This paper aims at investigating the robustness of the Balassa-Samuelson (BS) effect to alternative proxies for a panel of 38 developing and emerging economies over the period 1980-2016. We examine the internal and external versions of the BS hypothesis using a total of five different measures. Relying on the Cross Sectional-Distributed Lag (CS-DL) approach, we show that the internal version of the BS hypothesis holds only if the labor productivity differential between the tradable and non-tradable sectors is used rather than the Gross Domestic Product Per worker. We also find evidence of a positive and robust effect of the relative price of the non-traded to traded goods on the real exchange rate. Overall, our findings highlight that while the verification of the internal version of the BS effect depends on the proxy considered for productivity, the validity of the external version is a general and robust result.