A cost-benefit analysis of climate change: Uncertainties and the value of information

Even though climate change is a dynamic problem, we argue that its analysis in terms of steady-state conditions is instructive for policy purposes, analogous to often-cited targets for the stabilized global average temperature rise or atmospheric CO2 concentration. We analyze climate change in a cost-benefit framework and specify CO2 emissions as annual averages of the time-dependent emission profiles of Wigley et al. (1996) while relating these to stabilized atmospheric CO2 concentrations and the corresponding global temperature increases. The resulting time-averaged model is simple enough to allow a fully transparent sensitivity study with respect to the uncertainties of all the damage and abatement cost parameters involved. Our result for the time-averaged optimal emission level Eo = 8.7 GtCO2/yr (about a third of current emissions) turns out to vary by at most 2.3 GtCO2/yr (or less than 30%) for the range of plausible parameter values. To assess the significance of uncertainties we focus on the social cost penalty, defined as the extra costs incurred by society relative to the overall social optimum if one makes the wrong choice of the time-averaged emission level as a result of errors in the estimates of the costs and benefits of CO2 emissions abatement. In relative terms the cost penalty turns out to be remarkably insensitive to errors. For example, if the true damage costs are three times larger or smaller than the estimate, the total social cost of global climate change increases by less than 20% above its minimum at the true optimal time-averaged emission level. However, because of the enormous magnitude of the total costs involved with climate change (mitigation), even a small relative error implies large additional expenses in absolute terms. To evaluate the benefit of reducing cost uncertainties, we plot the cost penalty as function of the uncertainty in relative damage and abatement costs, expressed as geometric standard deviation and standard deviation respectively. Suppose continued externality analysis reduces the geometric standard deviation of relative damage cost estimates from 10 to 5, the benefit is 0.5% of Gross World Product (about 250 billion €). If further research reduces the standard deviation of relative abatement costs from 1 to 0.5, the benefit is 0.06% of Gross World Product (some 30 billion €).