The Long and Short of the Canada-U.S. Free Trade Agreement By Daniel Trefler* The Canada-U.S. Free Trade Agreement provides a unique window onto the effects of a reciprocal trade agreement on an industrialized economy (Canada). For industries that experienced the deepest Canadian tariff cuts, the contraction of low-productivity plants reduced employment by 12 percent while raising industry-level labor productivity by 15 percent. For industries that experienced the largest U.S. tariff cuts, plant-level labor productivity soared by 14 percent. These results highlight the conflict between those who bore the short-run adjustment costs (displaced workers and struggling plants) and those who are garnering the long-run gains (consumers and efficient plants). (JEL F13, F14, F15, F16, D24) The central tenet of international economics is that free trade is welfare improving. We express our conviction about free trade in our textbooks and we sell it to our politicians. Yet the fact of the matter is that we have one heck of a time explaining these benefits to the larger public, a public gripped by Free Trade Fatigue. Why is the message of professional economists not more persuasive? To my mind there are two reasons. First, in examining trade liber- * Joseph L. Rotman School of Management and Department of Economics, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada, M5S 3E6, Canadian Institute for Advanced Research (CIAR), and the National Bureau of Economic Research (NBER) (e-mail: dtrefler@ rotman.utoronto.ca). I have benefited from the research assistance of Yijun Jiang, Huiwen Lai, Runjuan Liu, and Susan Zhu. Michael Baker, Erwin Diewert, Peter Dungan, Gerry Helleiner, Pravin Krishna, Aloysius Siow, and members of the CIAR provided key comments that dramatically improved the paper. Much more than is usually the case, this paper was dramatically overhauled in response to the creative input from two anonymous referees. Many members of Statistics Canada provided advice on data issues including Richard Barnabe (Standards Division), Jocelyne Elibani (International Trade Division), Richard Landry (Investment and Capital Stock Division), Jean-Pierre Maynard (Productivity Division), Bruno Pepin (Industry Division), and Bob Traversy (Industry Division). I am especially grateful to John Baldwin (Director, Micro-Economic Studies and Analysis Division) for making available the plant-level data. Alia Lileeva was kind enough to analyze the plant-level data in Ottawa. Research support from the Social Sciences and Humanities Research Council of Canada (SSHRC) is gratefully acknowledged. Views expressed in this paper do not necessarily reflect those of Statistics Canada or SSHRC. alization we treat short-run transition costs and long-run efficiency gains as entirely separate areas of inquiry. On the one hand are those who study the long-run productivity benefits of free trade policies, e.g., James R. Tybout et al. (1991), James Levinsohn (1993), Ann E. Harrison (1994), Tybout and M. Daniel Westbrook (1995), Pravin Krishna and Devashish Mitra (1998), Keith Head and John Ries (1999a, b), and Nina Pavcnik (2002). On the other hand are those who study the impacts of freer trade on short-run worker displacement and earnings, e.g., Noel Gaston and Trefler (1994, 1995), Ana Revenga (1997), Levinsohn (1999), Eugene Beaulieu (2000), and Krishna et al. (2001). Only Janet Currie and Harrison's (1997) study of Morocco examines both labor market outcomes and productivity. In assessing free trade policies there is clearly a bias introduced when looking only at the long-run benefits or only at the short-run costs. Nowhere is this more apparent than for the Canadian experience with the Canada-U.S. Free Trade Agreement (FTA) and its extension to Mexico. The FTA triggered ongoing and heated debates about freer trade. This heat was generated by the conflict between those who bore the short-run adjustment costs (displaced workers and stakeholders of closed plants) and those who garnered the long-run efficiency gains (stakeholders of competitive plants and users of final and intermediate goods). There is another reason why the free trade message is not more persuasive. While case-study 870 VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 871 evidence abounds about efficiency gains from liberalization (e.g., Anne O. Krueger, 1997), solid econometric evidence for industrialized countries remains scarce. When I teach my students about the effects of free trade on productivity I turn to high-quality studies for Chile (Tybout et al., 1991; Pavcnik, 2002), Turkey (Levinsohn, 1993), Cote d'lvoire (Harrison, 1994), Mexico (Tybout and Westbrook, 1995), and India (Krishna and Mitra, 1998) among others. I find these studies compelling, but I wonder whether they can be expected to persuade policy makers and voters in industrialized countries such as Canada and the United States. What is needed is at least some research focusing on industrialized countries. The Canada-U.S. Free Trade Agreement offers several advantages for assessing the short-run costs and long-run benefits of trade liberalization in an industrialized country. First, the FTA policy experiment is clearly defined. In developing countries, trade liberalization is typically part of a larger package of market reforms, making it difficult to isolate the role of trade policy. Further, the market reforms themselves are often initiated in response to major macroeconomic disturbances. Macroeconomic shocks, market reforms, and trade liberalization are confounded. Indeed, Gerald K. Helleiner (1994, p. 28) uses this fact to argue that "Empirical research on the relationship between total factor productivity (TFP) growth and ... the trade regime has been inconclusive." His view is widely shared, e.g., Harrison and Gordon H. Hanson (1999) and Francisco Rodriguez and Dani Rodrik (2001). In contrast, the FTA was not implemented as part of a larger package of reforms or as a response to a macroeconomic crisis. Second, as Harrison and Revenga (1995, abstract) note, "Trade policy is almost never measured using the most obvious indicators— such as tariffs." Tybout (2000) echoes this criticism. My study of the FTA is particularly careful about constructing pure policy-mandated tariff measures. Third, the FTA is not just about import-liberalizing policies. It is a reciprocal agreement that includes export-liberalizing policies as well. It should therefore be expected to induce a pronounced general-equilibrium relocation of resources out of import-competing sectors and into export-oriented sectors. I will examine these FTA effects on a large number of Canadian plant and industry outcomes. At the plant and industry levels the outcomes include employment and earnings of both production and nonproduction workers, skill upgrading, earnings inequality, hours of work, plant size, and labor productivity. Fourth, the FTA is a preferential trading arrangement. Such arrangements need not be welfare-improving. I will examine the two conditions usually put forward as sufficient—at least informally—for welfare gains. These are that trade creation must dominate trade diversion and that import prices must not rise (Arvind Panagariya, 2000; Krishna, 2003). Both conditions are satisfied. The backdrop of the FTA—an industrialized country, a clean policy experiment, the direct policy lever of tariffs, general-equilibrium reciprocity effects, and the long list of outcomes including employment, productivity, and prices— will be my basis for a rigorous and detailed examination of the short-run costs and long-run benefits of trade liberalization. The FTA has been the subject of several studies since its implementation on January 1, 1989. Gaston and Trefler (1997) found that the FTA had no effect on earnings and only a modest effect on employment. Beaulieu (2000) found that the employment effect was primarily driven by modest nonproduction worker employment losses. Kimberly A. Clausing (2001) found evidence that the FTA raised U.S. imports from Canada (trade creation), but did not divert U.S. imports away from other U.S. trading partners. John Romalis (2004) found both trade creation and diversion. The most intriguing FTA study is by Head and Ries (1999b). They found that the FTA had little net effect on industry-level average output per plant (which they take as a proxy for scale) and a puzzling effect on Canadian plant exit (exit was induced by falling Canadian tariffs and by falling U.S. tariffs). Unfortunately, none of these papers uses plant-level data. Further, I will argue below that at least some of these papers (including my own), suffer specification issues that substantively mar the inferences drawn about the effects of the Canada-U.S. Free Trade Agreement. 