Growth accounting exercises point to aggregate TFP differences as the dominant source of the large cross-country income differences. In this paper, I ask which sectors account for the aggregate TFP gap between rich and poor countries. Data limitations for developing countries have led researchers to use indirect methods for estimating sectoral TFPs. This paper proposes a new approach for estimating sectoral TFP using panel data on sectoral employment shares and GDP per capita. The approach builds a model of structural transformation and uses it to infer sectoral TFP time series consistent with the reallocation of labor between sectors and GDP per capita growth of a set of developing countries over a 40-year period. I fi nd that relative to the US, developing countries are the least productive in agriculture, followed by services and then manufacturing. While these fi ndings are consistent with empirical studies, they differ from fi ndings in the growth literature.