve. Topics Econ Anal Policy 4, Article 16 Courchene TJ (2003) FTA at 15, NAFTA at 10: a Canadian perspective on North American integration. N Am J Econ Finan 14: Deardorff AV (1998) Determinants of bilateral trade: does gravity work in a neoclassical world? In: Frankel JA (ed) The regionalization of the world economy. The University of Chicago Press, Chicago pp 7 32 External Affairs Canada (1987) The Canada US free trade agreement: tariff schedule of Canada. External Affairs Canada, Ottawa Feenstra RC (2002) Border effects and the gravity equation: consistent methods for estimation. Scottish J Polit Econ 49(5): Frankel J, Stein E, Wei S (1995) Trading blocs and the Americas: the natural, the unnatural, and the supernatural. J Dev Econ 47(1):61 95 Gould DM (1998) Has NAFTA changed North American trade? Federal Reserve Bank of Dallas Econ Rev 1st Quarter:12 23 Greene WH (2000) Econometric Analysis,. 4th ed. Prentice Hall, Upper Saddle River Halvorsen R, Palmquist R (1980) The interpretation of dummy variables in semilogarithmic equations. Am Econ Rev 70(3): Helliwell JF (1996) Do national borders matter for Quebec s trade? Can J Econ 29(3): Helpman E, Krugman PR (1985) Market structure and foreign trade: increasing returns, imperfect Competition, and the international economy. The MIT Press, Cambridge Hoare AG (1993) Domestic regions, overseas nations, and their interactions through trade: the case of the United Kingdom. Environ Plan A 25: Hummels D, Levinsohn J (1995) Monopolistic competition and international trade: reconsidering the evidence. Quar J Econ 110(3): Kennedy PE (1981) Estimation with correctly interpreted dummy variables in semilogarithmic equations. Am Econ Rev 71(4):802 Krueger AO (1999) Trade creation and trade diversion under NAFTA. NBER Working Paper National Bureau of Economic Research, Cambridge Linneman H (1966) An econometric study of international trade flows. Amsterdam, North Holland McCallum J (1995) National borders matter: Canada US regional trade patterns. Am Econ Rev 85: Pöyhönen P (1963) A tentative model for the volume of trade between countries. Weltwirtschaftliches Archiv 90(1):93 99 Romalis J (2005) NAFTA s and CUSFTA s impact on international trade. NBER Working Paper National Bureau of Economic Research, Cambridge Statistics Canada (1992) 1991 census of agriculture. Statistics Canada, Ottawa Statistics Canada (1986) Census of agriculture, Statistics Canada, Ottawa Statistics Canada (1998) Interprovincial trade in Canada, Statistics Canada, Ottawa Statistics Canada (1999) 1996 census of agriculture. Statistics Canada, Ottawa Statistics Canada (2000) Interprovincial and international trade in Canada, Statistics Canada, Ottawa Statistics Canada (2002) 2001 census of agriculture. Statistics Canada, Ottawa Statistics Canada (2004) Canadian international trade, International Trade Division, Statistics Canada, Ottawa Statistics Canada s (2005a) Canadian socio-economic information management System (CANSIM). Statistics Canada, Ottawa

15 The geographical effects of the NAFTA on Canadian provinces 265 Statistics Canada (2005b) Census of agriculture, Statistics Canada, Ottawa Tinbergen J (1962) Shaping the world economy. The Twentieth Century Fund, New York United States Department of Agriculture (2000) Summary report: 1997 natural resources inventory (revised December 2000). Natural Resources Conservation Service Statistical Laboratory, Iowa State University, Washington, DC Wall HJ (2000) Gravity model specification and the effects of the Canada US border. Federal Reserve Bank of St. Louis Working Paper Series: Working Paper A Wall HJ (2003) NAFTA and the geography of North American trade. Federal Reserve Bank of St. Louis Rev March/April 2003:13 26