Abstract
This paper assesses the effects of reducing tariffs under the Doha Round on market access for developing countries. It shows that for many developing countries actual preferential access is less generous than it appears because of low product coverage or complex rules of origin. Thus, lowering tariffs under the multilateral system is likely to lead to a net increase in market access for many developing countries, with gains in market access offsetting losses from preference erosion. Furthermore, comparing various tariff-cutting proposals, the research shows that the largest gains in market access are generated by higher tariff cuts in agriculture.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Other issues, such as loss in tariff revenue, are not addressed in this paper.
A bound tariff is the maximum tariff that a country can set, as agreed under WTO negotiations. The MFN tariff is the tariff rate applied to all WTO member countries that do not receive special preferences.
Simulations with alternative elasticity of substitution assumptions are also presented to show robustness of the results.
This infinite supply elasticity assumption differs from that made in Subramanian (2003) and Alexandraki and Lankes (2004), which assume a supply elasticity of 1 and do not consider terms of trade effects. Their assumptions deliberately bias the results in favor of overstating losses from preference erosion in order to minimize the risk of overlooking individual countries that might face losses. In addition, they also assume 100 percent utilization of preferences.
Note that these numbers are likely to understate the potential gains in market access and the number of gaining countries because only the “intensive margin” of trade is modeled; that is, if there were no exports of a particular good from a particular country to the United States or EU before the tariff cut, there will also be no exports to those countries following the cut. This is particularly relevant for high-tariff goods where tariff cuts could lead to changes in the “extensive margin” that are not captured in these simulations.
Romalis (forthcoming) finds high supply elasticities, but these estimates were not confined to developing countries. The effects of other proposals in the Doha Round such as cuts in export subsidies on agricultural products are not considered here. The focus is on changes in market access resulting from tariff cuts.
Papers that assume finite supply elasticities also find small losses from preference erosion. In simulations following a 40 percent cut in MFN rates, Subramanian (2003) finds that losses from preference erosion for LDCs as a whole are very small and likely to be less than 2 percent of exports, and only two countries face losses greater than 10 percent of exports. Alexandraki and Lankes (2004) extend this analysis to middle-income developing countries and also find the overall impact to be small, between 0.5 and 1.2 percent of total exports, but it could be much higher for a subset of countries that are overwhelmingly dependent on a few export products, namely sugar, bananas, and to a lesser extent, textiles. Note, that there could also be further gains owing to productivity improvements caused by lower tariffs on inputs. See Amiti and Konings (forthcoming).
The countries marked with an asterisk do not receive preferences from the United States. All EU members are labeled as “developed.”
Product coverage is much lower when mineral products are excluded (HS Chapters 25–27, mostly oil), collapsing to 3.9 percent (1.5 billion out of 3.9 billion of dutiable imports).
Previous studies have also identified limitations of GSP. For example, not all developing countries are included (Baldwin and Murray, 1977). Programs typically exclude products for which developing countries have the greatest comparative advantage (DeVault, 1996). Export eligibility ceilings are often binding (MacPhee and Rosenbaum, 1989). The programs impose strict rules of origin requirements (UNCTAD, 2001) and do not remove nontariff barriers. Up to 42 countries have temporarily dropped or have been permanently “graduated” by the United States at some time since 1976 (Ozden and Reinhardt, 2003). The United States has allowed the GSP to lapse on occasion, including one period in excess of a year, increasing uncertainty for exporters. Mattoo, Roy, and Subramanian (2002) highlight that the stringent rule of origin requiring exporters to source certain inputs from within Africa or the United States severely restricts the potential benefits from the preferences granted under AGOA.
Brenton and Ikezuki (2004) also point out that products that are excluded from AGOA preferences are high-duty products, and the United States is not a major export destination for many AGOA country exports. These low utilization rates are likely due to high compliance costs, such as paperwork and red tape. Francois, Hoekman, and Manchin (2005) find a threshold preference margin of 4 percent below which preference margins are irrelevant because of these costs.
Note that over 99 percent of U.S. and EU tariffs are bound (WTO, 2002).
EU preferential tariffs for developing countries are reduced using the formulas in European Commission (2003).
Approximately 10 percent of tariff lines include specific tariffs, which have been converted to ad valorem equivalent tariffs by the United Nations Conference on Trade and Development (UNCTAD) in the case of EU tariffs and by dividing actual duty paid by the value of imports in the case of the United States.
