This article examines the charge that Thirlwall's law is a theoretical tautology. It shows that a certain approach to empirical testing of that law can sometimes – under conditions analysed here – result in econometric estimates that reflect an approximate identity or ‘near-tautology’. Nevertheless, other methods of empirically testing the law are not subject to the near-tautology critique, and hence the theory itself is not a tautology. Econometric estimates for the US and Mexico reveal that the near-tautology critique applies to data for the former but not the latter; the difference in these results is explained by exactly the reasons discussed here. The article offers an alternative interpretation of Thirlwall's law as implying a benchmark for analysing whether national income, rather than relative prices, is the main adjusting factor in response to current-account imbalances in the long run.