In this paper, we update our earlier estimates of the cost of the Kyoto Protocol using the G-Cubed model, taking into account the new sink allowances from recent negotiations as well as allowing for multiple gases and new land clearing estimates. We then compare the protocol to an alternative policy outlined in McKibbin et al. (Brookings Policy Brief, No. 17. June, The Brookings Institution, Washington, 1997; Climate Change Policy After Kyoto: A Blueprint for a Realistic Approach, The Brookings Institution, Washington, 2002a; J. Econom. Perspect. 16(2) (2002b) 107) that does not impose rigid emissions targets. We focus particular attention on the sensitivity of compliance costs under each policy to unexpected changes in future economic conditions. To illustrate the issue, we evaluate the policies under two plausible alternative assumptions about a single aspect of the future world economy: the rate of productivity growth in Russia. We find that moderate growth in Russia would raise the cost of the Kyoto Protocol by as much as 50 percent but would have little effect on the cost of the alternative policy. We conclude that the Kyoto Protocol is inherently unstable because unexpected future events could raise compliance costs substantially and place enormous pressure on governments to abrogate the agreement. The alternative policy would be far more stable because it does not subject future governments to adverse shocks in compliance costs.