of unobservable country-specific fixed effect motivates time differentiation of the regression
equation (1) to remove it:14
∆ln(Xict)=β0+β1FTAi+β2∆ln(Yit)+β3∆ERit+εict
where ∆Zit = Zi,2005 - Zi,1997. This simple specification is augmented with other vari-
ables:
∆ ln(Xict) = β0 + β1FTAi + β2∆ ln(Yit) + β3∆ERit+
+β4∆ALi + β5 ln(Di ) + β6ComLangi + β7Borderi + (2)
+β8Llockedi + β9Islandi + εict
The estimator for FTA dummy, which takes the value of one for Canada-Chile country-
pair observations, is an adjusted difference-in-difference estimator of the effect of an FTA,
where the control group is a set of countries that have no free arrangements with Canada. In
the benchmark specification we look at the change in the growth rate of Canadian agricultural
trade with Chile relative to the growth rate of imports from other countries during the period
1997-2005 when the Chilean market was completely liberalized for Canada. These results are
not very different from the analysis in level form: the flow of Chilean exports to Canada has
increased by around 60 percent as a result of the agreement, while the effect of the CCFTA
on Canadian exports is insignificant.
One of the possible reasons for the asymmetric effect of the CCFTA on agricultural trade
between member countries is the difference in tariff elimination schedules. In 1997, Canada
and Chile started with very similar protection levels for their agricultural products. As it
14 Differencing out the country specific fixed effects (country fixed effects are constant over the two periods
so differencing will cancel them out)
15