872 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 The Average Canadian Tariff Rate Against: ......______Rest of World ....._______ United States^ --1-!-1-1-1-1-1-1-1-1-1-1-1-1 83 84 85 86 87 88 89 90 91 92 93 94 95 96 The Average U.S. Tariff Rate Against: 10------------------------------------------------------------- g --------------------------------------------.......----------- 0 -i-1-1-1-1-1-1-1-1-1-1-1-,-1- 83 84 85 86 87 88 89 90 91 92 93 94 95 96 Figure 1. Canadian and U.S. Bilateral Tariffs in Manufacturing (In Percents) I. The FTA Tariff Cuts: Too Small to Matter? This paper deals with the impact of FTA-mandated tariff cuts. The top panel of Figure 1 plots Canada's average manufacturing tariff against the United States (solid line) and Canada's average manufacturing tariff against the rest of the world (dashed line). The bottom panel plots the corresponding U.S. tariffs against Canada (solid line) and the rest of the world (dashed line). In 1988, the average Canadian tariff rate against the United States was 8.1 percent. The corresponding effective tariff rate was 16 percent.1 Perhaps most importantly, tar- iffs in excess of 10 percent sheltered one in four Canadian industries. Given that these industries were almost all characterized by low wages, low capital-labor ratios, and low profit margins, the 1988 tariff wall was indeed high. Similar comments apply to the U.S. tariff against Canada, albeit with less force since the average 1988 U.S. tariff was 4 percent. That one in four Canadian industries had tariffs in excess of 10 percent depends crucially on the level of aggregation. I am working with 4-digit Canadian SIC data (213 industries). If one aggregates up even to 3-digit data (105 industries), almost no industries had 1988 tariffs in excess of 10 percent. This is important because studies of trade liberalization typically do not work with such disaggregated tariff data. For example, papers by Tybout et al. (1991), Levinsohn (1993), Harrison (1994), Tybout and Westbrook (1995), Gaston and Trefler (1997), Krishna and Mitra (1998), and Beaulieu (2000) are never at a finer level of aggregation than 3-digit ISIC with its 28 manufacturing sectors. The core feature of the FTA is that it reduced tariffs between Canada and the United States without reducing tariffs against the rest of the world. Graphically, the FTA placed a gap between the dashed and solid lines of Figure 1. Letting i index industries and t index years, my measures of the FTA policy levers will be rftA: The FTA-mandated Canadian tariff concessions granted to the United States. In terms of the top panel of Figure 1, this is the solid line minus the dashed line. t%s: The FTA-mandated U.S. tariff concessions granted to Canada. In terms of the bottom panel of Figure 1, this is the solid line minus the dashed line. rf(A and t^5 capture the core textual aspects of the FTA.2 1 Both the nominal and effective tariff rates were aggregated up from the 4-digit SIC level using Canadian production weights. The standard formula used to calculate the effective rate of protection appears in Trefler (2001, p. 39). Details about construction of the tariff series appear in Appendix A. 2 Given that tariffs are positively correlated with effective tariffs and nontariff barriers to trade (NTBs), the coefficients on rf,A and tfs will capture the effects of FTA-mandated reductions in tariffs, effective tariffs, and nontariff barriers. This is exactly what I want: When VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 873 II. Econometric Strategy In this section, I lay out econometric strategies for analyzing the plant- and industry-level data. I begin with the latter. Let / index industries, let t index years, and let Yit be a Canadian outcome of interest such as employment or productivity. The FT A mandates that tariffs be reduced once a year on January 1, starting in 1989. I have data for the FTA period 1989-1996. In what follows I will define the pre-FTA period as the years 1980-1986. As will be shown in detail, this choice is useful for dealing with business fluctuations. Let Ayis be the average annual log change in Yit over period s where 5=1 indexes the FTA period and 5 = 0 indexes the pre-FTA period. That is, (1) Ayn ss (in F,1996 - In 7,1988)/(1996 - 1988) and Ay,-0 = (In F,1986 - In Yit1980)/(1986 - 1980). The FTA period changes use 1988 data because I am interested in comparing the FTA-period outcome Yil996 with its baseline level, i.e., with its level before the first round of tariff reductions on January 1, 1989.3 For k = CA and k = US, define (2) At* - (t?i1996 - t*1988)/(19 9 6 - 1988). Arff measures the change in the FTA-mandated tariff concessions extended by Canada to the United States. Likewise, At{{5 analyzing tariff concessions i am actually capturing a broader set of FTA trade-liberalizing policies. 3 Since this may cause some confusion, consider by analogy a cholesterol-reducing drug trial in which the drug is given once a year on January 1 (starting in 1989) and the patient's cholesterol level Yit is measured once a year on December 31 (starting in 1988). To measure the long-term effects of the drug one looks at Yil996 — Yil9S8 rather than ^i.i996 ~ ^i,i989 because Yil9gs describes the patient cholesterol baseline without drugs. The same logic holds for the "drug of free trade." The FTA mandates that tariffs be reduced once a year on January 1 (starting in 1989) and the plants are surveyed once a year as closely as possible to December 31. Therefore, the appropriate baseline is y/>19gg. measures the change in the FTA-mandated tariff concessions extended by the United States to Canada. What of pre-FTA period tariff concessions, which I denote by At*0? Except for the 1965 Canada-U.S. Auto Pact, all tariff rates were extended on a Most Favored Nation (MFN) basis prior to 1988. Thus, define At*0 = (ff.1986 - ifW^86 - 198°) when industry i is in the automotive sector and At! = 0 otherwise. As will be shown, setting At70 = 0 for all i or omitting the automotive sector entirely from the analysis makes no difference to the results. Additional details about At*,, including a list of industries with large absolute values of At^4 and At^5, appear in Appendix A. I am interested in a regression model that explains the impact of the FTA-mandated tariff concessions on a variety of industry outcomes: (3) Ay,, = Bs + PcaAtIa + PusAt»s + els, 5 = 0,1 where 6S is a period fixed effect. There is an obvious problem with estimating equation (3). I have no deeply satisfying way of controlling for the lack of randomization in the tariff concessions. I must thus take particular care to control both for the endogeneity of tariffs and for sources of industry-level heterogeneity that might contaminate the estimates of (3 and Pus. I turn to this task now. A. The Secular Growth Control For political economy reasons, one might expect declining industries to have high tariffs and hence deep FTA tariff concessions, e.g., Trefler (1993). To prevent mistakenly attributing secular growth trends to the FTA tariff concessions, I introduce a growth fixed effect a, into equation (3): (4) Ay,, = a, + 9S + fiCAArfsA + jS^Alf + 6fa, 5 = 0,1. As a result, fiCA and (}us only pick up FTA 874 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 impacts on industry growth that are departures from industry trend growth. B. Industry-Specific Shocks A number of Canadian industries experienced reversals of fortune in the sense that employment growth in the pre-FTA and FTA periods had opposite signs. For these industries similar reversals also appeared in their U.S. counterparts. This is indicative of industry-specific demand and supply shocks. If these reversals of fortune are a characteristic of highly protected industries, the reversals might contaminate the estimates of j3CA and f$us. Controlling for reversals of fortune begins with the observation that many industry-specific shocks that appeared in Canada also appeared in Canada's major trading partners. For example, higher oil prices affected the petroleum industry in Canada and all its major trading partners. I have industry-level data for Canada's three largest trading partners: the United States, Japan, and the United Kingdom. I use these data to control for industry-specific shocks. More formally, let Ay4 be data on Ay,-, for economy j, e.g., if Ay,, is Canadian employment growth then Ay 4 is country fs employment growth. I control for industry-specific shocks by including Ay 4 in equation (4). Note that Ay 4 may be endogenous, especially for j = US, so I will employ instrumental variables (IV) techniques. Finally, for expositional ease I will refer to Ay4 as the "U.S. control" and simply write 0.50 t Figure 2. Real Canadian Manufacturing GDP Note: gdp at fa4ctor cost, 1992 dollars. constructed shortly. Introducing Abis and Ay^5 