This paper reports estimates of the productivity premium of German firms exporting to the Euro-zone and beyond, controlling for unobserved time-invariant firm-specific effects. Furthermore, it tests for self-selection of more productive firms into exporting beyond the Euro-zone. The main contribution of the paper is to correct a serious flaw in hitherto published studies that ignore the potentially disastrous consequences of extreme observations, or outliers. The paper shows that estimates of the exporter productivity premium by destination are driven by a small share of outliers that comprises some five per cent of all firms. These outliers are identified by a recently developed highly robust estimator for models with fixed effects. Using a 'clean' sample without these outliers, the estimated productivity premium of firms that export to the Euro-zone only is no longer statistically different from zero at a conventional error level, and the estimated premium of firms exporting beyond the Euro-zone, too, over firms that serve the German market only is tiny. Furthermore, an ex-ante differential that is statistically significant and large only shows up for enterprises that exported to the Euro-zone already and start to export to countries outside the Euro-zone. These conclusions differ considerably from those based on non-robust standard regression analyses.