and EU estimates and the change from the Japanese and Chinese estimates to the matched estimates
derived from the Kee, Nicita, and Olarreaga (2005) database. Table A.48 provides these paired “past”
and “present” estimates for the four countries, made possible by our concordance in table A.47. We then
calculate a weighted average (weighted by total 1990 imports for each country) from the average percent
change in the tariff equivalents of NTBs for each of the four countries. We apply this figure, namely, a
51.1 percent change from “past” to “present,” to every “present” GTAP sector estimate to calculate “past”
NTB rates.8 Specifically, multiplying a “present” tariff equivalent of an NTB by 1 divided by 1 — 0.511
gives the “past” NTB estimate. The estimates of past and present NTB rates for the 18 countries are
displayed in tables A.49 through A.66.
Transportation Costs
The GTAP7 database contains 2004 (i.e., “present”) estimates of transportation costs. The manner
that transportation costs are modeled in GTAP does not allow us to simply insert “past” estimates of
transportation costs. We must therefore determine how much to shock present transportation costs to
simulate past transportation costs. To do this, we use a database from Hummels (2007) that contains
the transportation costs and value of US imports from over 100 countries from 1974 to 2004. The
database is disaggregated at the leaf level (i.e., the most disaggregated level, which is either 4-digit or
5-digit depending on the good) of the Standard International Trade Classification (SITC) revision 2.
We calculate both “past” and “present” ad valorem equivalents of transportation costs (transportation
charges divided by import value) from the database; comparing the two tells us how much to shock the
transportation costs already in GTAP to simulate “past” costs.
Using a concordance from HS 1996 through SITC revision 2 to GTAP developed by the Centre
D'Etudes Prospectives et D'Informations Internationales (CEPII), we collapse the SITC leaf level
codes into GTAP sectors for each of the 17 US partner countries in our analysis for available data from
1979 to 1981 and 2002 to 2004.9 We take the sum of the value of transportation charges (i.e., freight
and insurance) and import values for the underlying SITC codes by GTAP sector. We then divide the
transportation charges by import value to determine the ad valorem equivalent of transportation costs. To
control for variability we take the simple average of available data from 1979 to 1981 for “past” estimates
8. We were also able to observe the percent change in the ad valorem equivalents of Norwegian NTBs from past to
present. Interestingly, the average percent change, namely a 51 percent decline, was nearly identical to our estimate. This
does prove that our estimate is right, but it is an interesting result. The Norwegian data are presented in Deardorff and
Stern (1997) and were originally calculated by Holmoy and Haegeland (1994).
9. The concordance created by CEPII has, at times, multiple GTAP sectors paired with one SITC code. In these instances,
we consider the mode (i.e., most frequent) GTAP sector. If more than one mode exists we use the first listed mode. For the
purposes of filling in gaps in the data later we also collapse the SITC rev.2 data into GTAP sectors for several other years.
The leaf level of SITC codes refers to the end code for any commodity type; most codes terminate at the 5-digit level,
others at only the 4-digit level.
45