Trade Creation and Trade Diversion in the North American Free Trade Agreement: The Case of the Agricultural Sector

Headnote

This paper examines the effect of the U.S.-Mexico trade agreement under the North American Free Trade Agreement (NAFTA). The results suggest that U.S. agricultural imports from Mexico have been responsive to tariff rate reductions applied to Mexican products. A one percentage point decrease in tariff rates is associated with an increase in U.S. agricultural imports from Mexico by 5.31% in the first 6 years of NAFTA and by 2.62% in the last 6 years of NAFTA. U.S. imports from Mexico have also been attributable to the pre-NAFTA tariff rates. Overall, the results indicate that the U.S-Mexico trade agreement under NAFTA has been trade creating rather than trade diverting.

Key Words: agricultural sector, NAFTA, panel data, tariffs, trade creation, trade diversion

JEL Classifications: F10, F15, Q17, Q18, C31, C33

The surge of free trade agreements (FTAs) has raised the question of their effect on the countries included in the FTA and on the rest of the world (Bhagwati and Krueger; Krueger 1997). It is an issue that economists have long debated. The debate has divided economists between those who support FTAs and those who oppose them. The former group emphasizes trade-creating effects. By reducing (eliminating) trade barriers among members, FTA can improve resource allocation within a region and improve income for member countries. Production shifts toward the most efficient producers of specific commodities within the FTA and consumers are better off because they can purchase goods at lower prices. The latter group argues that FTAs are by definition discriminatory because they lower/eliminate barriers on internal trade while retaining barriers to trade with nonmembers and are, therefore, trade diverting. Even if an FTA results in internal trade creation, these proponents believe that such gains are likely to be outweighed by their trade-diverting effects. In general, one would expect an FTA to result in some amount of both trade creation and trade diversion (Krueger 1997; Venables). If the trade diversion is sufficiently large relative to the trade creation, the agreement could conceivably end up being harmful to the member countries.

The North American Free Trade Agreement (NAFTA) is one of the most comprehensive agreements in history. Like many other FTAs, the creation of NAFTA has been a subject of bitter discussions and division among politicians and economists, focusing on the effect that NAFTA might have on trade and economic welfare (Fukao, Okuba, and Stern). When NAFTA was being negotiated in the early 1990s, for example, many countries voiced concern that their exports to the United States (and, to a lesser extent, to Canada and Mexico) would be displaced by NAFTA exports, even though in many products and industries those countries could be more competitive than NAFTA producers (Lederman, Maloney, and Serven). From the viewpoint of Mexico, this trade diversion is also important because it would entail a loss of fiscal revenues from replacing imports from third countries subject to tariffs with duty-free imports from the United States or Canada.

Despite the growing concern of the debate, NAFTA was expected to create new trade among the member countries. Through progressive elimination of tariff and nontariff barriers, bilateral trade flows among the United States, Canada, and Mexico were expected to increase. A number of reports have shown evidence of increased trade flows. The Congressional Budget Office (CBO), for example, analyzed that by 2001, NAFTA had increased U.S. exports to Mexico by 11.3% and had increased U.S. imports from Mexico by 7.7% (CBO). On the other hand, the report also pointed out that the agreement had almost no effect on the U.S. trade balance with Mexico and little effect on the change in U.S. gross domestic product (GDP).

Considerable concern is also expressed about the welfare implications of increased trade among the NAFTA member countries. Agriculture is one of the sectors in which there is considerable concern about the potential effects of free trade agreements on domestic producers and consumers (Miljkovic and Paul). Before NAFTA implementation, for example, the effect of NAFTA on Mexican agriculture received a lot of rather pessimistic attention (Levy and van Wijnbergen). Recently, it has also become the subject of political controversy as a consequence of the liberalization of certain sensitive products for Mexico, which was implemented in January 2003 (Lederman, Maloney, and Serven).

Quantitative economic analysis of the potential effects of NAFTA has been done. However, few studies addressed whether new trade in the agricultural sector has been created at all. The objective of this study is to estimate and evaluate the benefits of NAFTA, emphasizing trade creation and trade diversion in the U.S-Mexico agreement. This is particularly important because the liberalization of U.S-Mexico trade is in an advanced stage. However, many crucial trade disputes remain between the two countries, such as in the case of sugar and high-fructose corn syrup and, notably, concern for the effect that NAFTA might have on trade and economic welfare.

