This paper explores the link between exports and total factor productivity (TFP) in Brazilian manufacturing firms over the period 2000–2008. The Brazilian experience is instructive as it is a case of an economy that expanded aggregate exports significantly, but with stagnant aggregate TFP growth. The authors first estimate firm-level TFP under alternative assumptions (exogenous and endogenous law of motion for productivity) following Wooldridge’s (On estimating firm-level production functions using proxy variables to control for unobservables, 2009) GMM procedure. In turn, they use stochastic dominance techniques to assess whether the ex-ante most productive firms are those that start exporting (self-selection hypothesis). Finally, they test whether exporting boosts firms TFP growth (learning-by-exporting hypothesis) using matching techniques, to control for the possibility that selection into exports may not be a random process. Their results confirm the self-selection hypothesis and show that starting to export yields an extra TFP growth that emerges since a firm’s first year of exporting but lasts only one year. Further, this extra TFP growth is much higher under the assumption of an endogenous law of motion for productivity, which reinforces the importance of accounting for firm export status to study the evolution of productivity.