The most recent available data for the United States is 2004 and for the EU it is 2003.
Thus these results understate the gains from the Doha Round because the simulations do not take account of tariff cuts by other countries. Yang (2005) points out that African countries can increase their gains by also seeking greater market access in developing countries as well as making their own liberalization commitments.
Note that total duties collected were unavailable for the EU. The estimated tariff paid on a product exported by a particular country is a weighted average of the EU MFN tariff for that product and the lowest tariff that product may be eligible for under various EU preference arrangements. The weight on the lowest tariff is the preference utilization rate for exports of that product fromthat country. Detailed preference utilization rates were obtained fromthe EU. Estimated average tariffs for a group of products and/or exporting countries are trade-weighted averages of the estimated tariffs for each product and exporting country.
For each region, comparative advantage in each good is identified using the Balassa index of revealed comparative advantage, defined as
where x
ij
is the industry i exports in region j, x
j
the total exports by region j, x
i
the total industry i exports in the world, and X is the total exports in the world. A number greater than 1 indicates revealed comparative advantage in that industry. Note that a country's comparative advantage is endogenous, and these are presented only for the purposes of illustrating why developing countries might be receiving inferior market access.See Dean and Wainio (2005) for detailed measures of size, utilization, and value of U.S. nonreciprocal trade preferences.
It is impossible to know exactly what the tariff cut will be under the next Doha Round. This number is based on cuts in previous rounds and pre–Doha Round informal discussions.
Jean, Laborde, and Martin (2005) also base their numbers on the Harbinson proposal, with some variations, arguing that although the proposal was not adopted its transition points are likely to reflect a great deal of deliberation.
These results are consistent with Francois, Hoekman, and Manchin (2005), which shows the potential magnitude of preference erosion is reduced owing to the high compliance costs of obtaining preferences. In their policy experiment, all OECD members abolish all trade distorting policies.
U.S. tariffs on petroleum are 5.25 cents per barrel for light crude oil, 10.5 cents per barrel for heavier grades of crude oil, and 52.5 cents per barrel for more refined products such as gasoline. Preferences (zero tariffs) are given to most developing countries (but not to most OPEC (Organization of Petroleum Exporting Countries) members) and to partners of free trade agreements. So the preference is small, but because oil is by far the biggest export from African LDCs to the United States, it plays a large part in the simulation results. However, it is unlikely that African LDCs would suffer falls in aggregate petroleum exports in the short term. What is more likely to happen is that U.S. tariff reductions for other suppliers cause a redirection of petroleum exports because the direction of commodity trade tends to minimize transport costs plus taxes.
where x
ij
is the industry i exports in region j, x
j
the total exports by region j, x
i
the total industry i exports in the world, and X is the total exports in the world. A number greater than 1 indicates revealed comparative advantage in that industry. Note that a country's comparative advantage is endogenous, and these are presented only for the purposes of illustrating why developing countries might be receiving inferior market access.References
Alexandraki, Katerina, and Hans Peter Lankes, 2004, “The Impact of Preference Erosion on Middle-Income Developing Countries,” IMF Working Paper 04/169 (Washington, International Monetary Fund).
Amiti, Mary, and Jozef Konings, forthcoming, “Trade Liberalization, Intermediate Inputs, and Productivity: Evidence from Indonesia,” American Economic Review.
Baldwin, R.E., and T. Murray, 1977, “MFN Tariff Reductions and Developing Country Trade Benefits Under the GSP,” Economic Journal, Vol. 87 (March), pp. 30–46.
Brenton, Paul, and Takako Ikezuki, 2004, “The Initial and Potential Impact of Preferential Access to the U.S. Market under the African Growth and Opportunity Act,” Policy Research Working Paper No. 3262 (Washington, World Bank).
Clausing, K.A., 2001, “Trade Creation and Trade Diversion in the Canada-United States Free Trade Agreement,” Canadian Journal of Economics, Vol. 34 (August), pp. 677–696.
Dean, Judith M., and John Wainio, 2005, “Quantifying the Value of U.S. Tariff Preferences for Developing Countries,” Policy Research Working Paper No. 3977 (Washington, World Bank).
DeVault, James, 1996, “Competitive Need Limits and the U.S. Generalized System of Preference,” Contemporary Economic Policy, Vol. 14 (October), pp. 58–66.