To evaluate the effects of NAFTA, U.S. import demand functions from both Mexico and rest of the world (ROW) are analyzed.1 Certain aspects make this study different from previous empirical work. First, it focuses on the agricultural products within the fourdigit level of the Harmonized Tariff Schedule (HTS). More importantly, this study looks at commodities that were subject to the nonzero pre-NAFTA tariff rates. Second, this study also aims to measure the time-varying dimension by allowing the effects of tariff reduction to differ during the NAFTA period. Third, this study seeks to assess the differential effects of pre-NAFTA tariffs on the U.S. imports from Mexico. This is important because initial tariff rates are predetermined and are therefore useful for identifying the effect of tariff changes on trade flows (Clausing).

A Brief Review of Welfare Analysis

Modern analysis of FTAs dates back to Viner, who drew the distinction between the tradecreating and trade-diverting effects of FTA formation. According to Viner, reducing tariff barriers through the formation of FTAs will not necessarily improve welfare. That is, whether a particular FTA (customs union) is a move in the right or in the wrong direction depends on the magnitude of trade creation and trade diversion (p. 44). This is in line with the trade argument in which the gains apply if all trade barriers are reduced. Because the purpose of an FTA is solely to reduce trade barriers among country members and retain them from nonmembers, it is basically partial and discriminatory; therefore, discrimination between sources of supply is not eliminated, it is just shifted (Venables). If a member of an FTA displaces higher cost domestic production of a good with imports of the good from another member of the FTA, there will be gains. Viner refers these gains as "trade creation." It is also possible, however, that a partner country production might displace lower cost imports from nonmember countries. This is called welfare-reducing "trade diversion." It is then noticeable that the benefit of an FTA as a whole derives from the portion of the new trade among the member countries (trade creation) and each particular portion of the new trade among the member countries that is a substitute for trade with nonmembers (trade diversion).

To illustrate the model, Figure 1 shows an analysis of a commodity between the United States and Mexico. S^sub US^ and D^sub US^ are, respectively, the domestic supply and demand curves for the United States. Initially, assume that the United States has a nondiscriminatory tariff t (as represented by the dotted arrow line) set on imports from both Mexico and ROW. The tariff raises the domestic supply prices to P^sup T^^sub M^ for Mexico and P^sup T^^sub ROW^ for ROW. Assuming that ROW is capable of supplying all the domestic demand for the United States, the entire quantity demanded, S^sub 1^D^sub 1^, is imported from P^sup T^^sub ROW^. The price paid by consumers is P^sup T^ROW and tariff revenue is given by the area C + E. The United States will not trade initially with Mexico because Mexican suppliers are not competitive.2

Figure 1. Trade Creation and Trade Diversion

Once the tariff is eliminated, as it should be under the NAFTA imposition, the Mexican duty-free price, P^sub M^, is lower than the tariffinclusive ROW price, P^sup T^^sub ROW^. The United States now purchases its imports from Mexico rather than from ROW at price P^sub M^. With the lower price, imports would rise to S^sub 2^D^sub 2^, creating new trade shown by the area B + D. One should also note that the nondistorted (i.e., free trade) price in ROW is conceivably less than the Mexican price. Trade is substituted from a more efficient supplier (ROW) to a less efficient supplier (Mexico) because imports from Mexico replace those from the ROW. The agreement is said to be trade diverting. The United States loses the tariff revenue C + E, with E used to pay for the higher production cost in Mexico. Domestic consumers gain areas A + B + C + D, and domestic producers lose area A.

As Viner pointed out, whether an FTA will bring net benefits to the member countries depends on the relative magnitudes of trade creation and trade diversion. In practice, area B + D of Figure 1 can be larger than area E, establishing the possibility that a wholly tradediverting FTA can lead to the improvement of welfare (Panagariya).

Empirical Studies of Free Trade Agreements

Much empirical work has been devoted toward evaluating trade and welfare effects of FTAs. These studies have taken two main forms: computer simulation studies of the full general equilibrium effects of FTA membership and econometric studies of changes in trade flows (Burfisher, Robinson, and Thierfelder; Venables). The ex ante studies with computable general equilibrium (CGE) models use various simulation methods to analyze a calibrated model economy for a particular base year. Virtually most of the studies that analyze the effects of an FTA on member as well as nonmember countries have used a CGE model and find that trade agreements have been welfare improving; that is, trade creation outweighed trade diversion (Burfisher, Robinson, and Thierfelder; Kehoe and Kehoe; Krueger, 2000). For the case of NAFTA, all the models agree that NAFTA would provide positive gains to member countries. Although Mexico would enjoy the biggest gains, the United States would experience a marginal increase in the economy, and Canada would expect only minimal effects.