into equation (4) yields (5) Ay,, - a,- + 6S + PcaAt™ + jB^At™ + yAy,f + 8Abis + e,„ s = 0, 1. D. Estimation Differencing (5) across periods yields my difference-of-differences baseline specification: (6) (Ay,, - Ay,0) = 0 + /3™(At 0) and nonexporters (for which (3US < 0). She has linked the Canadian plant-level data to data on the exporter status of the plant. While the match precludes using my difference-of-differences methodology, she has nevertheless been able to show that $us is positive for exporters and hugely negative for non-exporters. Why? The U.S. tariff concessions had the unexpected effect of encouraging Canadian exporters to expand their domestic operations at the expense of Canadian nonexporters. Since the majority of plants are nonexporters, pooling across exporters and nonexporters yields estimates of (ius that are close to 0. Returning to the plant-level estimates in Table 1, row 11 excludes the plant-level controls. Comparison with row 10 shows that (3CA or J3US are unaffected by the exclusion of the plant-level controls. Rows 12-15 report the IV results. A key issue is the identification of variables that satisfy the two requirements of an instrument. The most likely candidates for valid instruments are variables measuring the level of industry characteristics in 1980. For one, these level characteristics are unlikely to be correlated with the residuals because the latter are twice-differenced. Such difference of differences are far removed from levels. For another, the 1980 characteristics determine the 1980 levels of pro- tection which in turn are correlated with the tariff changes. I therefore use an instrument set that consists of 1980 log values for: (1) Canadian hourly wages, which captures protection for low-wage industries as in W. M. Corden's (1974) conservative social welfare function, (2) the level of employment, which captures protection for large industries as in the J. Michael Finger et al. (1982) high-track protection for large industries, (3) Canadian imports from the United States, and (4) U.S. imports from Canada. I also include squares and cross-products as well as any exogenous regressors. The first-stage R2s are between 0.30 and 0.40 for almost all the results in this paper. Row 12 repeats the specification of row 1, but with the two tariff regressors instrumented. $CA and $us are now much larger. Also, (5US reverses signs, suggesting that the U.S. tariff concessions raised Canadian employment. However, these results do not pass the Hausman test. The "Overld/Hausman" column reports p-values for overidentification and Hausman tests. In row 12, both the overidentification test (0.60) and the Hausman test (0.65) are above 0.01 which indicates that the instruments are valid at the 1-percent level and that endogeneity is rejected at the 1-percent level. Given the poor small-sample properties of IV estimators (Charles R. Nelson and Richard Startz, 1990), I use the 1-percent cut-off, i.e., p-values below 0.01. Row 13 reports the IV estimates for the case where the U.S. control is instrumented along with the two tariff concessions. Comparing row 13 with row 12, it is clear that endogenizing the U.S. control has no impact on the estimates of $CA and fius. Further, endogeneity continues to be rejected.8 Rows 14 and 15 repeat the IV exercises of rows 12 and 13, respectively, but starting with If this is not clear consider the following. Let xikl be some characteristic of plant k in industry i in year t, let sikl be plant k's market share and let xit = ~Lk xik^ikt be the average value of xikt. Using obvious difference notation, Axit = 2, bxuaSuu + 2, Asiktxikyt^u i.e., the total industry change can be decomposed into a within-plant change (the first term) and a between-plant or market-share shift change (the second term). The plant-level regressions deal with \xikt and thus capture within-plant changes. The industry-level regressions deal with Axit and thus capture both within-plant and market-share shift changes. 8 As someone who has tried to build a career on the endogeneity of protection (Trefler, 1993), I am surprised by the rejection of endogeneity. To investigate further, I have experimented with a much larger set of instruments drawn from 1980 and 1988 characteristics of Canadian and U.S. industries. I have also experimented with a drastically reduced instrument set. None of this makes any difference to the conclusion that endogeneity is rejected. As a result, I will report the industry-level IV results, but downplay them. Interestingly, endogeneity only comes into play when the dependent variable is imports. See below. VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 879 the plant-level baseline specification of row 10. As with the industry-level results, the 0CA and $us are much larger, but endogeneity is rejected. Indeed, endogeneity is easily rejected for every plant-level specification reported in this paper. This likely reflects the fact that tariffs, even if endogenous to the industry, are exogenous to the plant. V. Coefficient Magnitudes I have not yet properly explained the magnitudes of j3CA and 0US. Since the distribution of tariff concessions is skewed, it is of interest to know the effect of the Canadian tariff concessions on the most impacted, import-competing group of industries, i.e., on the one-third of industries with the most negative values of At£a. This group has 71 (=213/3) industries, tariff concessions ranging from —5 to —33 percent, and an average tariff concession of —10 percent. The industries are listed in Appendix Table Al. For any industry i, the Canadian tariff concessions are estimated to change employment by /3caAt£a log points. For the most impacted, import-competing group as a whole this change is given by $caAt?a where ArfjA is a weighted average of the ArfA with weights that depend on industry size. (See Appendix B for details about the weights.) It is fica&t?a that is reported in the j3 column of all the tables in this paper. From row 1 of Table 1, the most impacted, import-competing group as a whole experienced a 12-percent employment loss. A similar discussion of coefficient magnitudes applies to the most impacted, export-oriented group of industries, i.e., the one-third of industries (71 industries) with the most negative values of A-rf{5. For this group the estimated impact of the U.S. tariff concessions on employment is given by $u*At%5 where At.Y5 is the weighted average of the Aiff. $u*Ary/ is reported in the fius column of all the tables in this paper. From row 1 of Table 1, this group experienced a statistically insignificant and nonrobust 3-percent employment loss. The "Total FTA impact" (TFI) columns in this paper present the joint effect of the tariff concessions on manufacturing employment as a whole. This effect is just (8) TFI = j3caAt^a + j^AT.f where ArfjA and At.Ys are now defined as averages across all 213 industries. From the TFI column of row 1 in Table 1, the FTA reduced manufacturing employment by 5 percent. This impact is statistically significant and quite similar across all the OLS specifications. It stands in sharp contrast to Gaston and Trefler (1997) who found economically small and statistically insignificant effects of the FTA. The difference in conclusions reflects both the better data and the better methodology of the current study. Employment losses of 5 percent translate into 100,000 lost jobs and strike me as large, not least because only a relatively small number of industries experienced deep tariff concessions. Indeed, most of these lost jobs were concentrated in the most impacted, import-competing industries. For this group, with its 12-percent job losses, one in eight jobs disappeared. This number points to the very large transition costs of moving out of low-end, heavily protected industries. It reflects the most obvious of the costs associated with trade liberalization. It is difficult to be sure whether these transition costs were short-run in nature. However, two facts drawn from the most recent seasonally adjusted data suggest that they probably were short-run costs. First, the FTA had no long-run effect on the Canadian employment rate which was 62 percent both in April 1988 and April 2002. Second, Canadian manufacturing employment has been more robust than in most OECD countries. For example, between April 1988 and April 2002, manufacturing employment rose by 9.1 percent in Canada, but fell by 12.9 percent in the United States and by 9.7 percent in Japan. This suggests, albeit not conclusively, that the transition costs were short run in the sense that within ten years the lost employment was made up for by employment gains in other parts of manufacturing. VI. Labor Productivity It would be best to examine productivity using a total factor productivity (TFP) measure. Unfortunately, the Canadian ASM does not record capital stock or investment data. There is 880 THE AMERICAN ECONOMIC REVIEW Table 2—Detailed Results for Labor Productivity SEPTEMBER 2004 Canadian tariffs U.S. Business U.S. control Total FTA „ . Atca tariffs At^ conditions Ab Ayus ,. n ,, impact Construction - - - - Adjusted Overld/ - ofAfc PCA t pus t 8 t y t R2 Hausman TFI t Industry level, OLS 1 gdp, rer (2) 0.15 3.11 0.04 1.14 0.25 8.30 0.16 1.99 0.31 0.058 3.79 2 gdp, rer (0) 0.15 2.77 0.02 0.40 0.13 1.79 0.28 3.05 0.09 0.050 2.87 3 gdp {2) 0.17 3.21 0.04 1.17 0.25 5.19 0.21 2.43 0.18 0.065 3.87 4 — 0.16 2.85 0.01 0.34 0.29 3.23 0.08 0.051 2.89 5 gdp, rer (2) 0.14 2.79 0.05 1.36 0.26 8.77 0.05 0.31 0.29 0.058 2.46 6 gdp, rer (2) 0.14 2.96 0.05 1.44 0.27 8.82 0.30 0.059 3.89 7 — 0.15 2.58 0.03 0.76 0.04 0.053 2.98 8 gdp, rer (2) 0.17 2.97 0.04 0.98 0.26 8.34 0.16 1.95 0.30 0.061 3.76 9 gdp, rer (2) 0.16 3.27 0.02 0.49 0.26 8.61 0.18 2.24 0.33 0.051 3.36 Plant level, OLS 10 gdp, rer (2) 0.08 1.70 0.14 3.97 0.12 3.95 0.11 1.51 0.06 0.074 4.92 11 gdp, rer (2) 0.09 1.92 0.11 3.02 0.10 3.18 0.14 1.79 0.01 0.066 4.39 Industry level, IV 12 gdp, rer (2) 0.15 1.10 0.10 0.86 0.26 8.09 0.14 1.53 0.30 0.86/0.43 0.081 3.41 13 gdp, rer (2) 0.13 0.89 0.13 1.01 0.28 6.99 -0.08 -0.28 0.28 0.87/0.51 0.083 3.40 Plant level, IV 14 gdp, rer (2) 0.22 1.67 0.05 0.49 0.11 3.20 0.17 1.80 0.06 0.06/0.77 0.082 2.53 15 gdp, rer (2) 0.79 2.58 -0.49 -1.73 -0.19 -1.29 2.07 2.29 0.05 0.76/0.52 0.050 0.39 Notes: The dependent variable is the log of labor productivity. The estimating equation is equation (6) for the industry-level regressions and equation (7) for the plant-level regressions. The number of observations is 211 for the industry-level regressions and 3,726 for the plant-level regressions. See the notes to Table 1 for additional details. In rows 4 and 7, the business conditions variable is omitted so that business conditions are controlled for implicitly by double-differencing Ay;l — Avl0. In row 5 the U.S. control is replaced by the Japan-U.K. control discussed in the text. In row 8, the two "outlier" observations with the largest Canadian tariff cuts are omitted. In row 9, all nine observations associated with the automotive sector are omitted. In row 11, the plant controls are omitted. In rows 12 and 14, only the Canadian and U.S. tariff variables are instrumented. In rows 13 and 15, the two tariff variables and the U.S. control are instrumented. thus little alternative but to work with labor productivity. I define labor productivity as value added in production activities per hour worked by production workers.9 I deflate using 3-digit SIC output deflators.10 Table 2 reports the labor productivity results. The table has the exact 9 Trefler (2001) extensively examined the sensitivity of results to alternative definitions of labor productivity. Appendix D of the current paper shows that the results are not sensitive to redefining labor productivity as total value added (in production plus nonproduction activities) per worker (production plus nonproduction workers). This definition does not correct for hours; however, it is useful in that it is directly comparable to the way in which I am forced to define U.S. labor productivity in Ayjf. (The U.S. ASM does not report value added in production activities.) 10 Appendix D also shows that the results do not change when labor productivity is deflated by the available 2-digit SIC value-added deflators. I am indebted to Alwyn Young for encouraging me to carefully examine the issue of deflators. same format as the Table 1 employment results so that I can review it quickly. As in Table 1, endogeneity is always rejected11 and all the industry-level OLS results are similar so that I can focus on the baseline row 1 specification. From the industry-level OLS results, the Canadian tariff concessions raised labor productivity by 15 percent in the most impacted, import-competing group of industries (t = 3.11). This translates into an enormous compound annual growth rate of 1.9 percent. The fact that the effect is smaller and statistically insignificant at the plant level (row 10) suggests that much of the productivity gain is coming from market share shifts favoring high-productivity plants. Such share shifting would come about 11 The Table 2 plant-level IV results are based on an instrument set without squares or cross-products because these are rejected by the overidentification tests. VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 881 from the growth of high-productivity plants and the demise and/or exit of low-productivity plants. From the plant-level OLS results (row 10), the U.S. tariff concessions raised labor productivity by 14 percent or 1.9 percent annually in the most impacted, export-oriented group of industries (t = 3.97). This labor productivity gain does not appear at the industry level ($us = 0.04, t = 1.14) which is likely due to the fact that the U.S. tariff concessions encouraged entry of plants that are less productive by virtue of being young. (On the low productivity of young plants see John R. Baldwin, 1995, for Canada and Andrew B. Bernard and J. Bradford Jensen, 1995, for the United States.) The importance of controlling for plant age can be seen by comparing rows 10 and 11 since the latter excludes the plant age control and has a lower 0US 12 The last column of Table 2 looks at the total FTA impact on all of manufacturing. The plant-level numbers of row 10 indicate that the FTA raised labor productivity in manufacturing by 7.4 percent or by an annual compound growth rate of 0.93 percent (t = 4.92). The industry-level numbers are about the same. These numbers, along with the 14-15 percent effects for the most impacted importers and exporters, are enormous. The idea that an international trade policy could raise labor productivity so dramatically is, to my mind, remarkable. VII. Import Prices and Trade Creation/Diversion: Implications for Welfare Preferential trade arrangements, including the FTA, need not be welfare improving. The literature identifies two conditions which, if satisfied, increase the likelihood of welfare gains for a representative domestic agent. These are that trade creation "dominates" trade diversion and 12 Another contributing factor to the difference between the 0US at the industry and plant levels is that the U.S. tariff concessions encouraged Canadian plants to enter the U.S. market. This must reduce average productivity because new Canadian exporters are less productive than old Canadian exporters (Baldwin and Wulong Gu, 2003). Expansion into the U.S. market therefore increases the market share of lower productivity new exporters, thus reducing the industry-level productivity effect. that import prices do not rise (Panagariya, 2000; Krishna, 2003). This section explores these conditions. A. Trade Creation and Trade Diversion Krishna (2003) offers a precise expression for welfare gains in terms of the relative sizes of trade creation and diversion. Let A In misj be the log change in Canadian imports of industry / in period s from region j = US or j = ROW (rest of the world). Let At,-^ be the corresponding change in the Canadian tariff. Krishna shows that a sufficient condition for welfare gains is (9) _o.8^!S-o.2^P=>0 At.-h/s At.-u/s where 0.8 is the share of Canadian imports originating from the United States.13 The first term is proportional to a utility-relevant measure of trade creation and is positive because A In mnus/ATnus < 0. The second term is proportional to a utility-relevant measure of trade diversion and is likely negative because A In mnROW/ATnus is likely positive. I examine equation (9) empirically as follows. The first row in Table 3 reports estimates of my standard equation (6) using Canadian imports from the United States as the dependent variable. Note that there is no U.S. control in this regression because it makes no sense in an 13 To derive equation (9), start with equation (10) in Krishna: riusdmius/dTius + TiROWdmiROW/dTiUS where all variables relate to 1988. Since tws = riROW in 1988, this expression can be rewritten as --1-[6iusd In m,vs/d miUS "r" miROW + (1 - 6ius)d \nmiFOW/dTiUS] where 8iUS = mius/(mius + miROW) = 0.8 is the U.S. import share. Krishna's analysis looks at a representative consumer in an economy with a single final good. The generalization to many goods is trivial as long as expenditure shares for each good are independent of the tariff, e.g., Cobb-Douglas preferences. In examining equation (9) empirically, i ignore the fact that Krishna's miUS and miROW are compensated demands for imports. 