This paper estimates structural vector autoregression models of output, the real exchange rate and trade balance for the group of seven leading advanced economies (G-7). Unlike previous studies, we do not impose long-run purchasing power parity as an identifying assumption; instead, the shocks underlying the model are structurally identified using a set of theory-consistent sign restrictions. Empirical results show that nominal shocks account for most of the long-run variability in trade balances across the G-7 countries. We are able to attribute this finding to long-run movements in the real exchange rate, as the real exchange rate is significantly affected by nominal shocks in the long run.