European Commission, 2003, “User's Guide to the European Union's Scheme of Generalised Tariff Preferences” (Brussels), Available via the Internet: http://europa.eu.int/comm/trade/issues/global/gsp/gspguide.htm.
Evans, Carolyn L., and James Harrigan, 2004, “Tight Clothing: How the MFA Affects Asian Apparel Exports,” NBER Working Paper No. 10250 (Cambridge, Massachusetts, National Bureau of Economic Research).
Francois, Joseph, Bernard Hoekman, and Miriam Manchin, 2005, “Preference Erosion and Multilateral Trade Liberalization,” CEPR Discussion Paper No. 5153 (London, Centre for Economic Policy Research).
Hummels, David, 2001, “Toward a Geography of Trade Costs,” (unpublished; West Lafayette, Indiana, Krannert School of Management, Purdue University). Available via the Internet: http://www.mgmt.purdue.edu/faculty/hummelsd/research/toward/TGTC.pdf.
Jean, Sébastien, David Laborde, and Will Martin, 2005, “Consequences of Alternative Formulas for Agricultural Tariff Cuts,” Chapter 4 in Agricultural Trade Reform and the Doha Development Agenda, ed. by Kym Anderson and Will Martin (Washington, World Bank).
Lai, Huiwen, and Daniel Trefler, 2002, “The Gains from Trade with Monopolistic Competition: Specification, Estimation, and Mis-Specification,” NBER Working Paper No. 9169 (Cambridge, Massachusetts, National Bureau of Economic Research).
MacPhee, Craig, and David I. Rosenbaum, 1989, “The Asymmetric Effects of Reversible Tariff Changes Under the United States GSP,” Southern Economic Journal, Vol. 56 (July), pp. 105–125.
Mattoo, Aaditya, Devesh Roy, and Arvind Subramanian, 2002, “The Africa Growth and Opportunity Act and Its Rules of Origin: Generosity Undermined?” IMF Working Paper 02/158 (Washington, International Monetary Fund).
Ozden, Çaglar, and Eric Reinhardt, 2003, “The Perversity of Preferences: The Generalized System of Preferences and Developing Country Trade Policies, 1976–2000,” Policy Research Working Paper No. 2955 (Washington, World Bank).
Romalis, John, (forthcoming), “NAFTA's and CUSFTA's Impact on International Trade,” Review of Economics and Statistics.
Subramanian, Arvind, 2003, “Financing of Losses from Preference Erosion,” paper prepared by IMF staff for the World Trade Organization, WT/TF/COH/14/14, Geneva.
United Nations Conference on Trade and Development (UNCTAD), 2001, Improving Market Access for Least Developed Counties, UNCTAD/DITC/TNCD/4. Available via the Internet: http://www.unctad.org/en/docs/poditctncd4.en.pdf.
United Nations Conference on Trade and Development (UNCTAD), 2003, Trade Preferences for LDCs: An Early Assessment of Benefits and Possible Improvements, UNCTAD/ITCD/TSB/2003/8 (New York and Geneva). Available via the Internet: http://www.unctad.org/en/docs/itcdtsb20038_en.pdf.
World Trade Organization, 2002, “Market Access: Unfinished Business—Post-Uruguay Round Inventory and Issues,” Special Studies No. 6 (Geneva). Available via the Internet: http://www.wto.org/english/res_e/booksp_e/special_study_6_e.pdf.
World Trade Organization, 2003, “Negotiations on Agriculture: First Draft of Modalities for the Further Commitments,” TN/AG/W/1Rev.1, March 18 (Geneva). Available via the Internet: http://www.wto.org/English/tratop_e/agric_e/negoti_mod2stdraft_e.htm.
Yang, Yongzheng, 2005, “Africa in the Doha Round: Dealing with Preference Erosion and Beyond,” IMF Policy Discussion Paper No. 05/8 (Washington, International Monetary Fund).
Additional information
*This paper was written when both authors were at the IMF Research Department. Mary Amiti is a senior economist at the Federal Reserve Bank of New York, and John Romalis is an associate professor at the University of Chicago. The authors thank Kalpana Khochar, Raghu Rajan, Arvind Subramanian, Shang-Jin Wei, and an anonymous referee for helpful comments. The views expressed in this paper are those of the authors and do not necessarily reflect those of the Federal Reserve Bank of New York or the Federal Reserve System.
Appendices
Appendix I
Technical Information
The detailed steps involved in calculating the change in market access and average preference margins are as follows.
