We consider the large range of estimated costs of meeting U.S. climate policy targets. Some of this range is due to different studies using different cost measures, and in principle such differences could be eliminated by more careful comparisons and greater transparency in studies that estimate costs. Still another source of differences is how the proposed policy is represented. Here, again, this source of difference could be reduced by more careful comparison and by better definition of implementation details. Even if these sources of difference could be eliminated, there would remain substantial uncertainties because of the difficulties of projecting economic activity over the long horizon of proposed policies. We show the importance of several of these factors, using a consistent modeling framework. Copyright © 2010 by Annual Reviews.