Since Rose’s (2000) path-breaking study, much work has been written on the effect of currency unions on trade. The striking finding reported in that seminal paper, the empirical evidence that sharing a currency union more than triples trade between countries, together with the creation of the euro, has motivated an intense debate on this issue and, in particular, on the effect of the Economic and Monetary Union (EMU) on trade, both in academic and policy making circles. The empirical work on the EMU effect on trade uses different specifications of the gravity equation of international trade. Despite the fact that the empirical literature has progressively improved its theoretical foundations and the econometric specifications to account for potential sources of bias, almost all results rely on gravity equations with serious theoretical and econometric problems. This paper investigates the EMU effect on trade by member country and the direction of trade flows. The empirical analysis uses a recent econometric development, which allows us to estimate gravity equations dealing with heteroskedastic residuals and zero bilateral trade flows on a large span of data (across both countries and periods). Our results provide evidence of no EMU effect that is robust to alternative samples of countries and periods as well as to the use of data for consecutive years or at five-year intervals. However, this consistent result across samples masks heterogeneous effects across member countries. The analysis by country shows robust evidence of a positive effect of EMU on both exports and imports for Spain and Portugal. For Austria and Cyprus we also find a positive effect in all samples, but it disappears when we include a dummy variable specific for the European Union in the regression. By contrast, the EMU has had a negative effect on import flows for the Netherlands and Malta, while we find a negative impact for exports from Greece.