Change in Market Access
Step 1
-
Calculate total U.S. imports for each product i in the base period 0 (year 2003).
Denote total imports in the base period M0i=Σ j M0ijp, where M ijp is U.S. imports of product i from country j that enters under tariff program p. This calculation is performed at the tariff-line level (10-digit level).
Step 2
-
Estimate total U.S. consumption for each product i.
Denote total consumption in the base period C0i=M0i/m:c i , where m:c i is the estimated ratio of imports to consumption calculated from the OECD's STAN database of domestic production, imports, and exports. The STAN database includes data for approximately 30 primary and secondary industries and is concorded to each tariff line.
Step 3
-
Calculate the new tariff rates t 1 ijp using existing tariff rates t 0 ijp as the base rates.
The new tariff rates will include a 40 percent tariff cut as the benchmark. In the second set of simulations 3 percent of the highest tariff rates will be excluded, and in the third set of simulations a tiered formula will be applied to agriculture with no other exclusions.
Step 4
-
Estimate new U.S. imports of each product i from each country j under each import program p.
The utility function is assumed to be Cobb-Douglas, which implies an elasticity of substitution of 1 between different goods at the HS 10-digit level. Hence, a fixed proportion of income is spent on each good.
Within these 10-digit categories, countries produce different varieties. U.S. consumers allocate their demands across products i. The import quantity demanded for country j goods under program p is given by maximizing the utility function subject to the budget constraint
where p0ijp is the free-on-board price, t0ijp the tariff rate, P the price index of all substitute varieties, and C0i is the expenditure on product i in period 0. Multiplying both sides by p gives the value of imports in period 0, M0ijp. Analogously, the total quantity of imports demanded from each country can be written as follows:
after substituting in for the price index and incorporating price changes from period 0 to period 1 that arise from changes in tariffs. Note that C0i−M0i is expenditure on domestically produced goods. The elasticity of substitution between different “varieties,” σ, is assumed to be 6. A “variety” is defined as the interaction of country j product and import program p.
Step 5
-
Calculate the change in “market access.”
The change in market access is defined as the change in U.S. demand for imports from each country as ΔMA j =100 (Σ ip M1ijp/Σ ip M0ijp−1). It is assumed that the export elasticity is infinite, thus the exporting country does not change its export prices exclusive of tariffs.
Step 6
-
Repeat the process for EU imports, with some modifications.
These modifications were necessary because the EU data on preference utilization, though detailed, is not as comprehensive as the U.S. data.
-
1)
Information on total imports in the base period, M0ij, for the EU is available, but not the imports under different preference programs, M0ijp. Detailed EU preference utilization data were obtained from the EU, indicating by eight-digit product and by exporting country the value of imports that were covered by a tariff preference and the value that actually entered under a preference. The exact preference scheme was not provided, only whether the applicable tariff under that preference was zero or positive. It is always assumed that trade entering under a preference always enters under the most favorable scheme. Thus M0ijp is estimated from M0ij using this utilization data.
-
2)
The analysis assumes that the tariff reductions for sugar, bananas, and rice for LDCs under the EU EBA program has already been implemented to avoid counting these changes as gains or losses in market access arising from the Doha Round. This requires a prior adjustment of import values for sugar, bananas, and rice in the base period using a formula equivalent to Equation (A.2).
Average Preference Margin and Average Preference Margin Including Domestic Production
The “average preference margin” enjoyed by country j in the United States (EU) is simply a weighted average difference between the tariffs paid on U.S. (EU) imports from country j and the MFN tariff applicable to such imports, where the weights are given by country j's trade with the United States (EU):
where t0iMFN is the MFN tariff applicable to product i and all other variables are defined in the “Change in Market Access” section above.
The “average preference margin including domestic production” enjoyed by country j takes account of preferential access enjoyed by other producers and the zero tariff paid on U.S. output sold in the United States and EU output sold in the EU:
where t0i_AVERAGE is the tariff revenue collected on U.S. (EU) imports of product i divided by U.S. (EU) consumption of product i:
Identifying the Elasticity of Substitution in Demand
The estimation approach was developed in a paper by John Romalis (forthcoming). Demand elasticities are identified by examining where the U.S. and the EU source their imports of different products before and after the implementation of the Canada-U.S. Free Trade Agreement and the North American Free Trade Agreement (collectively referred to as NAFTA). Changes in U.S. import sources are explained using changes in the tariff preference afforded to products of North American origin. The idea is that where North American output is afforded no new preference (where the MFN tariff rate is zero, for instance), NAFTA's only impact should come through a general equilibrium effect on output prices, or through reductions in “border effects” owing to NAFTA provisions that go beyond tariff liberalization. When NAFTA causes a new preference to open up for North American goods, the preference should have an additional effect causing U.S. consumers to substitute toward newly preferred goods and away from other sources of supply. This strategy can be derived from a simple model.