CGE-based FTA studies are not without criticism, however. Kehoe, for example, argues that the CGE model greatly underestimates the increases in trade resulting from NAFTA. The CGE model is also considered to lack detailed, up-to-date policy coverage and product disaggregation (Beghin and Aksoy). In response to these criticisms, researchers have used econometric methods with historical time series (and cross-sectional) data to analyze the effects of FTAs on trade flows and welfare. This approach seeks to quantify the changes in trade flows attributable to membership in an FTA and thereby identify trade creation and trade diversion. The most common approach is the gravity model, which regresses trade flows among trading partners on their respective economic size (i.e., GDP) and geographic distance as proxy for transportation costs. Dummy variables are typically used to capture the effects of various preferential trading agreements on trade flows. Some examples are studies given by Gould; Krueger (1999); Zahniser et al.; and Lederman, Maloney, and Serven. In general, these studies agree that NAFTA was not a trade-diverting agreement.

Similar to the CGE model, there are problems with the gravity approach. Clausing provided three points of weakness with regard to the gravity equation model. First, the use of dummy variables is considered to be inadequate in capturing the effects of preferential trade liberalization. Second, the gravity model does not indicate the extent of trade creation and trade diversion; hence, it is difficult to assess the net effects of the agreements. Third, the data used in the analyses, in most cases, was highly aggregate, such that it is difficult to exploit variations in the extent of trade liberalization across goods or industries (Clausing, p. 680).

An extension of the gravity approach has been used to assess the effect of FTAs on trade and welfare. Unlike the gravity model that takes the model as an ad hoc representation, the current approach is developed on the basis of a better grounded economic theory such as demand theory. The work of Karemera and Koo; Clausing; and Fukao, Okuba, and Stern are some examples. Karemera and Koo analyzed the trade effects of removing tariff and nontariff barriers between the U.S.Canadian free trade agreements with the use of quarterly data from 1970 to 1987. They applied seemingly unrelated regression estimation technique to estimate the demand functions on the basis of the Standard International Trade Classification (United States) and Standard Industrial Classification (Canada). They conclude that U.S. imports of Canadian goods were more sensitive to domestic and bilateral import prices than were Canadian imports of U.S. goods. They also find that tariff and nontariff elimination would increase bilateral trade volume across all commodities traded, primarily trough trade creation and trade diversion.

Clausing was first to exploit tariff variation at the detailed commodity level with U.S. import data from 1989 to 1994. He finds that tariff liberalization was responsible for the growth in U.S. imports, with little evidence of trade creation. Fukao, Okuba, and Stern analyzed U.S. imports at the Harmonized System two-digit level and four-digit level for the period 1992-1998. They find that NAFTA tariff preferences had a significant effect on U.S. imports in 15 cases. Their results also show evidence of trade diversion, especially in U.S. imports of textiles and apparel products from Mexico.

Another similar approach that uses tariff data to examine trade effects is McDaniel and Agama, who estimated the effects of NAFTA on U.S. import demand for Mexican goods and Mexico's demand for U.S. exports. The results suggest that U.S. import demand for Mexican goods was responsive to tariff preferences, especially during the NAFTA years. Similarly, they find that Mexico's demand for U.S. exports was also responsive to the NAFTA preference.

Econometric Analysis

Empirical Specification

The import demand for the U.S. agricultural products from ROW is constructed to measure the trade diversion that might occur during the implementation of NAFTA. The independent variables consist of the tariff rates set by the United States against ROW (TR^sup ROW^^sub it^), tariff rates against Mexico (TR^sup MEX^^sub it^), DNAFTA, GDP (Y^sup US^^sub t^), import prices from ROW (PROW^sub it^), trade-weighted exchange index (RER^sup ROW^^sub t^), and quarterly dummy variables. We expect that an increase in T^sup ROW^^sub it^ would negatively affect the U.S. imports from ROW, and an increase in TR^sup MEX^^sub it^ would positively affect the U.S. imports from ROW. The central issue is overall on the coefficient of TR^sup MEX^^sub it^. If in fact trade diversion occurs, then TR^sup MEX^^sub it^ must have a positive sign, meaning that a decrease in tariff rates against Mexico would reduce U.S. imports from ROW. With respect to DNAFTA, this variable may have a positive or negative sign. However, we expect that DNAFTA will have a positive effect because this variable captures not only NAFTA per se but also other events not included in the model.

