Is the impact of AGOA heterogeneous?

The literature reports the mean impact of trade preferences. The literature on the agoa impact is no exception. The mean impact can be sensitive to heterogeneity in the adoption of preferences by recipients. Nonetheless, the choice of countries included in the sample can play a role in determining the level of the impact. A small increase at the bottom of the distribution is more likely to present a large impact (in percentage terms) compared to the top 20\% -- 5\%. In this paper, we investigate the gains to developing countries focusing on where they lie on the export distribution. This way, heterogeneity can be controlled for and the impact at various percentiles of the distribution can be estimated. We carry out a quantile regression on a sample of countries selected by matching countries on an estimated propensity score as well as the full sample for comparison. Secondly, we decompose the impact using methodology in the spirit of the Oaxaca-Blinder decomposition found in Machado and Mata (2005) for quantile regressions. We find that gains to agoa recipients is confined to the top 5\% of the export distribution. On the contrary the gains to the recipients from exporting to the EU is mainly at the bottom 25\% and the median. There is an unambiguous decline in their exports to the rest of the world relative to the counter-factual countries they are compared with. The decomposition exercise supports the quantile results and shows that both coefficient and the covariate differences between the two groups explain the difference in the total change at the various quantiles.