Abstract
Although there is evidence suggesting that the effects of trade liberalizations likely
vary across the distribution of trade flows, trade economists have focused almost entirely
on conditional mean estimates of the trade elasticity. We propose the novel use of
Poisson-based expectile regressions to estimate the heterogeneous effects of EIAs across
the entire conditional distribution. Like standard Poisson regression, this method does
not need the dependent variable to be logged, accommodates a mass of observations
at zero, and is easy to implement, allowing the estimation of gravity equations with
the standard three-way fixed effects specification. Using the proposed estimator, we
find systematic evidence that agreements have larger effects at the lower tail of the
conditional distribution. We then use the proposed method to investigate the causes
of this heterogeneity, and our results suggest that the success of trade liberalizations
strongly depends on potential for expansions along the extensive margin.