Estimation

Our empirical assessment of the specified equations is based on the panel data analysis. Within this framework, we are able to explore possible explanations for the heterogeneity in commodity groups or commodity characteristics. Potential reasons for the heterogeneity include different responses of import demand because of expected reductions in tariff rates (i.e., tariff schedules under NAFTA agreements and unobservable individual specific characteristics).

Furthermore, v^sub i^ and e^sub t^ can either be random or nonrandom, and ε^sub it^ is the classical error term with zero mean and homoscedastic covariance matrix. The nature of the error structures leads to different estimation procedures depending on the specification. Because our tests (Hausman's tests for random effects) show that the fixed effects model is preferred to the random effects model, our reports are only based on the fixed effects specification. Because of the presence of autocorrelation, the specified models are estimated under firstorder autocorrelation. The lag lengths of tariff rates, prices, and exchange rates were based on Hall's general to specific procedure.

Data

According to the U.S. HTS, all of the products found in Chapters 1-24, with the exception of fishery products in Chapters 3 and 16, are considered agricultural products (USDA). Certain other products outside of Chapters 1-24 are also considered agricultural products, particularly essential oils (Chapter 33), raw rubber (Chapter 40), raw animal hides and skins (Chapter 41), and wool and cotton (Chapters 51 and 52). We adopt this classification for the definition of agricultural sector. The data consist of the four-digit HTS system and range from 1989-2005 in a quarterly basis. Because most agricultural products traded between the United States and Mexico were already subject to free trade before NAFTA was in effect (i.e., zero tariff rates), we did not use all the four-digit level classified under agricultural products. Instead, we selected commodity groups in which they were subject to nonzero tariff rates before the implementation of NAFTA. Our selection of the commodity groups was also based on the consistency of the data during the selected period. The main reason for the use of the nonzero tariff rates is to obtain variation in the tariff rates to help identify the effects of tariff liberalization.

Table 1. Characteristics of Sample Data

The value of U.S. imports from Mexico is used to represent the quantity of import, and the applied U.S. tariff is based on detailed data on import duties collected. The tariff rates for each commodity classification are calculated as the ratio of calculated duties to customs value. The drawback of this approach is that tariff rates can only be observed when there is trade (Romalis). When there is no trade, we estimate the tariff rates by taking the average of two surrounding available tariffs. The customs value and calculated duties are extracted from the United State International Trade Commission database. Data on U.S. GDP are from the Bureau of Economic Analysis database and are converted into real values with the use of consumer price indices published by the Bureau of Labor Statistics. The unit values (border prices) for the fourdigit HTS level were constructed from a USDA database. The Mexico-U.S. real exchange rates are from the USDA, and the trade-weighted exchange index is obtained from the Federal Reserve Bank.

Summary statistics of the sample data are presented in Table 1. As shown, the range of import values was quite wide, from $20 million to $388 million (for commodities imported from Mexico) and from $41 million to $531 million (for commodities imported from ROW). The average tariff rates from 1989 to 2005 for commodities imported from Mexico and ROW were approximately the same, with values of 3.9% and 3.8%, respectively. The average pre-NAFTA tariff for Mexican products entering the United States (1989-1993) was 8.2%. In the first 6 years of NAFTA, this value was about 3.3%, and in the last 6 years 0.8%.

Results

U.S. Imports from Mexico

Table 2 presents the econometric results for U.S. import demand from Mexico. Specification 1 shows the effects of tariff rates on U.S. imports by controlling NAFTA (DNAFTA) and allowing the effects to differ during the NAFTA period. Specification 2 reports the effect of tariff rates without controlling NAFTA. In general, most of the estimated coefficients are significant and posses the expected signs, with the exception of exchange rates. The F-statistics for testing the joint significance of the individual (group commodity) effects strongly suggest the presence of an individual heterogeneity in the data.

Table 2. Effects of Tariff Rates and NAFTA on U.S. Imports from Mexico

Before turning to the detailed discussion of the tariff and its effect on the U.S. imports from Mexico, we will give a quick evaluation of the income, exchange rates, and price variables. As shown in Table 2, the coefficients of income are found to be significant in each specification, with values of 1.4 and 1.6 for Specification 1 and Specification 2, respectively. These suggest that U.S. agricultural imports from Mexico are income elastic. We also found that border price is significant in both specifications, suggesting that price is an important factor of U.S. agricultural imports; however, the magnitude of its parameter shows that U.S. agricultural imports from Mexico are price inelastic. With respect to exchange rates, we found negative signs, contrary to expectation. However, they are not significant. The insignificance of exchange parameters can partly be explained because U.S.-Mexico trade has been established such that U.S. importers did not consider exchange rate as the determining factor.

