Intensive agriculture is often bad for wildlife. Does this imply that a goal to boost wildlife on agricultural land is best met through a general reduction in intensity? We argue that such an approach may not be optimal, since cost functions for provision of wildlife on agricultural land may be non-convex, due to fixed costs associated with such provision. This implies that, even when farms are identical, it may be preferable to split them into groups of high providers and low providers.

We test our hypothesis in a study of the optimal management of mown grasslands in southern Sweden, where the two products are silage and successful reproduction of ground-nesting birds, and the variable controlled by the farmer is the date of the first mowing. We show that the optimal solution is likely to involve some farmers maintaining profit-maximizing practices while other—identical—farmers delay their first mowing significantly. The superiority of such split solutions may have major implications for agricultural policy.