Policies designed to avoid "dangerous anthropogenic interference" with the climate system often focus on targets for atmospheric carbon dioxide between 450 and 550 ppm. At the same time, economic analyses of policies designed to achieve these outcomes yield conflicting results, with the variability in global cost estimates in some recent comparisons spanning more than an order of magnitude. In this talk, I revisit the mitigation cost question in one well-documented integrated assessment model (MERGE), focusing on the sensitivity to (a) baseline assumptions, (b) the rate of energy efficiency improvements, (c) the availability of future carbon-free energy and (d) the magnitude of the future carbon sink. By considering these effects in isolation, I control for many of the factors that prohibit mechanistic attribution in larger inter-model comparison studies. The results can be interpreted in one of two ways: as uncertainty analysis in which different combinations of parameters represent different possible future worlds outside the control of policy, or alternatively, as policy assessment, in which different collections of assumptions represent the endpoints of deliberate policy intervention. I discuss the implications of the latter interpretation in the context of recent political developments.