9 July 2014 Why is Risk Aversion Unaccounted for in Environmental Policy Evaluations? By Dr. Noah Kaufman Government regulations can reduce the uncertainty associated with large environmental risks, such as catastrophic events caused by climate change, hydraulic fracking and The final version of this nuclear power plant meltdowns. Individuals dislike large risks—insurance companies publication appeared in earn profits because of the risk premiums that are paid to avoid risks such as sickness, Climatic Change Volume fires, floods and car wrecks—so there are considerable benefits to society associated 125, Issue 2, July 2014, with regulations that reduce or remove environmental risks. In welfare assessments, pp 127-135. The final economists typically use concave utility functions and estimate “option prices” to publication is available account for risk aversion. at http://rd.springer.com/ article/10.1007/s10584-014- Nevertheless, environmental policy evaluations in the U.S. customarily disregard these 1146-8. risk-reduction benefits. Environmental regulations are increasingly influenced by cost- benefit analyses that are performed based on the guidance of the Office of Management and Budget (OMB).1 The guidance of OMB for benefit-cost analyses is to estimate expected benefits and costs in monetary terms (i.e. a weighted average is calculated using the probabilities and monetary net benefits of all potential outcomes). Unlike a rigorous economic analysis, policy evaluations that follow this OMB guidance do not account for the effects of risk aversion.2 This essay offers two potential explanations for why risk aversion is typically unaccounted for in environmental policy evaluations. First, there is an extensive public economics literature on conditions whereby governments should behave in accordance with risk neutrality (i.e. zero risk aversion) when evaluating public investments with uncertain costs and benefits. Arrow and Lind (1970) showed that when populations are relative large, the risk premiums for small public investments with uncertain effects converge to zero because they can essentially be “spread out” among constituents. Indeed, both the U.S. and U.K. government documents that provide the official guidance on proper regulatory analysis specifically reference the uncertain costs and benefits of a regulation when recommending risk neutrality as the default assumption. A second rationale for ignoring risk aversion is computational and theoretical difficulties. Any attempt to quantify “societal risk premiums” will run into significant computational and theoretical problems. There is no well-accepted level of societal risk aversion, and no generally accepted methodology for converting monetary values into estimates of societal well-being that account for risk aversion. The choices of these modeling parameters would be controversial in any policy evaluation. Neither of these two rationales stands up to scrutiny. First, computational and theoretical difficulties are of course not a valid justification to disregard risk-reduction benefits. Second, Arrow and Lind (1970) is an endorsement for risk neutrality only for regulations that have uncertain costs and benefits, not for regulations that reduce uncertainty that exists in the absence of environmental policy (“baseline” or “business-as-usual” uncertainty). Policies that reduce pre-existing environmental uncertainty will provide risk-reducing benefits to all affected risk-averse individuals, and in no sense is the risk “spread out” across constituents. Policy evaluations should therefore account for risk aversion in situations when pre-existing uncertainty is significant. The implications of not doing so can be dramatic, as shown by Anthoff and Tol (2009) in sensitivity analysis on estimates of the social costs of carbon dioxide emissions. However, given the computational and theoretical difficulties of estimating risk premiums, each policy evaluations cannot be tasked with determining an appropriate methodology for estimating risk premiums. Instead, general guidance should be provided on how to incorporate risk aversion into policy evaluations, as has been done in the case of discounting future benefits and costs to present value terms. Despite the contentious ongoing academic debates surrounding the appropriate “social discount rate,” most U.S. policy evaluations follow the guidance of the Office of Management and Budget without controversy. Similarly, an expert panel could provide guidance on when and how the uncertainty-reducing benefits of regulations should be incorporated into environmental policy evaluations. The remainder of this essay is structured as follows. The next section provides an overview of the benefits of regulations that reduce pre-existing