This paper estimates an agency model of contracts used in California's processing-tomato industry. Model estimation proceeds in three stages. We first estimate growers' stochastic production possibilities, and then, for a given vector of preference parameters, compute an optimal compensation schedule. Finally, we compare computed compensations with actual compensations and choose preference parameters to minimize distance between the two. Assuming perfect competition and risk neutrality for processors, we obtain an estimate of .08 for growers' measure of constant absolute risk aversion (where returns are measured in units of $100/ton), and find that growers' effort cost is 1.8% of total operating cost. Welfare losses from information constraints are estimated at .59% of mean compensation, and quality measurement improves efficiency (measured as a percentage increase in expected quality) by 1.08%.