The purpose of this study is to understand the implications of farm-to-farm and regional variations in nitrogen runoff and leaching for targeting specific policies to reduce nutrient contamination. To do this, we estimate 3D-year distributions of nitrate runoff and leaching for individual soils on nearly 150 farms in three farm production regions of New York and rank the distributions according to second degree stochastic dominance criteria. Based on these rankings, it is evident that cropland across farms and regions of New York is so heterogeneous that it is impossible to target policies to reduce nitrate contamination based on farm or regional characteristics. A much clearer ranking is found if soils are grouped by productivity group as measured by corn yield. Based on the estimated elasticities of nitrate runoff and leaching with respect to nitrogen application, one can target those areas where contamination problems are I most severe by focusing on soils with potential yields greater than 125 bu.lac. For it to make sense to target lower productivity soils, the productivity of additional nitrogen application at the margin on the highest yielding soils would have to be about double that of the lower yielding group. Evidence indicates that the ratios of productivities are less than unity in all three production regions.