When moral hazard concerns are present, standard contract theory predicts the "Marshallian inefficiency" of sharecropping contracts, in the sense that, ceteris paribus, sharecropping tenants will use different amounts of inputs than owner operators. In this paper, we examine this issue using a unique dataset collected in 1993 in the Tunisian village of El Oulja, thanks to the financial support of the PARADI program. We focus our attention on four questions that have been neglected by previous studies, namely: (1) cost sharing between landlords and tenants; (2) management inputs provided by landlords; (3) direct supervision of tenants by landlords;(4) repeated interaction between landlords and tenants. We implement panel estimation with household-specific fixed effects and control for the censoring of the dependent variable using the trimmed LAD estimator proposed by Honoré (1992). Our empirical results show that moral hazard is indeed an issue in tenancy contracts in the village, but that its quantitative importance in determining input use, in comparison with other factors, is relatively small. It follows that sharecropping is probably not chosen because of moral hazard concerns, and that other motivations, such as risk sharing or transaction costs, may be more important determinants of contractual choice.