This paper examines whether sharecroppers and fixed-rent tenants in the rice farms of South Asia are distinguished by their farming skills. The idea that fixed-rent contracts are typically given to relatively skilled tenants dates back to the agricultural (tenancy) ladder hypothesis of Spillman . The screening models [e.g. Hallagan 1978] that have attempted to formalize this idea assume that landlords do not observe the tenants' skill levels. This assumption is restrictive, and has found little support in empirical studies. The principal-agent model proposed in this paper focuses on the differences between time-intensive and skillintensive labor tasks. I show that tenancy contracts are designed to match the provision of these tasks with the owners of time and skill inputs. Sharecropping, in this model, provides an incentive scheme that allows for the specialization between a time-abundant tenant and a skill-abundant landlord. The second part of the paper empirically explores this result with household-level data from Sri Lanka. A two-stage model that distinguishes the choice of contract from the extent of land leased is used. The results clearly show that relatively skilled farmers are more likely to become fixed-rent tenants. I also find that, conditional on contract choice, farming skills do not affect the extent of land leased. A substantial part of the empirical analysis is devoted to the measurement of farming skills. I interpret farming skills as the contribution of observed farmer characteristics to the technical efficiency of the farm. This measure recognizes that many dimensions of skills are observed, and the use of weights computed from a production function to construct the skill index is theoretically more appealing than the ad hoc selection of proxy variables.