Which sectors make the poor countries so unproductive?

Standard growth accounting exercises find large cross-country differences in aggregate TFP. Here we ask whether specific sectors are driving these differences, and, if this is the case, which these problem sectors are. We argue that to answer these questions we need to consider four sectors. In contrast, the literature typically considers only two sectors. Our four sectors produce services (nontradable consumption), consumption goods (tradable consumption), construction (nontradable investment), and machinery and equipment (tradable investment). Interacting the data from the 1996 benchmark study of the Penn World Tables with economic theory, we find that the TFP differences across countries are much larger in the two tradable sectors than in the two nontradable sectors. This is consistent with the Balassa--Samuelson hypothesis. We also find that within the tradable sectors the TFP differences are much larger in machinery and equipment than in consumption goods. We illustrate the usefulness of our findings by accounting for the conflicting results of the existing two--sector analyses and by developing criteria for a successful theory of aggregate TFP.