A model of a voluntary "green" payment program is developed to control nitrate leaching and runoff from corn production in New York. The program achieves environmental goals through self-interested choices of farmers, grouped by the productive and environmental characteristics of soils. It considers randomness in prict}~, production, and environmental damage. Farmers are assumed to maximize expected utility subject to chance constraints on severe levels of nitrate contamination. This program compensates farmers for applying environmentally safe levels of nitrogen fertilizer. If information is symmetric, program participation conditions require that the post-policy expected utility is at least as large as pre-policy expected utility. Under asymmetric information payments must be set so that farmers in group i always prefer their own policy over group's policy. If information is symmetric, the two groups have separate optimal policies. If information is asymmetric, separate policies are optimal if, and only if, the group with higher marginal productivity of nitrogen can meet environmental standards more easily; otherwise only a single policy need be specified. An empirical application of the model to three New York farming regions is based on estimated yield and environmental damage relationships from New York soils data. Asymmetric information between producers and the government would impose a cost burden on society. Separate policies are specified for the two groups, and the cost of information is as high as $11 per acre. The group most susceptible to nitrate leaching and runoff receives a windfall benefit, but this group makes up only about 10% of the total corn acreage. Two alternative methods of defining environmental quality standards are examined. First, environmental standards impose relative (percentage) reductions from pre-policy levels of nitrate loss. The second method requires an absolute level of environmental quality. The optimal payments range from $1 to $28 per acre. Under relative standards, payments are highest in the region with the highest level of pre-policy environmental quality. This situation is reversed when absolute standards are imposed. At an aggregate level, the program payments would range from $0.5 million to $3.5 million over the regions combined, representing between 3% and 18% of total government payments received by farmers in the three regions in 1992. The effect of risk aversion on program payments depends on whether nitrogen fertilizer is a risk increasing or risk reducing input. For the New York case studied, nitrogen was a risk reducing input for the group with higher yielding soils, and a risk increasing input for the other. Higher levels of risk aversion increase payments for the first group and decrease payments for the second.