Balassa–Samuelson in the Long Run: Qualitative Success, Quantitative Limits

I qualitatively and quantitatively revisit the Balassa–Samuelson (BS) mechanism in the long run. Traditional panel regression specifications without time fixed effects are fragile, but adding time fixed effects yields a stable, positive BS elasticity across samples and frequencies—evidence that the data support BS qualitatively on average across countries. Quantitatively, however, a standard multi-country trade model fed only by observed sectoral productivity cannot match country paths and delivers too small magnitudes. These failures persist with costly trade, multi-country, multi-sector settings, input–output linkages, and time-varying trade costs.