882 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 Table 3—Trade Diversion/Creation and Import Prices Canadian tariffs U.S. tariffs Total FTA impact Business conditions Adjusted R2 Overld/ Hausman Observations Variable ßCA t pus t TF1 t S Canadian imports from the United States OLS Industry 0.54 -4.67 0.16 -2.16 0.01 0.83 0.22* 0.24 211 IV Industry 2.32 0.80 -0.86 -0.40 -0.15 -0.48 0.30 0.15 NA/0.28 211 Canadian imports from the rest of the world OLS Industry -0.40 2.67 0.08 -0.17 0.03 0.12 0.11* 0.05 211 IV Industry -1.60 -0.54 1.24 0.48 0.22 0.47 0.08 0.04 NA/0.75 211 Canadian import prices OLS Product -0.004 0.20 0.00 4,700 IV Product -0.073 2.26 0.00 0.51/0.03 4,700 Canadian import quantities OLS Product 0.70 15.12 0.05 4,700 IV Product 1.02 12.68 0.04 0.87/0.00 4,700 Notes: The dependent variable is indicated in bold font at the start of each block of results, e.g., "Canadian imports from the United States." All dependent variables are in log changes. The estimating equation is equation (6) for the industry-level Canadian imports regressions and equation (10) for the product-level import price and quantity regressions. The business conditions variable is the same as in the Table 1, row 1 baseline specification. The U.S. control is not included because it makes no sense in a bilateral import context. $CA and fius are scaled as described in the notes to Table 1. An asterisk indicates statistical significance at the 1-percent level. The "Overld/Hausman" column reports p-values for the overidentification and Hausman tests. Rejection of the instrument set or exogeneity are indicated by p-values of less than 0.01. Blank entries indicate OLS estimation. The product-level import results use wages, employment, squares, and cross-products as instruments. Based on the overidentification test, the industry-level import results drop the squares and cross-products from the instrument set. It is thus just identified (NA). import context. The Canadian tariff concessions raised Canadian imports from the United States by 54 log points. I therefore set A In mausl Arilus equal to —0.54. The third row in Table 3 reports my OLS estimates of equation (6) using Canadian imports from the rest of the world as the dependent variable. The Canadian tariff concessions lowered Canadian imports from the rest of the world by 40 log points. I therefore set A In milROW/hTnus equal to +0.40.14 Plugging —0.54 and +0.40 into equation (9) yields -0.8 X (-0.54) - 0.2 X (0.40) = 0.35 (t = 3.62). Since this number is statistically greater than zero, Krishna's (2003) welfare condition is satisfied. This conclusion is robust to the many alternative specifications described in Tables 1-2. Thus, FTA trade creation dominated FTA trade diversion enough to ensure that the FTA improved the welfare of the "representative" Canadian. B. Prices A preferential trading agreement will not likely be welfare improving if it raises prices (Panagariya, 2000). Clearly the FTA is unlikely to have raised import prices—this would require either some unusual change in the strategic interactions between firms or a rise in tariffs against non-FTA trading partners. More likely the FTA reduced import prices by allowing U.S. producers to send larger quantities per shipment, thus spreading fixed shipping costs over a larger number of units. Fixed costs of shipping are sufficiently large that reducing them has been a key focus of Canadian public policy.15 Surprisingly, there exists very little econometric work on the effects of trade liberalization on import prices. J. Richard Huber (1971) is a rare exception. To investigate, I examine the relationship 14 Using U.S. rather than Canadian imports, Romalis (2004) finds large impacts of both the FTA and NAFTA on U.S. trade creation and diversion. 15 See the C. D. Howe Border Papers series for reviews of the public policy discussions, e.g., Wendy Dobson (2002). VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 883 between tariff cuts and changes in import unit values. Both these variables are available at the 10-digit Harmonized System (HS10) level. While unit values are difficult to interpret as prices, the hope is that at this detailed level of disaggregation, changes in unit values over the FT A period reflect changes in prices. Note that I am looking only at unit-value changes within an HS10 item. This is very different from and less problematic than the typical use made of unit values. Typically, researchers draw conclusions from the fact that one HS10 item has a higher unit value level than another. Since unit values are based on actual payments net of import duties, freight, insurance, and other charges, I will interpret changes in unit values as changes in producer prices. Canadian trade data was first collected in the HS system in 1988.16 Let At,v be the FTA period change in Canada's tariff against country j for HS10 product /. Let A In paj be the corresponding log import price change. Since I do not have pre-FTA data on import price changes at the HS10 level (A In pi0j), I cannot estimate my standard equation (6) with A In pnus -A In pmuS as the dependent variable. However, if the FTA had never been implemented one expects A In pnus to have evolved in the same way that Canada's import prices from other advanced economies evolved. I thus estimate (10) Alnp„us-Aln PiWECD = a + J3ca(At/1£/s - At,10£CD) + e, where A In pnoECD is the simple average of the A In pnj for the United Kingdom, Germany, France, and Japan. Likewise for &TilOECD. The third block of results in Table 3, labeled "Canadian import prices," reports the estimates. The OLS estimate indicates that the FTA did 16 In matching 1988 data with 1996 data I lose 33 percent of the 1988 HS10 items. There is some evidence that the loss is nonrandom in that the average tariff on the unmatched commodities is 0.5 percentage points lower than on the matched commodities. This reflects the fact that many of the unmatched commodities are in high-tech industries. For example, Intel's introduction of the 486 CPU in 1989 quickly led to the demise of the 386 CPU. (Don't date yourself by admitting you remember this!) not raise import prices ($CA = —0.004). There is modest evidence of endogeneity at the 3-percent level and the IV estimates indicate that the FTA reduced import prices by 7 percent for the most impacted, import-competing products. One wonders if the HS10 import price changes are so noisy that these results are meaningless. Import prices are defined as import values divided by import quantities so that any noisiness in prices must come from noisiness in quantities. To investigate the role of noise, I reestimated equation (10) using log import quantity changes as the dependent variable. The fourth block of results in Table 3 reports the results. The FTA raised import quantities by 70 percent and the r-statistic is huge (15.12). Further, for the first time in this paper I obtain the expected strong rejection of the exogeneity of tariffs. Thus, noise does not appear to be a problem. To summarize, two conditions increase the likelihood that a preferential trade arrangement is welfare improving: trade creation must dominate trade diversion and import prices must not rise. Both of these condition are met in the FTA context. VIII. Employment of Production and Nonproduction Workers I am now in a position to quickly review the results for other outcomes. The data distinguish between workers employed in manufacturing activities and nonmanufacturing activities. I will refer to these as production and nonproduction workers since the distinction broadly follows that used in the U.S. ASM. In particular, nonproduction workers are more educated and better paid. The top block of results in Table 4 reports a limited number of specifications for the employment of production workers. My baseline industry- and plant-level specifications appear in rows 1 and 10, respectively. (Row numbers match those of Table 1 so that the reader can always remind herself of the specification details of any row by referring back to the detailed discussion surrounding Table 1.) The results indicate that the Canadian tariff concessions reduced employment by a large amount, 14 percent using industry-level estimates (t - -2.44), and 9 percent using 884 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 Table 4—Employment and Skill Upgrading Total FTA Business U.S. Canadian tariffs U.S. tariffs impact conditions control Adjusted R2 Overld/ Hausman Variable PCA t |8ro t TFI t 8 y Employment—Production workers 1 Industry -0.14 -2.44 -0.07 -1.56 -0.08 -3.44 0.37* 0.16 0.33 4 Industry -0.13 -1.99 -0.07 -1.36 -0.08 -2.89 0.21 0.07 6 Industry -0.16 -2.93 -0.08 -1.71 -0.09 -4.08 0.37* 0.32 12 Industry -0.20 -1.28 0.03 0.17 -0.06 -1.60 0.37* 0.16 0.32 0.59/0.70 10 Plant -0.09 -2.58 -0.03 -0.87 -0.04 -3.01 0.17* 0.29* 0.04 Employment- -Nonproduction workers 1 Industry -0.06 -0.71 0.05 0.79 0.00 0.02 0.36* 0.07 0.26 4 Industry -0.07 -0.77 0.05 0.73 -0.00 -0.09 0.14 0.00 6 Industry -0.06 -0.79 0.04 0.71 -0.00 -0.12 0.36* 0.26 12 Industry 0.01 0.06 0.11 0.52 0.05 1.22 0.36* 0.11 0.25 0.18/0.36 10 Plant -0.14 -3.02 0.04 1.19 -0.03 -1.72 0.02 0.15 0.01 Skill upgrading 1 Industry 0.11 1.41 0.10 1.67 0.08 2.72 0.47* 0.24 0.48 4 Industry 0.08 0.79 0.11 1.26 0.07 1.81 0.24 0.01 6 Industry 0.12 1.63 0.10 1.56 0.08 2.82 0.47* 0.48 12 Industry 0.11 0.50 0.15 0.74 0.10 2.21 0.47* 0.25 0.48 0.11/0.83 10 Plant -0.01 -0.30 0.04 1.48 0.01 0.96 0.05* 0.17 0.01 Notes: The dependent variable is indicated in bold font at the start of each block of results, e.g., "Employment—Production workers." The estimating equation is equation (6) for the industry-level regressions and equation (7) for the plant-level regressions. Row numbers correspond to those in Table 1 so that the reader can refer to Table 1 for details of the specification. Rows 1 and 10 are my baseline specifications. See notes to Table 1 for further details, including the scaling of the flCA and fius. An asterisk indicates statistical significance at the 1-percent level. Skill upgrading is the log of the ratio of nonproduction workers to production workers. All dependent variables are in logs. The number of observations in the industry-level (plant-level) regressions is 211 (3,742) for production workers, 212 (3,539) for nonproduction workers, and 211 (3,489) for skill upgrading. plant-level estimates (t = —2.58). The effects of the U.S. tariff concessions are less clear. They reduced employment by 7 percent using industry-level estimates, but this is not statistically significant and virtually disappears in the plant-level estimates. The total FTA impact of 8 percent (industry level) and 4 percent (plant level) are both economically large and statistically significant. Rows 4, 6, and 12 present alternative specifications. In rows 4 and 6 the business conditions control and the U.S. control are excluded, respectively. This does not affect the 0CA or |8 . In row 12, the industry-level IV results are reported. Endogeneity is strongly rejected (p = 0.99). I do not report the plant-level IV results because endogeneity is always strongly rejected at the plant level. In contrast to the results for production workers, nonproduction worker employment is esti- mated to have been unaffected by the U.S. tariff concessions. Finally, the "Skill upgrading" block of results in Table 4 show that there has been FTA-induced skill upgrading, i.e., an increase in the ratio of nonproduction workers to production workers. This happened at the industry level much more than at the plant level which means that market shares have shifted in favor of nonproduction-worker-intensive plants. Possibly these workers are a fixed cost that is needed to penetrate U.S. markets. IX. Earnings Most commentators expected Canadian wages to fall in response to competition from less unionized, less educated workers in the southern United States. Table 5 revisits this question using payroll statistics. Since the VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 885 Table 5—Earnings, Wages, Hours, Inequality, and Output Canadian tariffs U.S. tariffs Total FTA impact Business conditions U.S. control Adjusted R2 Variable /3CA t ßus t TFI t 5 7 Earnings—All workers 1 Industry 0.05 2.43 0.03 1.92 0.03 3.80 0.34* 0.25* 0.20 10 Plant 0.04 2.92 0.04 3.60 0.03 5.64 0.17* 0.19* 0.03 Earnings—Production workers 1 Industry 0.04 2.12 0.00 -0.02 0.02 3.61 0.16* 0.11 0.07 10 Plant 0.05 3.25 0.03 2.57 0.03 4.74 0.12 0.21 0.02 Earnings—Nonproduction workers 1 Industry 0.01 0.30 -0.01 -0.29 0.00 0.02 0.18* 0.12 0.08 10 Plant 0.04 1.48 0.06 2.87 0.03 3.67 0.11 0.11 0.01 Hourly wages of production workers 1 Industry 0.05 3.15 0.03 1.84 0.03 4.37 0.60* 0.13 0.33 10 Plant 0.06 3.23 0.02 1.40 0.03 4.04 0.20 0.16* 0.01 Annual hours of production workers 1 Industry -0.01 -0.48 -0.02 -1.75 -0.01 -1.94 0.02 0.14 0.01 10 Plant -0.02 -0.90 0.01 0.80 0.00 -0.12 0.03 0.07 0.00 Earnings inequality 1 Industry -0.04 -1.32 -0.01 -0.55 -0.02 -1.66 0.42* 0.05 0.21 10 Plant -0.01 -0.46 0.02 0.97 0.00 0.41 0.13* 0.08 0.00 Gross output per plant in production activities 1 Industry -0.05 -0.65 0.03 0.54 0.00 -0.05 0.30* 0.18 10 Plant -0.05 -1.36 0.06 2.01 0.01 0.72 0.16* 0.05 Notes: The dependent variable is indicated in bold font at the start of each block of results, e.g., "Earnings—All workers." The estimating equation is equation (6) for the industry-level regressions and equation (7) for the plant-level regressions. Row numbers correspond to those in Table 1 so that the reader can refer to Table 1 for details of the specification. Rows 1 and 10 are my baseline specifications. See notes to Table 1 for further details, including the scaling of the fiCA and f3us. An asterisk indicates statistical significance at the 1-percent level. Earnings inequality is the ratio of nonproduction-worker earnings to production-workers earnings. The U.S. control is not included in the output equations because the published data on the number of U.S. plants are only available at five-year intervals. All dependent variables are in logs. The number of observations in the industry-level (plant-level) regressions is 213 (3,801) for the earnings of all workers, 211 (3,742) for the earnings of production workers, 212 (3,526) for the earnings of nonproduction workers, 211 (3,738) for wages, 211 (3,738) for hours, 211 (3,489) for earnings inequality, and 211 (3,751) for output. industry-level results are robust and since endo-geneity is strongly rejected, I do not report the specifications that appeared as rows 4, 6, and 12 of Table 4. For all workers, the tariff concessions raised annual earnings. For example, the total FTA impact is a rise of 3 percent at both the industry level (t = 3.80) and the plant level (t = 5.64). At the plant level, earnings rose for both production and nonproduction workers. At the industry level, earnings gains were concentrated among production workers.17 I have re- fined this observation by looking at hourly wages and hours worked by production workers. As shown in Table 5, there are wage effects and no hours effects. These earnings and wage effects are large in a statistical sense, but small in an economic sense. For example, a 3-percent rise in earnings spread over eight years will buy you more than a cup of coffee, but not at Starbucks. The important finding is not that earnings went up, but that earnings did not go down 17 My earnings results contrast sharply with those of Gaston and Trefler (1997) and Beaulieu (2000). Gaston and Trefler found no statistically significant effect of the tariff concessions on earnings. The only effect Beaulieu finds is the positive effect of U.S. tariff concessions on nonproduc- tion worker earnings (an effect I find only in the plant-level data, not the industry-level data). Once again, my improved data and methodology means that my results supersede older results. 886 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 in response to competitive pressures from the U.S. South. There are a number of reasons why earnings may have risen slightly at a time when employment was falling. First, there may have been end-game bargaining on the part of unions seeking to extract rents from nearly bankrupt firms as in Colin Lawrence and Robert Z. Lawrence (1985). To investigate, I use the Canadian Labour Force Survey which reports unionization rates in 1996 for a classification in which manufacturing is divided up into 16 industries. The correlation of Canadian tariff concessions with union membership rates and union coverage rates is 0.016 and 0.002, respectively. Thus, unionization does not offer an explanation of modestly rising earnings. Another possibility is that workers in the most impacted industries upgraded their skills, possibly through the attrition of less-skilled workers. The Labour Force Survey is the most detailed source of data on education by industry. It reports education on a consistent basis back to 1988 (but not 1980). The correlation of Canadian tariff concessions A iff1 with 1988-1996 log changes in average years of schooling is —0.28 which supports the view that the tariff cuts were associated with educational upgrading. However, this correlation is almost completely driven by the Clothing industry. The correlation falls to —0.06 when Clothing is omitted. Note of course that the Clothing industry is too important for an analysis of the FTA to simply be dismissed as an outlier. Thus, while there is some evidence that the earnings effect is driven in part by educational upgrading, this conclusion must be tentative. The explanation of modestly rising earnings best supported by the data is seniority-based worker attrition. The Labour Force Survey reports current job tenure over the 1980-1996 period. Let A In