The methodology proposed in this paper can be improved in several ways. Although the results
satisfy several robustness tests on distributional assumptions, using techniques that rely less on normality
for identification would be more appropriate. In addition, using a complete system of production functions
equations (including first order conditions for all endogenous inputs) and jointly estimating the stochastic
production frontier and the leasing model to allow for correlation between the errors are likely improve the
consistency and efficiency of our results. Finally, a data set that matches landlords with tenants and
contains detailed farm level land quality measures can help us better isolate farmer skills from technical
efficiency. Since all these improvements require significant advances in computational methodology or
considerably more detailed data sets, they are left as useful topics for future research.
43