Climate change, mitigation and adaptation with uncertainty and learning

Abstract

One of the major issues in climate change policy is how to deal with the considerable uncertainty that surrounds many of the elements. Some of these uncertainties will be resolved through the process of further research. This process of learning raises a crucial timing question: should society delay taking action in anticipation of obtaining better information, or should it accelerate action, because we might learn that climate change is much more serious than expected. Much of the analysis to date has focussed on the case where the actions available to society are just the mitigation of emissions, and where there is little or no role for learning. We extend the analysis to allow for both mitigation and adaptation. We show that including adaptation weakens the effect of the irreversibility constraint and so, for this model, makes it more likely that the prospect of future learning should lead to less action now to deal with climate change. We review the empirical literature on climate change policy with uncertainty, learning, and irreversibility, and show that to date the effects on current policy are rather small, though this may reflect the particular choice of models employed.

Introduction

One of the major issues in climate change policy is how to deal with the considerable uncertainty that surrounds many of the elements needed, in terms of the scientific understanding of the processes driving climate change, for example, the risk of possible catastrophic effects of climate change, the impacts of any such changes on society and the economy, the extent to which these impacts might be ameliorated by future adaptation and the economic values to be attached to these impacts, for example, the appropriate social discount rate or equity weights.

This raises the obvious question of how such uncertainties should affect decisions about the current level of abatement of greenhouse gas (GHG) emissions and hence about the correct amount of mitigation and adaptation. What would be the difference if explicit account is taken of uncertainties as opposed to ignoring such uncertainties?

However, there is a more subtle aspect to the treatment of uncertainties, which is the focus of this paper. For some of these uncertainties will be resolved through the process of further research, for example, as exemplified in the reports of the IPCC. This process of learning raises a crucial timing question: should society delay taking action to reduce GHG emissions in anticipation of obtaining better information, or should it accelerate action, because we might learn that climate change is much more serious than expected but find that by the time we learn this information we have too little time to adjust. But this timing question also needs to take account of the fact that there are constraints on how quickly the stock of GHGs can be reduced in the future. So, if society learns that the damages from climate change are much greater than currently expected, it may not be able to take as much action to reduce the stock of gases as it would ideally like to take. This constraint is referred to as an irreversibility constraint. Much of the analysis of this timing question to date has focussed on the case where the actions available to society are just the mitigation of emissions. We extend the analysis to allow for both mitigation and adaptation, by extending the model of Ulph and Ulph (1997). We show that the inclusion of adaptation weakens the impact of the irreversibility constraint, so that for this model it is more likely that the prospect of future learning should lead society to take less action now to deal with climate change.

In Section 2, we summarise the theoretical results, which are set out fully in Appendix A. Firstly, those for where the only policy with regard to climate change is the mitigation of GHG emissions. Starting from a two-period model in which there is no uncertainty and no irreversibility, irreversibility, and uncertainty with the possibility of learning are introduced into this model, the consequences for optimal mitigation are derived. Then adaptation is introduced into the model in two ways. The first way is that adaptation acts in a way equivalent to a reduction in the stock of GHGs, the second is that adaptation reduces marginal damage costs. The general conclusion from these admittedly very simple models is that allowing for adaptation reduces the relevance of the irreversibility constraint, and results depend more on the pure learning effect.

In Section 3, we survey the attempts that have been made to assess the implications of uncertainty, irreversibility, learning and precaution in the context of empirical models of climate change. These studies suggest that the prospect of obtaining better information (earlier resolution of uncertainty) should lead to a reduction in current abatement levels, although these effects are small. Contrary to the simple intuition of the precautionary principle, with the exception of one scenario, all these studies suggest that the prospect of obtaining better information (earlier resolution of uncertainty) should lead to a reduction in current abatement levels, although these effects are small. However, the theoretical analysis of Section 2 would suggest that these results are not very surprising given the choice of utility functions (either logarithmic or quadratic) and that the environmental irreversibility constraint does not seem to bite even in worst-state scenarios.