partners in 1980, 1990, and 2003. Due to data limitations, we use the estimate of transportation costs
for US imports as an estimate of transportation costs for US exports in the same periods. The data for
these rates come from Hummels (2007); ad valorem rates of transportation costs are calculated from
this dataset as the total of insurance and freight charges divided by import values. According to our data,
average transportation costs faced by US partners exporting to the United States (and, by proxy, US
firms exporting to those same partners) have been low since 1980, with a 4.3 percent rate in 1980, a 3.7
percent rate in 1990, and a 3.2 percent rate in 2003. In other words, the decline in transportation costs
has not been a big factor in trade growth.
In table 3 we present the actual tariff rate applied on US imports purchased from the 17 major
partners. The table follows the same method as columns III and IV of table 1 (“actual tariffs faced for
US exports”) but uses 1990 and 2004 US imports to calculate weighted averages. Other US rates (i.e.,
MFN applied, bound, and NTB) are displayed in table 4, which includes all the average tariff rates we use
for the partial equilibrium analysis. US tariff or NTB rates (i.e., rates on US imports) in the “past” and
“present” are lower than average tariff or NTB rates applied against US exports. For example, US MFN
applied tariffs went from 5.7 percent in 1990 to 3.8 percent in the “present,” while the weighted average
of US partner MFN applied tariffs dropped from 10.3 percent in 1990 to 7.4 percent in the “present.”
Our method for carrying out the partial equilibrium analysis of changed protection from “past”
to “present” loosely follows from Hufbauer and Elliott (1994). In the present analysis we are concerned
only with US trade in the aggregate, so we jump directly to price elasticities of demand for US exports
and imports. Table 5 shows various estimates of price elasticities for US exports and imports. Crane,
Crowley, and Quayyum (2007) and Mann and Plück (2005) provide useful surveys of the literature. We
use weighted average estimates from Kee, Nicita, and Olarreaga (2004), who take the novel approach
of calculating price elasticities at the tariff line level. We use a US import price elasticity (—1.30) that is
somewhat larger (in absolute terms) than our US export price elasticity (—1.17). These estimates are what
we consider “responsibly high” for the literature. We are comfortable with high estimates because the
partial equilibrium approach we use probably does not account for the full impact of closer economic
integration realized through policy liberalization or transportation cost declines.8
In table 6 we conduct the calculations for the six scenarios enumerated above. We calculate
the percentage point change in ad valorem rates for each scenario and apply it to the relevant price
elasticity to construct an “impact on trade” figure for each part (merchandise exports or imports) of the
six scenarios. We then multiply the relevant trade flow by one minus the “impact on trade” figure to
determine the hypothetical level of US trade with the policy reversion or transportation cost increase. We
subtract the hypothetical trade figure to determine the impact on annual US trade in each scenario.
8. See Bradford, Grieco, and Hufbauer (2006) and Yi (2003) for discussion.