In the standard problem of mechanism design under adverse selection, it is well known that the transfer function from the principal to the agent will be increasing in the agent s unknown productivity when there exists an incentive compatible mechanism. Cost-sharing contracts in LDC agriculture are a particularly interesting form of such mechanisms. In this paper, we develop a simple model of cost sharing contracts and construct a measure of potentially unobservable household productivity by estimating a plot level production function with household-specific fixed effects, which are then purged of observable household characteristics. We then use the implications of the model and our measure of potentially unobservable tenant productivity to test whether the productivity of tenants is indeed unobservable to landlords. Our empirical results strongly suggest that adverse selection concerns are not empirically important : in the Tunisian village we consider, it would appear that peasants do much better than is expected by Spence and Mirlees.