A consistent finding in the development literature is that average non-farm labor productivity is higher than average farm labor productivity. These differences in average productivity are sometimes used to promote policies which advance the non-farm sector. In this paper, we analyze the importance of two specification choices when comparing productivity gaps, using detailed household panel data from Malawi. Importantly, we are able to calculate both average revenue products (ARPLs) – similar to most of the sectoral productivity gap literature – as well as marginal revenue products (MRPLs). We show that the choice of productivity measure combined with the choice of production function specification can lead to different sectoral productivity rankings. MRPLs from translog production functions suggest the household farm sector is more productive than the household non-farm sector, while MRPLs from a Cobb-Douglas and ARPLs from both a translog and a Cobb-Douglas find the opposite ranking.