Tenureis be the average annual log change in tenure in the pre-FTA period (s = 0) or FTA period is = 1). Figure 3 plots A In Tenuren — A In Tenurei0 against ATnA — At^4. That is, it has the form of my usual difference-of-differences estimator. As is apparent, industries that experienced the deepest tariff cuts (and hence the deepest employment losses) also experienced the largest increases in current job tenure. The correlation is —0.45. o 3 u J3 u 00 o -J 0.015 0.010 0.005 0000 ■ -0.005 -0.010 -0.015 ■ -0.020 ■ -0.025 - -0.030 > Clothing »Textiles »Furniture ß = -1.56 p = -0.45 • Chemicals -0.014 -0.012 -0.010 -0.008 -0.006 -0.004 -0.002 -0.000 - 0.002 Canadian Tariff Concessions (r™ - r,™) Figure 3. Current Job Tenure Changes [(1996-1988) less (1986-1980)] vs. Canadian Tariff Concessions The wage results point to a potential caveat for the labor productivity results. The 5-percent earnings rise associated with the Canadian tariff concessions may in part reflect a rise in labor quality. At one extreme, if the earnings rise was entirely due to increased labor quality then labor productivity rose not by 15 percent, but by 15 — 5 = 10 percent. This translates into a compound annual growth rate of 1.2 percent, still an enormous number. At the other extreme, if productivity increases drove wage increases (i.e., if there was no labor quality increase), then no correction to the productivity numbers is needed. There is a presumption in the popular press that anything to do with globalization will worsen income inequality. It is thus reassuring that there is absolutely no evidence that the FTA worsened income inequality. In the last block of results in Table 5, where inequality is measured as the earnings of nonproduction workers rela- are tive to production workers, ß effectively 0. CA and j8 us X. What Underlies Rising Labor Productivity? To the extent that the labor productivity benefits of the FTA reflect gains in technical efficiency (as opposed to allocative efficiency), it is VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 887 of interest to know how this came about. This section examines three possibilities. First, plants may have moved down their average cost curves. To examine this I estimated my industry-level equation (6) for average output per plant and my plant-level equation (7) for plant output. The results appear at the bottom of Table 5. The industry-level $CA and pus are comparable in magnitude to those estimated by Head and Ries (1999b) though my significance level is much lower.18 Their finding of statistical significance may reflect their decision to work with annual changes without correcting for serial correlation. The more interesting results are at the plant level since these are more readily interpretable as moving along an average cost curve. The results indicate that the Canadian tariff concessions led the most impacted, import-competing plants to contract by 5 percent (t = 1.36) while the U.S. tariff concessions led the most impacted, export-oriented plants to expand by 6 percent (t = 2.01). These are not statistically significant results. Thus, this is not strong evidence in support of a simple scale-effects explanation of labor productivity gains. Second, the popular press reports that U.S.-owned multinationals have been reorganizing their Canadian plants in order to produce fewer product lines, each with a global mandate. This is consistent with Baldwin et al. (2002) who find that for foreign-owned plants operating in Canada, increases in exports are associated with reductions in the number of commodities produced. Thus, plant rationalization may have contributed to rising productivity. Third, it is possible that my FTA-induced labor productivity gains do not extend to TFP gains. However, this seems unlikely since there is little evidence of capital deepening, more intensive use of intermediate inputs, or rising markups. Specifically, using my difference-of-differences methodology, Trefler (2001) finds (1) no evidence of capital deepening at the 3-digit SIC level (capital stock is not available at the 4-digit level), (2) evidence of only very 18 Head and Ries (1999b) find #CA = -0.11 with t = 3.08 and $us = 0.06 with t = 2.74. (For comparability, I have scaled their estimates.) modest increases in the usage of intermediate inputs at the 4-digit SIC level, and (3) no evidence of increased markups (not a surprise given that the most impacted import-competing industries are low-end manufacturing industries with low markups to begin with). Thus, the Robert E. Hall (1988) TFP calculation shows that TFP must have risen substantially. More exactly, Trefler (2001) argues that the FTA-induced TFP changes are roughly half of the labor productivity changes. That is, the TFP changes are huge. XL Conclusions There are many ways in which the Canada-U.S. Free Trade Agreement provides a unique window onto the effects of freer trade. The FTA was a relatively clean policy experiment, untainted by macro shocks or financial crises. It was an agreement between two industrialized countries. It was a reciprocal agreement, which means it affected exporters, not just importers. In contrast, most previous studies of trade liberalization have dealt with the unilateral trade actions of a developing country. Several strong conclusions emerged from the analysis. First, the FTA was associated with substantial employment losses: 12 percent for the most impacted, import-competing group of industries and 5 percent for manufacturing as a whole. These effects appear in both the industry- and plant-level analyses. Second, the FTA led to large labor productivity gains. For the most impacted, export-oriented group of industries, labor productivity rose by 14 percent at the plant level. For the most impacted, import-competing group of industries, labor productivity rose by 15 percent with at least half of this coming from the exit and/or contraction of low-productivity plants. For manufacturing as a whole, labor productivity rose by about 6 percent which is remarkable given that much of manufacturing was duty-free before implementation of the FTA. Third, the FTA created more trade than it diverted and possibly lowered import prices. Thus, the FTA likely raised aggregate welfare. The FTA is the wellspring of one of the most heated political debates in Canada. This heat is 888 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 generated by the conflict between those who bore the short-run adjustment costs (displaced workers and stakeholders of closed plants) and those who are garnering the long-run gains (stakeholders of efficient plants, consumers, and purchasers of intermediate inputs). One cannot understand current debates about freer trade without understanding this conflict. Unfortunately, much of the academic debate has been fragmented: one set of researchers has focused on the short-run adjustment costs of worker displacement while another has focused on the long-run productivity gains. While this paper does not provide the silver bullet that makes the case either for or against free trade, I believe that it has considerably refined the question. My hope is that the results here take us one step closer to understanding how freer trade can be implemented in an industrialized economy in a way that recognizes both the long-run gains and the short-run adjustment costs borne by workers and others. Appendix A: Tariff Details The Canadian tariff data were supplied by Statistics Canada at the 4-digit SIC level. The U.S. tariff data were constructed as follows. The 1980-1988 data were converted from the TSUSA classification system (approximately 10,000 products) to SITC (revision 2) (approximately 800 products) using Feenstra's (1996) converter. It was then converted to Canadian SIC (213 industries) using a converter supplied by Statistics Canada. This converter was largely unique but, where not, weights for prorating data across SIC industries were supplied by Statistics Canada. For 1989-1994 tariff rates, the same procedure was followed, but starting from HS10 rather than TSUSA. For 1996 data, I converted the Census Bureau's "U.S. Imports of Merchandise: December 1996" (CD-96-12) data from HS10 to SITC (revision 3) using the supplied converter. I then converted the data to SITC (revision 2) using an almost 1:1 converter supplied by Feenstra (1996) and proceeded as with the 1980-1988 data. Of Canada's 225 4-digit SIC industries, four were excluded from the analysis because of incomplete data and another 16 were aggregated into eight categories in order to ensure consistency of the trade and tariff data over time. The aggregated industries are: 1094 and 1099; 1511 and 1599; 1995 and 1999; 2911 and 2919; 2951 and 2959; 3051 and 3059; 3351 and 3359; 3362 and 3369. The tariff data are defined as