Specification 1 of Table 2 shows the effects of tariff rate reduction on U.S. imports from Mexico. As shown, the coefficient of tariff rates is negative and significant. Controlling NAFTA and other variables, the estimated parameter shows that a one percentage point decrease in the tariff rates against Mexico would increase U.S. imports of agricultural products by 2.4%. Note that this estimate indicates the effect of tariff rates during the whole period of the study. When considering the multiplicative effects, the sum of α^sub 1^ and α3 indicates the effects of tariff rate reductions for the period 1994-1999, and the sum of α^sub 1^ and α^sub 4^ shows the effects of tariff reductions for the period 2000-2005. The effects of tariff rates during the 12 years of NAFTA is represented by the sum of α^sub 1^, α^sub 3^, and α^sub 4^. As can be seen, the coefficient of α^sub 3^ is significant and negative as expected. This means that the effects of tariff reductions during the first 6 years has been a 5.31% increase of every one percentage point decrease in tariff rates against Mexico.

The coefficient of α4 is negative and not significant. The relatively low parameter estimate is as expected because of the declining effects of tariff reductions. The insignificant parameter might be justified by the fact that some of the commodities have been subject to trade liberalization (zero tariff rates) or at least have experienced a period of low tariff level. Hence, any reduction of tariff levels in this period might not significantly affect U.S. imports from Mexico, as indeed was shown in this study. Regardless of significance level, the effects of tariff reductions on U.S. imports from Mexico have been an increase of 2.62% during the last 6 years of NAFTA and 5.5% during the NAFTA period. By estimating the model without NAFTA dummy variable and its associated multiplicative effects, we found a slightly lower magnitude of the effects of tariff reductions. As shown in Specification 2, the coefficient of α^sub 1^ is negative and significant at -3.69%. This indicates that a one percentage point reduction in tariff rates has increased U.S. imports from Mexico by 3.69%.

Other important results are also given in Table 2. Researchers argue that the effects of NAFTA could be subject to the tariff levels before NAFTA implementation. Our specification enables us to track such effects in that it can show the different effects of tariff levels before NAFTA. The coefficients of α^sub 5^ and α^sub 6^ in the two specifications clearly show that the lower the pre-NAFTA tariff rates, the lower the effect on the increase of U.S. imports from Mexico as tariff rates are reduced. From Specification 1, U.S. agricultural imports from Mexico during the NAFTA period were approximately 26% higher than pre-NAFTA for group commodities with pre-NAFTA tariff rates less than 10%. On the other hand, this figure was 37% for commodities that fall into the category of pre-NAFTA tariff rates greater than 15%.4

Finally, the effect of DNAFTA on U.S. agricultural imports from Mexico was significant. U.S. imports from Mexico during NAFTA were 54% higher than the entire period of analysis. However, one should note that these coefficients do not explain the effect of NAFTA per se because this variable captures not only NAFTA but also other events that are not accounted for in the model.

U.S. Imports from ROW

One way to investigate the presence of trade diversion can be done by regressing U.S. imports from ROW on tariff rates set by the United States against Mexico. A positive sign of this variable indicates that trade diversion exists. Table 3 reports the econometric results for U.S. import demand from ROW. Specification 1 shows estimation results for U.S. import demand from ROW with tariff rates against Mexico that is intended to determine whether trade diversion occurred while controlling for NAFTA. It also provides estimates of NAFTA dummy variables to test whether U.S. imports from ROW increased during the NAFTA period. These two variables can be used jointly to justify whether trade diversion indeed occurred. Similarly, Specification 2 gives estimates of the effects of both tariff rates against ROW and tariff rates against Mexico on U.S. imports from ROW without controlling for NAFTA.

Except for tariff rates against Mexico, exchange rates, and NAFTA dummy variables, all coefficients are significant at reasonable levels. They also possess expected signs. The coefficient of determination is 0.77 and the F-statistics for fixed effects tests show the presence of heterogeneity in commodity characteristics. The seasonal dummy variables show significant differences in U.S. imports from ROW from quarter to quarter. In all cases, the U.S. imports from ROW in the fourth quarter are higher than the first three quarters.

