We investigate the cost of business-cycle uncertainty (lack of firm knowledge about the prevailing state of the economy) in a setup where the economy switches between booms and recessions at random intervals. Calibrating an exchange economy model to match the properties of the postwar U.S. data, we find that giving consumers additional information beyond that already contained in the endowment growth rates yields only moderate gains. In a second stage, we investigate the effect of nonperfect information processing in this setting. Surprisingly, we find that opting for slow learning might yield large utility gains, especially for consumers with a strong preference for early resolution of uncertainty. © 2011 Cambridge University Press.