Second thoughts on the exporter productivity premium

Contrary to the prevailing interpretation, this paper shows that the central models of trade with heterogeneous firms (<link>Melitz 2003</link>; <link>Bernard et al. 2003</link>) exhibit ambiguous predictions for the exporter productivity premium. This prospect arises because of differences between theoretical and empirical representations of firm productivity. Instead of marginal productivity, we examine in both models the theoretical equivalent of empirically observable productivity (value-added per employee). Given the presence of fixed export costs or heterogeneous mark-ups and trade costs, the observable productivity of exporters in proximity to the export-indifferent firm turns out to be lower than that of non-exporters; that is, the productivity distributions overlap. The paper reviews empirical literature that reports non-positive exporter productivity premia in firm-level data and discusses implications for empirical research on exporter performance, including learning and the role of non-parametric regressions (stochastic dominance, quantile regressions), fixed costs, and productivity distributions.