believe that a swing this large in protection would be unrealistic, so they also estimate the impact of a
reversion from current tariffs to the highest MFN applied tariff over the last 13 years by product for every
country. In this second scenario, they estimate a decrease in world trade of 3.2 percent.
Before going further, we pause to note an important critique, developed by Kei-Mu Yi (2003), of
the method used by various scholars—Bouet and Laborde (2008), ourselves, and many others—who have
investigated the role of policy on trade expansion.5 These days, a great deal of trade involves vertically
integrated supply chains, where the same inputs may criss-cross the same border more than once in the
process of assembling the final product. This description characterizes the automobile industry in North
America (the United States, Canada, and Mexico) and many electronic goods. Under these circumstances,
any tariff reduction has a multiplied effect in enlarging trade flows, because it cuts the duty more than
once. By contrast, in standard models of the sort we use, trade gains are calculated as if the imported
input crosses the border only once.
To illustrate the difference, Yi (2003) analyzes the role of tariff liberalization on US trade growth
from 1962 to 1999 using a standard model and a model that takes into account vertical specialization—
i.e., criss-crossing trade. Yi (2003) finds, with one set of parameters, that the vertical model explains
35 percent of trade growth over the period while the standard model explains only 13 percent. Using a
different set of parameters, the vertical model explains 53 percent of trade growth versus only 29 percent
for the standard model.6 The difference between the vertical and standard models is more exaggerated
in later periods, suggesting an increasing role for vertical specialization. An inference from Yi’s (2003)
analysis is that our estimates may understate the impact of policy liberalization.
Turning first to our data and then to our findings, columns I and II of table 1 show estimates of
average MFN applied tariffs in the “past” and the “present” for the 17 major US partners. In general, the
“past” rates are the average of three years of available data from 1988 to 1993, with a preference for the
oldest data; the “present” rates are the average of three years of available data from 2002 to 2005, with
a preference for the most recent data. For most countries ad valorem equivalents of specific tariffs are
included in the calculations (e.g., a specific tariff of $100 per ton with a ton valued at $1,000 would be
expressed as 10 percent ad valorem). Using 1990 US export shares with the 17 partners as weights for the
“past” rates and 2004 export shares as the weights for the “present” rates, we determine that the average
MFN applied rate faced by the United States dropped from around 10.3 percent in the “past” (circa
1990) to about 7.4 percent in the “present” (circa 2004).
Columns III and IV of table 1 show our estimates of the average actual tariffs of US partners in the
5. Yi’s analysis was drawn to our attention by Alan Deardorff, professor of economics, University of Michigan.
6. Yi (2003) analyzes the export growth of manufactures in a two-country model, where the two countries are the United
States and the rest of the world treated as a single country.