Existence of large income gaps between agricultural and non-agricultural workers in developing countries is well known, but the exact source of the gaps is debated. The two main hypotheses, barriers to labor mobility and sorting of workers based on unobserved comparative advantage, have distinct predictions for aggregate efficiency but are difficult to distinguish using only cross-sectional data typically available for developing countries. We use panel data from Indonesia Family Life Survey to document that workers who move out of agriculture see an income gain of over 20% while those who move into agriculture see a similar income loss, even if they stay in the same location. We then ask whether even such within-worker sector premia tell us anything about the presence of barriers to sectoral mobility. By themselves, they do not. However, taking into account a richer set of moments of the joint sector-income distribution over time allows us to identify the role of self-selection across sectors and of barriers to sectoral mobility. Our estimates indicate that while self-selection is important, there are also barriers that significantly misallocate workers across sectors. Removing such barriers would lead 35% of workers to reallocate and as a result would increase aggregate output by as much as 21%.