The U.S. agricultural import demand from ROW seems to be income elastic with a magnitude of 2.4. It is surprising that the NAFTA dummy variable is not significant at any reasonable level, even though it has the expected sign. If it were significant, we could have expected that U.S. agricultural imports from ROW in the first 12 years of NAFTA were 9.1% higher than they would have been without NAFTA. Strong income effects in the United States likely negate this result. Similarly, we also found that exchange rates are not significant. The use of the tradeweighted exchange rate index, which was calculated on the basis of U.S. major trading partners, might contribute to the insignificance parameter.

Table 3. Effects of Tariff Rates and NAFTA on U.S. Imports from ROW

The estimated results for the coefficients of tariff rates against ROW suggest that a one percentage point reduction in tariff rates against ROW is associated with a 1.5% increase in U.S. agricultural imports from ROW. This is clearly less than one third of the effects of tariff reductions against Mexico on U.S. agricultural imports from Mexico (Specification 1 of Table 2). This evidence suggests that the United States gives more preference to Mexican agricultural products than ROW agricultural products. This is not surprising given that the United States and Mexico are tied to the NAFTA agreements, along with other advantages such as geographical proximity. The parameter estimates of tariff rates against Mexico are positive. As previously stated, a positive sign of tariff rates indicates the presence of trade diversion. However, because these coefficients are not significant, we would argue that there is no significant evidence that NAFTA has caused trade diversion in the agricultural sector, particularly as a result of the United States and Mexico trade agreements.

Trade Creation and Trade Diversion

As previously discussed, the net benefit of an FTA as a whole derives from the portion of the new trade among the member countries (trade creation) and each particular portion of the new trade among the member countries which is a substitute for trade with nonmembers (trade diversion). The regression results show that tariff reductions during the NAFTA period had significant effects on U.S. imports from Mexico, whereas the coefficient of tariff rates against Mexico regressed on U.S. imports from ROW was not significant, even though it has a positive sign. Before concluding that NAFTA has affected trade flows, it would be informative to highlight the changes in U.S. imports from NAFTA countries and ROW.

Figure 2. Shares of Real U.S. Imports from Country of OriginFigure 3. U.S. Real Imports from Country of Origin ($ Billion)

Figure 2 shows the shares of U.S. imports of the selected four-digit level HTS from Canada, Mexico, and ROW from 1989 to 2005. As the figure shows, U.S. imports from Mexico increased sharply immediately after NAFTA's inception, before experiencing a slight decrease in 1996 and 1997. The share of U.S. imports from Canada increased gradually, with a slight decrease in 2003. The share of U.S. imports from Mexico increased from 21% to 29% between 1989 and 2005. During the same period, the share of U.S. imports from Canada increased from 11.6% to 21%. In general, we can conclude that the share of U.S. imports from both Mexico and Canada trended upward during the NAFTA period, suggesting that NAFTA has been trade creating. Meanwhile, the share of U.S. imports from ROW has degraded continually since 1989. Notably, U.S. imports from ROW declined from 67.1% in 1989 to 50% in 2005. One should note that these figures were based on the selected commodity groups (i.e., product groups that were subject to nonzero pre-NAFTA tariff rates).

The increase in the share of U.S. imports from Mexico and Canada accompanied by a decrease in the share of U.S. imports from ROW has raised the question whether the United States shifted away its imports from ROW to the NAFTA members. It is very important to respond to such concern because if in fact the United States did shift its imports at the expense of ROW, there was clearly trade diversion. Figure 3 might clarify the issue. As depicted in this figure, U.S. agricultural imports from Mexico, Canada, and ROW increased substantially from year to year with a slight decline in particular years. In general, the trend of U.S. imports from NAFTA members and ROW suggest that the United States did not shift its imports away from ROW. The decline in the share of U.S. imports from ROW was particularly that U.S. imports from NAFTA members grew faster than those from ROW.

Historical data show evidence that U.S. agricultural imports from Mexico have increased since the inception of NAFTA. But to what degree the increase is attributable to NAFTA is difficult to examine. Other important factors have also been responsible for the increase. Krueger (1999, 2000), for example, noted that the economic growth and the change in exchange rates were responsible for the growth in trade flows in the NAFTA region. However, although such other factors are of full consideration, we argue that NAFTA has been trade creating. Our conclusion is also supported by regression results. As shown in Table 2 and the discussions that follow, tariff rate reductions have had a positive effect on U.S. imports from Mexico. This effect is even higher during the NAFTA period compared with the average of the entire period. The other regression results in Table 3 clearly indicate that reductions in tariff rates against Mexico did not significantly lower U.S. imports from ROW, suggesting that trade diversion did not occur to a significant degree.