Model description
Firms produce products under perfectly competitive conditions. Trade is driven by preference for variety and by products being differentiated by country of origin. Countries may impose ad valorem tariffs on imports. Countries may then enter into preferential trading agreements whereby each country in the agreement lowers tariffs on imports from partner countries but need not adjust the tariff on imports from other countries. This causes consumers to substitute toward the output of preferred countries and away from all other sources of supply, including domestic production. Factor supplies are not explicitly modeled. The model assumptions are set out in detail below.
-
1
Products and industries are indexed by i, countries are indexed by j, and time by t.
-
2
In each country j, every industry i produces a product i using an industry-specific factor under conditions of perfect competition with marginal cost c(q ijt s) (henceforth often denoted as c ijt ), where qs is the industry production. Note that marginal cost depends on the quantity produced and may vary across producing country and time.
-
3
In every period, consumers in each country are assumed to maximize Cobb-Douglas preferences over their consumption of the output of each industry, Q ijt , with the fraction of income spent on industry i being b ij (equation (A.6)). Expenditure shares for each industry are therefore constant for all prices and incomes.
-
4
The output of each industry is not a homogeneous good. Although firms in the same country produce identical goods, production is differentiated by country of origin. Q ijt can be interpreted as a subutility function that depends on the quantity of each variety of i consumed. We choose the constant elasticity of substitution (CES) function with elasticity of substitution σ>1. Let qijj′t denote the quantity of product i consumed in country j that was produced in country j′. Q ijt is defined as
-
5
There are transport costs for international trade. Transport costs are introduced in the convenient “iceberg” form; gijj′t units must be shipped from country j′ for one unit to arrive in country j; gijj′t=1, ∀ j .
-
6
For tariffs, τijj′t−1 is the ad valorem tariff imposed on product i imported by country j from country j′; τijj′t=1, ∀ j .
Equilibrium
In equilibrium, consumers maximize utility and firms maximize profits. Because of the assumption of perfect competition, prices (exclusive of tariffs and transport costs) are equal to marginal cost, c ijt . Consider the consumers in country 1, which we will call the United States. Tariffs and transport costs raise the price paid by U.S. consumers for goods imported from country j to c ijt gi1jtτi1jt. Let Y1t denote U.S. income. U.S. consumers maximize utility subject to expenditure being equal to income in every period:
Differentiating the Lagrangian for the consumers' constrained optimization problem with respect to consumption levels of each product, we find that tariffs on imported goods cause domestic consumers to substitute away from higher-taxed varieties. The amount of substitution depends on the level of the tariff and on the elasticity of substitution between varieties:
Equilibrium conditions for all other countries are symmetric, which will be exploited by the empirical work to control for the effect of unobserved movements in marginal cost that may be correlated with tariff movements.
Results
We use Equation (A.9) to derive estimating equations for demand elasticities. Equivalent equations exist for every other country; specifically, let country 2 be the aggregate of the 12 countries that were always members of the EU for the sample period 1989–99:
Using Equations (A.9) and (A.10) we can eliminate the marginal cost terms:
Elimination of the unobserved marginal cost terms is important because relative costs will shift following trade liberalization. Equation (A.11) can be transformed into an equation for c.i.f. import values, to match how EU trade data are collected:
As long as we examine only countries j and j′ for which the EU does not change its relative tariffs, 
is simply a product fixed effect. Because we do not have detailed transport cost data for EU trade, to identify σ we assume that relative transport costs of shipping products to the U.S. and the EU, 
, is the sum of a product fixed effect, a year fixed effect, and an error term that is orthogonal to U.S. tariffs. This produces the basic estimating Equation (A.13) based on c.i.f. import values, where D
i
and D
t
are full sets of product and year dummies respectively, while ɛijj′t is a random disturbance term:
Now consider country j to be Canada or Mexico and country j′ to be any other country. NAFTA's increase in the U.S. tariff preferences for Canadian and Mexican goods, 
, will increase the share of those goods in U.S. consumption relative to their share of EU consumption. The size of the increased share in an arbitrary industry i depends positively on the size of the increased U.S. tariff preference, and positively on the elasticity of substitution σ between varieties of i. The EU was chosen as country 2 for two main reasons. First, its detailed trade data are available electronically. Second, the EU is a relatively large trading partner for Canada and Mexico, which maximizes the number of products that can be used to estimate demand elasticities and increases the precision of the estimates.
