This paper investigates the measurement of risk exposure in agriculture and its linkages with input and output decisions. We develop a conceptual analysis of risk under general risk preferences, including cumulative prospect theory. The approach is applied to a sample of U.S. farms from 1996 to 2011. In a multi-input, multi-output framework, the analysis documents the effects of management on production risk exposure and estimates the cost of risk under alternative frameworks. We find that variable inputs contribute to increasing risk, while livestock contributes to reducing risk. Nonfarm income reduces the cost of risk.