Climate Change 2001: Working Group III: Mitigation

	Other reports in this collection

Continued from previous page

Figure 10.2: Optimal hedging strategy for low probability, high consequence scenario using a cost-benefits optimization approach.

One way to avoid the bias inherent in the framing of the emissions control problem under uncertainty is to reframe the problem as a decision tree structure within the context of cost–benefit analysis rather than cost-effectiveness analysis. This was the approach taken by the seven models used in an Energy Modeling Forum (EMF; Manne, 1995) exercise on climate change decision making under uncertainty (Weyant, 1997). The study focused on hedging strategies for low probability, high consequence scenarios in which uncertainty was not resolved until 2020. Two parameters were varied: the mean temperature sensitivity factor and the cost of damages associated with global warming. The unfavourable cases were defined as the top 5% of each of these two distributions. Two surveys of expert opinion were used to choose the distribution of these variables. For the opinion survey on climate sensitivity, see Morgan and Keith (1995), and for warming damages, see Nordhaus (1994b). Figure 10.2 (Manne and Richels, 1995; Manne 1995) shows what happens when the unfavourable case has a probability of 0.5 and the expected case a probability of 0.95 (the two parameter values assumed for the unfavourable case are shown in the surveys cited above as being in the upper 5% of each of the distributions of the two key parameters, i.e., climate sensitivity and climate damages). The dashed lines show what happens if perfect foresight is available and can make today’s decisions in the full knowledge of which of these outcomes will occur. The solid lines indicate the average results from an economically efficient hedging strategy. The analysis takes into account both the costs and benefits of emissions abatement. With a cost-benefit analysis, costs and benefits are balanced at the margin. Seven EMF modelling teams have confirmed these results (Weyant, 1997). The reason for so little hedging is the low probability of the extreme outcome, that is 0.25%. If one were to increases this probability, the desired degree of hedging would increase accordingly.

Another parameter for stochastic cost–benefit analysis is the importance of non-linearity in the impacts and the date at which some threshold is likely to occur. Peck and Tiesberg (1993) observed that optimal policies were more sensitive to uncertainty in the damage-function power parameter than to uncertainty in the scale parameter. Ha-Duong et al. (1999) confirm this view and demonstrate that introducing thresholds in the damage function leads to more significant decoupling from current emissions trends for a given probability distribution.

Ultimately, as recognized in the IPCC (1996c) one should try and assess the option value of the information incorporated in alternative emissions pathways, that is the capacity of society to adapt to any new information. As pointed out by Ulph and Ulph (1997), the environmental irreversibility has to be balanced against the technological irreversibility, including the crowding out between forms of technical progress. Ha-Duong (1998) finds, comparing Working Group I (WGI) and Wigley, Richels, Edmonds (WRE) strategies, that the magnitude of the value of information is significant compared with the opportunity costs of abatement. On the basis of nine scenarios he found that the information value of acting soon is, for most of them, higher than that of acting later, if low and high damages are assumed equally probable.

Whatever the approach, the basic message is quite similar. First the costs and benefits of quick action have to be balanced against those of delayed action; second, to assume that the concentration target is known with certainty is an over-simplification of the decision problem. What is needed is an approach that explicitly incorporates uncertainty and its sequential resolution over time. The desirable amount of hedging should depend upon assessment of the stakes, the odds, and the costs of policy measures. The risk premium – the amount that society is willing to pay to reduce risk – ultimately is a political decision that differs among countries.

Uncertainty also affects the choice of policy instrument. In principle many mechanisms can be employed to limit emissions, including, voluntary agreements among domestic and international parties, regulation, taxes, subsidies, and quotas or tradable permits (see Chapter 6). Economists have focused on the potential role of taxes and quotas because these tools hold potential for cost minimization. Although both instruments are equivalent in a world with complete information (the optimal quota leads to the same marginal abatement cost as the optimal tax level), Pizer (1999), building upon a seminal work by Weizman (1974), demonstrated that this is not the case if uncertainties about climate damage and GHG abatement costs are considered.

Indeed, welfare losses that result from imperfect foresight depend on whether the steepness of the marginal abatement cost curve is higher or lower than that of the damage curve. Hence the finding that a co-ordination through price is preferable as long as the probability of dramatic non-linearity in climate systems is not large over the middle term. This policy conclusion can be reverted if the transaction costs of adopting co-ordinated taxation, high level of risk-aversion to catastrophic events, or a large amount of “no regrets” policies are considered. The main message, however, is that in a tax co-ordination approach costs are observable (while the outcome is not predictable), but in a quota approach the outcome is observable although there is an uncertainty about the resultant costs. In this respect, emissions trading is logically a companion tool for a system of emissions quotas, to hedge against the distributional implications of surprises regarding abatement costs and emissions baselines.