Data
International trade data for almost the entire world is now collected according to the HS, a schedule that is standard across countries at the six-digit level, or approximately 5,000 products. The U.S. International Trade Commission (USITC) maintains a database at the 10-digit level (15,000 products) of U.S. imports classified by product, country of origin, import program, month, and port of arrival. Eurostat maintains a similar database for the EU.
Tariff data are based on either tariff schedules or detailed data on import duties collected. U.S. tariff schedules for 1997 to the current year are available from the USITC. We extracted U.S. tariff data for 1989–96 from USITC files. U.S. tariffs are almost invariably set at the HS eight-digit level (10,000 products). Tariffs are aggregated from the HS eight-digit level to the six-digit level in two different ways: by taking simple averages or by taking trade-weighted averages. There are several limitations to using tariff schedules to calculate tariffs. One limitation is the effect of the maquiladoras (export assembly plants that use imported inputs located near the United States-Mexico border) on Mexican exports to the United States. Under “production-sharing” provisions, duty does not have to be paid on the U.S.-sourced content of many exports to the United States, while the full value of those transactions is recorded in U.S. trade data. The tariff schedule will therefore often overstate the NAFTA preferences. A second limitation of the tariff schedule is that preferential tariff arrangements are often circumscribed by restrictive rules of origin that need to be satisfied to qualify for the tariff preference. To partly address these limitations we also calculate tariffs using data on actual import duty paid. The drawback of this approach is that tariff rates can only be observed when there is trade. Where there is no trade, we revert to the tariff schedule for that item. This alternative set of eight-digit “applied” tariffs are also aggregated to the six-digit level using simple averages and trade-weighted averages. This gives a total of four measures of tariffs at the HS six-digit level.
Quantitative restrictions on imports of many textile, clothing, and footwear products under the Multi-Fibre Agreement and of many agricultural products provide a further complication. Many of these restrictions are binding, although a large number are not (Evans and Harrigan, 2004). They are extremely difficult to account for, because many restrictions encompass many HS products and most apply bilaterally. The existence of binding quotas will tend to bias downward the estimated substitution elasticities. Eliminating products subject to quotas did not, however, lead to higher substitution elasticity estimates.
Elasticity of substitution estimates
The mean elasticity of substitution is estimated using Equation (A.13). We use HS six-digit trade and tariff data from 1989–99. Later years are omitted because the Mexico-EU free trade agreement commenced in 2000. Country j is alternatively Canada or Mexico, country j′ is the aggregate of all countries that did not substantially change their preferential trade relations with either the United States or the EU between 1989 and 1999. A list of these countries is provided in Table A.4. Four different measures of tariffs are used; depending on whether the tariff schedule or actual duty paid are used to calculate tariffs at the eight-digit level, and on whether tariffs were aggregated to the six-digit level using simple averages or trade weights.
Results are reported in Tables A.5 and A.6. Table A.5 reports results based on changes in the destination of Canadian exports and Table A.6 reports results based on the destination of Mexican exports. Reasonably precise estimates of the mean elasticity of substitution range between 6 and 11. Moving across the columns, the estimates are slightly sensitive to the choice of tariff measure—the estimates using Canadian exports are lower when the tariff schedule is used. The estimates based on Mexican exports tend to be higher than those based on Canadian exports. The estimates are very similar whether the “control” countries j′ are limited to those listed in Table A.4 or include all non-NAFTA countries. The estimates are similar in magnitude to elasticities estimated by Clausing (2001) and Lai and Trefler (2002). For the purposes of estimating the market access effects of proposed multilateral trade liberalization under the Doha Round, we choose to use the more conservative substitution elasticity estimate of 6. Note that in Table 8 we compare how sensitive these results are to choice of demand elasticity, using the higher estimate of 11 and an arbitrarily lower estimate of 3.5.