to reduce total exports. The estimated coefficient on the amount of agricultural land per
worker (β7) is negative giving further support to the Hecksher-Ohlin theory: countries that
are relatively scarce in agricultural land endowment tend to import agricultural products
from other countries, and from Canada in particular. Consistent with the predictions of the
gravity model, the coefficients on landlocked and island are positive.13
In general, the estimates of the modified gravity model are broadly consistent with the-
oretical predictions and we can use it to address the main question of the analysis: the
effect of the CCFTA on the flow of Canadian exports to Chile. Controlling for cross-country
differences, coefficient β1 remains insignificantly different from zero. Therefore, prior to the
CCFTA Chilean agricultural imports from Canada were not statistically different from the
average Chilean trade partner. Moreover, the coefficient on the CCFTA dummy is always
insignificant, implying that the FTA had no effect on the volume of Canadian agricultural
exports to Chile. Together with the previous result for Canadian imports, the main con-
clusion of the empirical analysis is that the CCFTA generated disproportionately higher
benefits to producers from one member of the agreement, Chile, while Canadian producers
of agricultural products seem not to have benefited.
A potential problem with specification (1) is a possible endogeneity problem due to
correlation of PC and FTA dummies with the error term. Country-specific fixed effect
not captured by the list of country-specific variables may lead to the biased estimates of
the model parameters. In particular, we can think of technological differences, differences
in land quality, country-specific policies in support of agricultural sectors, etc. Presence
13 However, the distance coefficient becomes positive and significant. This can possibly indicate higher
demand for Canadian agricultural goods in more distant countries.
14