Discussion

Daron Acemoglu opened the discussion by recognizing the importance of the topic. Then, in agreement with the discussants, he argued for a number of potentially fruitful extensions of the model. First, he expressed the desirability ofwidening the scope of the analysis to include issues such as irreversibility and tipping points. He added that it would be useful to jointly analyze emissions and mitigation actions, which are themselves uncertain or subject to uncertain learning. In the presence of uncertainty, it might be optimal to be conservative in emissions butmore experimental in mitigation policies such as investment in renewable energy and carbon sequestration. This is because the latter act as insurancemechanisms. Mar Reguant expressed concern that in the current model, dire scenarios can be avoided by losing only 20% of utility, and this seems infeasible. Lars Hansen agreed that he and his collaborators look forward to pushing the analysis in the ways mentioned and in other directions, while preserving tractability. The point of this paper is to argue for taking a broader approach to uncertainty than is typical in this and other literatures and to show that this can be quantitatively important. In regard to tipping points in particular, Hansen added that the introduction of a probabilistic structure on tipping points opens up new research challenges by pushing beyondpure riskmodels to amore nuanced approach that recognizes the limited knowledge of the timing and magnitude of such events. The discussion then centered around the issue of discounting. James Stock pointed out that implementing a stochastic approach to discounting under uncertainty is a complex procedure that requires information on outcomes of covariances and consumption paths, which might