Backus et al. in (Am Econ Rev 84(1): 84–103, 1994) found that the cross-correlation function between terms of trade and trade balance resembles the letter S and labeled it the S-curve. Support for S-curve is rather weak in some cases, most notably US when aggregate trade data are used. Empirical regularities based on aggregate trade data may suffer from a potential bias stemming from aggregation. Indeed, by employing US bilateral trade data excellent support for the S-curve is recovered. The support, however, was rather weak at the level of US–UK bilateral trade. Suspecting aggregation bias again, we employ US–UK trade data at the industry level and find overwhelming support for the S-curve. Furthermore, commodity attributes play no role.