If, however, one still believed that trade diversion existed because the sign of TR^sup MEX^^sub il^ in Equation (7) was positive, we still argue that NAFTA has been net trade creating because the absolute magnitude of the coefficient TR^sup MEX^^sub it^ in Equation (7) is far below the coefficient that measures the effects of TR^sup MEX^^sub il^ on U.S. imports from Mexico.

Conclusions

In this paper, we estimated U.S. agricultural import demand functions from both Mexico and ROW and examined the trade creation and trade diversion that might have occurred in the U.S.-Mexico agreements under NAFTA with the use of panel data of 35 selected fourdigit levels of HTS from 1989 to 2005. The 35 commodity groups were selected on the basis of tariff rates that were not subject to zero tariff rates before NAFTA. The use of more disaggregated data and the nonzero preNAFTA tariff rates has enabled us to examine the variations of tariff rates and also reduce bias that might have occurred if we included the zero pre-NAFTA tariff rates in the analysis.

The results suggest that U.S. agricultural imports from Mexico have been responsive to tariff rate reductions applied to Mexican products. A one percentage point reduction in tariff would increase U.S. agricultural imports from Mexico by 2.4% in the entire period. Results also show that during NAFTA, a one percentage point decrease in tariff rates would increase U.S. agricultural imports from Mexico by 5.31% in the first 6 years of NAFTA and by 2.62% in the last 6 years of NAFTA. The overall effect would be a 5.5% increase in U.S. imports from Mexico for a one percentage point decrease in tariff rates. The increase in U.S. imports from Mexico has also been attributable to the pre-NAFTA tariff rates. Higher pre-NAFTA tariff levels would result in a higher percent increase in U.S. agricultural imports from Mexico as tariff rates are reduced. Similarly, U.S. imports from ROW have also been significantly affected by tariff rates applied to ROW. The effect, however, is lower compared with Mexico. We also found that U.S. agricultural imports from Mexico during NAFTA were approximately 54% higher than the entire period. Conversely, U.S. imports from ROW were not significantly higher because of NAFTA.

Overall, we conclude that there is significant evidence that the U.S.-Mexico trade agreement under NAFTA has been trade creating rather than trade diverting. However, one should note that these conclusions are based only on evaluation of U.S. agricultural imports from Mexico and from selected agricultural products. This finding is especially important given arguments by Caribbean and other countries that U.S.-Mexico trade has diverted commercial sales. The extent to which these results might be applicable to other agreements, such as the Central AmericaDominican Republic Free Trade Agreement, is limited because each case must be verified empirically.

Footnote

1 Initially, we planned to estimate import demand for Mexico from the United States. However, because of data availability, especially the tariff rates for agricultural products set by the Mexican Government for U.S. agricultural products and the rest of the world, this analysis was not done. If such data were available, we would have been able to assess bilateral trade in a more complete setting. Because of such constraints, we focus on U.S. imports of selected agricultural products.

2 For trade to exist, the domestic price, P^sub US^ (not shown in the graph), is constructed such that P^sub US^ > P^sub MEX^ > P^sub ROW^. It is also assumed that ROW and Mexico do not trade with each other. This is true if Mexico applies a per unit tariff higher than P^sub MEX^ P^sub ROW^.

3 It is possible to define yearly multiplicative effects to see the year-to-year effect of tariff reductions during the NAFTA period. However, this might cause collinearity problems. Besides, observing NAFTA year-to-year might not capture its effects, particularly in the early years of its implementation. It is then reasonable to have an even split of the time frame (6year period) because we would be able to see whether the effects of trade agreements are front loaded or back loaded.

4 Our interpretation of the coefficient of dummy variables is based on suggestions by Kennedy. Suppose that c is the estimate of a dummy variable coefficient c; its effect on the dependent variable (which is a log value) is given by g* = exp[c- ½v(c)] 1, where v(c) is an estimate of the variance c. See also Halvorsen and Palmquist for interpretation of the coefficient of dummy variables in semilogarithmic equations.

References

[Received March 2006; Accepted August 2006.]

References

Baltagi, B.H. Econometric Analysis of Panel Data, 2nd ed. New York: John Wiley & Sons, 2001.

Beghin, J.C., and A. Aksoy. "Agricultural Trade and the Doha Round: Lessons from Commodity Studies." Briefing Paper 03-BP 42, Center for Agricultural and Rural Development, Iowa State University, Ames, 2003.