duties divided by imports. These data are collected at the tariff-line level (e.g., HS10 after 1988). I have compared a large number of the tariff rates so derived with published statutory tariff rates. The two tariff rate series are the same. A key issue is how to aggregate the tariff-line data up to the 4-digit SIC level. Since imports are the only data reported at a comparable level of disaggregation, I must follow what all empirical trade researchers do and aggregate using import weights. This is accomplished in the usual way as follows. Consider a single 4-digit SIC industry, let i be an HS10 item feeding into the industry, let / be the set of HS10 items feeding into the industry, let rit be the tariff rate, and let mit be the share of the industry's imports accounted for by product i. My tariff rate changes have the form At = S,-G/ ^/n^ — 2,-e/ Tit_xmit_v For later reference, At = 2,.e/ (jit - TUt-X)mit + (mit - mUt_x)TUt^v Ideally I would prefer to use fixed-weight tariffs ATflxed = 2i(E/ (t,7 — t; t_\)mit_,. However this cannot be calculated because about one-third of all 1988 HS10 items disappeared by 1996. (Companies often hire lawyers to have their HS10 product reallocated to a higher tariff HS10.) To get a handle on the difference between A'/lxed and At, I manipulated the estimates of ^xed that were used by the Government of Canada in its pre-FTA assessment of the likely impacts of the agreement (S. Magun et al., 1988). To understand what I did, note that most industries had their tariffs reduced to zero linearly either over five years or ten years. Using Magun et al. (1988) I classified 4-digit SIC industries into either the five- or ten-year category. (The Magun et al. study reported estimates of !A'flxed using an input-output table classification that breaks manufacturing into about 60 industries.) In the formula Ar'xed = 2,e/ (t(1996 — t; i988)m, ]988 I set ril996 = 0 for five-year industries and Ti,i996 = 0-20ti>1988 for ten-year industries. This allows me to compute A'/,xed. The outcome of this procedure is estimates of ^T^fixed and ^^■fixed where I am using the VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 889 notation of equation (2). Across 4-digit SIC industries the correlation of ATflA'fixed with At^ is 0.98 and the correlation of LrYiS,fixed with At^5 is 0.97. That is, my tariff rate changes are very similar to a best estimate of fixed-weight tariff changes. Not surprisingly, the two tariff-change series yield almost identical results for estimates of )3CA and (Sus. Trefler (2001, Appendix 2) discusses further aspects of aggregation. Table Al reports At^a and At(^s for the most impacted, import-competing industries. Appendix B: Scaling /3ca and fius and Defining "Total FTA Impact" Recall that FU988 is the level of, say, employment in industry i in 1988. The industry i change in employment over the FTA period is approximately 8(Avll)y, 1988, i.e., the log change times the initial level. Multiplying by eight converts the average annual changes for the eight FTA years into a total FTA period change. The change in employment among industries in any set / is approximately 8 S,e/ (Ava)y,. 1988. As a proportion of total employment it is 8 S,G/ Ayno),- where is requires the preliminary step of estimating j Aiy,-,= 0, + S 0i;A,zr_,. + t)„. j = 0 I use OLS since my only criterion is to minimize in-sample prediction error. This regression was estimated separately for each industry using 1983-1996 data. (I do not have data for 1982.) This leaves only 13 observations for estimating seven parameters. (6i0, 6n, and 6i2 are each tuples.) To modestly increase the degrees of freedom, I estimated the regression at the 3-digit SIC industry level rather than at the 4-digit SIC industry level. There is not much difference between the 3- and 4-digit Abis as can be seen from the fact that on average there are only 2.03 4-digit industries per 3-digit industry. Since Abis is a generated regressor, I reestimated all my results for the case where A6a - Abi0 is an endogenous regressor in equations (6) and (7). This had no impact on the results. Further tests of misspecification due to a generated regressor led to rejection of misspecification. Table A2 reports results for different choices of years. As is apparent, the results do not change substantially as long as the FTA baseline year is 1988. A referee has suggested that I also report results for the periods 1981-1988 and 1989-1996. Since the worst of the FTA adjustment happened immediately, the use of 1989 as the FTA baseline period means that I miss at least some of the adjustment. Indeed, the estimated coefficients are somewhat smaller. 19 There are some exceptions to this definition of o>;. For the cases of production worker earnings and wages, o>; is based on total hours worked by production workers. For the cases of skill upgrading and inequality o>, is based on total employment. For intraindustry trade, oj, is based on Canadian imports from the United States. Otherwise, if YiX9is is a ratio then to, is based on the numerator of the ratio, i.e., if Yil9SS = ailgsi/bilgss then to, = «,,i988/2/e/ aj,\9S8- 890 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 2004 Appendix D: Measuring Labor Productivity Table A3 reports the results for labor productivity using three alternative measures of labor productivity. The most commonly used measure of labor productivity at the industry level is value added per worker deflated by an output deflator. This is the third measure reported in Table A3. There are several defects with this measure, two of which are easily addressed. The first deals with the measurement of labor input. In Canada, but not in the United States, there has been a strong trend towards part-time employment. By not correcting for Canadian hours, measure 3 has a downward trend. Since this trend will be spuriously correlated with the downward trend in tariffs, the estimated effect of the FTA on productivity ($CA and $us) will be downward biased. The Canadian data allow for an hours correction. Unlike the U.S. data, value added is reported for production activities alone and thus can be directly compared with the data reported for hours worked. Measure 1 of Table A3 reports the estimates using Canadian real value added in production activities per hour worked and U.S. real value added in all activities per employee. This is the same measure used in Table 2. As expected, the estimates tend to be larger for measure 1 than for measure 3 (though both are large). Clearly, measure 1 is preferred. The second data issue deals with deflators. In Table A3, measures 1 and 3 use output deflators while measure 2 uses value-added deflators. Value-added deflators would have been preferable had the U.S. deflator not been seriously flawed for present purposes. It is at the 2-digit level (20 industries) and even at this highly aggregated level there are imputations for instruments (SIC 38) and electric and electronic equipment (SIC 36). Measure 2 of Table A3, the value-added deflated measure, thus has serious problems. This said, the ($CA, 0US) based on value-added deflators are very similar to the ($CA, $us) based on output deflators. This can be seen by comparing measures 1 and 2 in Table A3. See Trefler (2001, Appendix 4) for a detailed discussion of deflators. Appendix E: Plant Selection Issues As noted in Section II, subsection E, my results apply to long-form plants that were in existence in 1980, 1986, 1988, and 1996. These tend to be large plants. For example, in 1988 the average long-form plant was 2.2 times larger than the all-plant average. Note that the average long-form continuing plant was only 2.1 times larger than the all-continuing-plant average so that the large size of my plants is due to the fact that they are long-form rather than continuing per se. The available evidence suggests that long-form selection issues are of secondary importance in the current context. To see this, I begin by noting that almost every plant in Canada receives either a long-form or short-form survey so that almost the entire universe of Canadian plants is surveyed. Next, for the few industry outcomes available in the short-form survey (employment, earnings, output, and a measure of labor productivity), the estimates of /3CA and /Sus based on long-form and on long-form plus short-form plants are very similar. The exception is the estimate of f3us for employment. It implies employment losses of —4 percent using the long-form plants and —6.7 percent using long-form plus short-form plants. Thus, the conclusions from the long-form continuing plants appear to be broadly representative of all continuing plants. VOL. 94 NO. 4 TREFLER: CANADA-U.S. FREE TRADE 891 Table Al—The 71 Most Impacted, Import-Competing Industries SIC Industry description At