developed country in the trade flow. Trade flows involving less developed countries are
relatively low, but the coefficients for GDP may not fully capture this in estimating the
regression function. Some of it then will be captured by the coefficient for log distance.
After the introduction of country-specific dummies, the dummies for less developed
countries will generally reflect the downward effect of trade between developed and
developing countries.
The logic for a gravity equation specified with distance rather than its natural log, as in
specifications 5 and 6, lies in the impact on the effect of other variables related to
geographical distance. Because of the monotonously increasing, concave form, countries
which are relatively close geographically will yield more than proportionate variation in log
distance. As a result, the natural logarithm of distance gives more weight to countries that
are relatively close, in its partial explanation of trade, compared to the explanatory effect of
variation in distance itself. Consequently, log distance may substitute somewhat for the
influence of other bilateral explanatory variables that are negatively correlated to distance,
such as common border and common trade block. We can assess the proposition through
comparison of the results for specification 5, which introduces distance directly as
explanatory variable, with those for specification 3. The straightforward use of distance
leads to a much higher estimated effect for common border (from 79% trade increase to
371%) and a somewhat less dramatic increase for free trade blocks (still, from 146% to
213%). Also, the statistical significance of these effects improves. A similar change occurs
for specification 6, compared to model 4, where distance enters in a specification
characterized by country-dummies. The changes in effect size, though, are less remarkable.
The expected effect only occurs for common border, which indeed is most closely related
to small distance. The impact for the trade blocks variable, if any, is slightly negative.
20
Bilateral Trade Flows and Institutions