Focus
Does income in poor countries catch up to that of rich ones? And did that change around the turn of the century? Recent research says it did; other papers contend that the methods behind that finding cannot measure catch-up at all. Which position is right?
Contribution
I place the competing specifications inside a single model, so that the assumptions of each side can be priced rather than asserted. The model grants the critics their case: the speed at which a country closes its own gap cannot be measured over 60 years, since the error is set by the length of the sample, not by how often income is observed. What the recent research leaves open is whether the change in poor countries’ ability to catch up is itself real.
Finding
This change is real, and it has a shape. The growth advantage of being permanently richer (pointing to divergence) held steady for 40 years, then fell abruptly in the decade from 2003 to 2013, and it has stayed down since (between-country force). That is a single shift rather than a gradual drift; the data do not indicate a steady decline of this force. I consider other factors, as suggested by the critics, that could rationalise the change in convergence. Yet none can reproduce a steady decline of the between-country force, including a simulation built to the leading critic’s own design, which moves in the other direction. Nor is this decline driven by any one region: the shift survives when China, resource exporters, the transition economies and East Asia are dropped from the analysis. Countries do close the gap in their own income trend (within-country force), at no less than about two thirds of a per cent a year. But these trends change, and 60 years are indeed not long enough to say how fast. What can be established with confidence is the increase in unconditional convergence since the end of the 20th century.
Abstract
Kremer, Willis, and You (2022) report that over six decades richer countries lost their growth advantage. Acemoglu and Molina (2022) and Pesaran and Smith (2026) reply that such regressions cannot measure catch-up when countries head toward different moving destinations. We nest both in one model. The movement survives across countries: the between-country income gradient of growth fell abruptly in the decade from 2003. Nulls with fixed destinations, or built to Pesaran and Smith’s specification, do not reproduce it. Only a lower bound on convergence speed survives: the within-country bias cannot be separated from the estimate.
