Global Warming, Irreversibility And Learning

Analysis of policies to deal with global warming is complicated by three features of the problem: uncertainty about the extent of damages from global warming, the fact that global warming depends on stocks of greenhouse gases, which introduces an element of irreversibility into the problem, and the fact that decision makers in the future are likely to have better information about the possible damages from global warming. This raises the question whether the possibility of getting better information in the future about damages from global warming should lead policy makers to take more or less action now to abate emissions of greenhouse gases. A number of economists have argued that the literature on the "irreversibility effect", based on seminal papers by Arrow and Fisher (1974), Henry (1974a,b), suggests that society should take stronger action now to abate greenhouse gas emissions when there is the possibility of obtaining better information than would be the case if there was no possibility of obtaining better information. This view seems to be commonly held by those advising policy makers about policies towards climate change. However, Epstein (1980) showed that the analysis of the irreversibility effect by Arrow and Fisher, Henry and others was based on special cases, and he derived sufficient conditions for the irreversibility effect to hold, and for the opposite of the irreversibility effect to hold. In this paper we show that even the simplest model of global warming does not satisfy either of Epstein's sufficient conditions, so it is not possible to use Epstein's analysis to tell whether the irreversibility effect applies to models of global warming. We then derive our own sufficient condition for the irreversibility effect to hold. Finally we adapt an empirical model of global warming due to Maddison (1994) to include uncertainty, learning and irreversibility and show that for most parameter values current abatement of emissions of greenhouse gases should be lower when we allow for the possibility of obtaining better information about damages caused by global warming than when there is no possibility of obtaining better information.