A thought experiment is designed to investigate whether the structure of risk aversion (i.e., the changes in absolute or relative risk aversion associated with changes in wealth) can be estimated with reasonable precision from agricultural production data. Findings strongly suggest that typical production data are unlikely to allow identification of the structure of risk aversion. A flexible utility parameterization is found to slightly worsen technology parameter estimates. Results also indicate that even under a restricted utility specification, utility parameter estimates are biased. Further, their quality is much worse when shocks are not large or samples are small.