environmental uncertainty. The following section discusses potential reasons that risk aversion has not been accounted for in these policy evaluations. The final section provides recommendations for a way forward. The Benefits of Reducing Pre-existing Uncertainty In U.S. environmental policy evaluations, benefits and costs are nearly always calculated by estimating expected net benefits in monetary terms. Specifically, the net benefits for each potential resolution of uncertainty are estimated and assigned probably weights. Economic theory supports a different approach. The proper measure of a policy’s impact on social welfare is the willingness to pay of individuals prior to the resolution of uncertainty (Boardman et al., 2001). Economists often refer to this ex-ante calculation for measuring welfare impacts under uncertainty as the “option price” of a policy. www.nera.com 2 A key difference between the option price of a policy and an ex-post estimate of the policy’s expected net benefits is that the option price will account for preferences toward uncertainty. In other words, option prices account for risk aversion, whereas ex-post expected net benefit calculations typically assume a risk neutral society. If a risk cannot be eliminated by the purchase of insurance (at an actuarially fair price), then society will benefit from a public policy that reduces the risk (Boardman et al. 2001). Clearly, insurance is not available to protect against certain large-scale environmental risks, such as catastrophic climate events. A simple example may be useful. Suppose there are two potential states of the world that are equally likely in the absence of government regulation: (1) the Lucky state, in which the negative effects of a pollutant are relatively small and the consumption level is relatively high; and (2) the Unlucky state, in which the negative effects of the pollutant are relatively large and the consumption level is relatively small. Suppose further that public policy can be enacted that removes the pre-existing uncertainty. In this scenario, the negative effects of the pollutant are avoided, but consumption is decreased due to the cost of the regulation (to the Safe state, at the midpoint of Lucky and Unlucky). Figure 1 illustrates this example for two different assumptions on risk: (1) a risk neutral society, represented by a linear welfare function; and (2) a risk averse society, represented by a concave welfare function.3 Consumption levels are on the horizontal axis and societal welfare (i.e. utility) levels are on the vertical axis. Figure 1: Risk Reduction Benefits for Risk Neutral and Risk Averse Societies Risk Neutral (RN) Societal Welfare Utility Function Risk Averse (RA) URN(Lucky) = URA(Lucky) Utility Function URA(Safe) Ignored Risk- Reduction Benefit URN(Safe) = EURN = EURA URN(Unlucky) = URA(Unlucky) Unlucky Safe Lucky Consumption www.nera.com 3 To a risk neutral society, the expected welfare in the absence of regulation is EURN, which is the midpoint between “URN(Lucky)” and “URN(Unlucky).” This is equal to the expected welfare with the regulation, URN(Safe). In other words, this regulation provides zero benefits under risk neutrality. To a risk averse society, the expected welfare in the absence of regulation is EURA. With the regulation, expected welfare is higher, equal to URA(Safe). The benefit of the regulation to a risk averse society is the vertical distance between EURA and URA(Safe). Of course, the net benefits of environmental regulations in the real world will be a combination of the changes in the expected outcome and the changes in uncertainty. By holding constant the expected outcome, this example has shown that the effects of reducing uncertainty are unaccounted for under the assumption of risk neutrality.4 This problem will arise for any regulation that decreases pre-existing environmental uncertainty. Why Risk Aversion is Unaccounted for in Environmental Policy Evaluations Given how well-accepted risk aversion is as a preference trait, it is somewhat surprising that risk neutrality is typically assumed in environmental policy evaluations. There are at least two potential explanations: 1. A lack of distinction between regulations that reduce uncertainty versus regulations that cause uncertainty; 2. The computational and theoretical difficulties of incorporating risk aversion in cost- benefit analyses. Regulations that cause uncertainty versus regulations that reduce uncertainty There are two distinct types of uncertainty associated with most environmental regulations: 1. Baseline uncertainty, in which there is pre-existing uncertainty about environmental outcomes, independent of (or prior to) policy choices; and 2. Effectiveness uncertainty, in which the benefits and costs of an