Bhagwati, J., and A.O. Krueger. The Dangerous Drift to Preferential Trade Agreements. Washington, DC: AEI Press, 1995.

Burfisher, M., S. Robinson, and K. Thierfelder. "The Impact of NAFTA on the United States." Journal of Economic Perspectives 15(2001): 12544.

Clausing, K. "Trade Creation and Trade Diversion in the Canada-United States Free Trade Agreement." Canadian Journal of Economics 34(2001):677-96.

CBO (Congressional Budget Office). "The Effects of NAFTA on US-Mexican Trade and GDP." A CBO paper, Washington, DC, May 2003.

Fukao, K., T. Okuba, and R.M. Stern. "An Econometric Analysis of Trade Diversion under NAFTA." North American Journal of Economics and Finance 14(2003):3-24.

Gould, D.M. "Has NAFTA Changed North American Trade?" Economic Review of the Federal Reserve Bank of Dallas First Quarter (1998):12-23.

Halvorsen, R., and R. Palmquist. "The Interpretation of Dummy Variables in Semi Logarithmic Equations." The American Economic Review 70(1980):474-75.

Hsiao, C. Analysis of Panel Data. New York: Cambridge University Press, 2003.

Karemera, D., and W.W. Koo. "Trade Creation and Diversion Effects of the U.S.-Canadian Free Trade Agreement." Contemporary Economics Policy 12(1994): 12-23.

Kehoe, P.J., and T.J. Kehoe. "Capturing NAFTA's Impact with Applied General Equilibrium Models." Federal Reserve Bank of Minneapolis Quarterly Review 18,1(Spring 1994)

Kehoe, T.J. "An Evaluation of the Performance of Applied General Equilibrium Models of the Impact of NAFTA." Research Department Staff Report 320, Federal Reserve Bank of Minneapolis, August 2003.

Kennedy, P.E. "Estimation with Correctly Interpreted Dummy Variables in Semi Logarithmic Equations." The American Economic Review 71(1981):801.

Kruger, A.O. "Free Trade Agreements versus Customs Unions." Journal of Development Economics 54(1997): 169-87.

_____. "Trade Creation and Trade Diversion under NAFTA." Working Paper 7429, National Bureau of Economic Research NBER Working Paper Series, Cambridge, MA, December 1999.

_____. "NAFTA's Effects: A Preliminary Assessment." World Economy 33:6(2000):761-75.

Lederman, D., W.F. Maloney, and L. Serven. "Lessons from NAFTA for Latin America and the Caribbean Countries: A Summary of Research Findings." Washington, DC: The World Bank, 2003.

Levy, S., and S. van Wijnbergen. "Labor Markets, Migration, and Welfare: Agriculture in the North American Free Trade Agreement." Journal of Development Economics 43(1994):263-78.

McDaniel, C., and L. Agama. "The NAFTA Preference and US-Mexico Trade: AggregateLevel Analysis." World Economy 6(2003):939-55.

Miljkovic, D., and R. Paul. "Agricultural Trade in North America: Trade Creation, Regionalism and Regionalization." The Australian Journal of Agricultural and Resource Economics 43,3 (2003):349-66.

Panagariya, A. "Preferential Trade Liberalization: The Traditional Theory and New Developments." Journal of Economic Literature 38(2000):287-331.

Romalis, J. "NAFTA's and CUSFTA's Impact on International Trade." Review of Economics and Statistics, forthcoming, 2007.

USDA (United States Department of Agriculture): "The Harmonized Tariff Schedule of The United States", available at: http://www.fas. usda.gov/itp/us-tariffsch.asp (last accessed: February 8, 2007).

Venables, A.J. International Trade: Regional Economic Integration. Manuscript prepared for the "International Encyclopedia of Social and Behavioral Sciences," July 2000. Internet site: http://econ.lse.ac.uk/~ajv/regenc2.pdf (Accessed February 2006).

Viner, J. The Customs Union Issue. New York: Carnegie Endowment for International Peace, 1950.

Wooldridge, J.M. Econometric Analysis of Cross Section and Panel Data. Cambridge, MA: MIT Press, 2002.

Zahniser, S.S., D. Pick, G. Pompelli, and M.J. Gehlhar. "Regionalism in the Western Hemisphere and Its Impact on U.S. Agricultural Exports: A Gravity-Model Analysis." American Journal of Agricultural Economics 84(2002): 791-97.