Global Environmental Change 17 (2007) 311���325 Some economics of ���dangerous��� climate change: Reflections on the Stern Review Simon Dietza,b, , Chris Hopec, Nicola Patmored aDepartment of Geography and Environment and Centre for Environmental Policy and Governance, London School of Economics and Political Science (LSE), Houghton Street, London WC2A 2AE, UK bTyndall Centre for Climate Change Research, University of East Anglia (UEA), Norwich, UK cJudge Business School, University of Cambridge, Cambridge, UK dDepartment for Environment, Food and Rural Affairs (DEFRA), Ashdown House, London, UK Received 1 March 2007 received in revised form 25 May 2007 accepted 28 May 2007 Abstract The Stern Review on the Economics of Climate Change concluded that there can be ������no doubt������ the economic risks of business-as-usual (BAU) climate change are ������very severe������ [Stern, 2006. The Economics of Climate Change. HM Treasury, London, p. 188]. The total cost of climate change was estimated to be equivalent to a one-off, permanent 5���20% loss in global mean per-capita consumption today. And the marginal damage cost of a tonne of carbon emitted today was estimated to be around $312 [p. 344]. Both of these estimates are higher than most reported in the previous literature. Subsequently, a number of critiques have appeared, arguing that discounting is the principal explanation for this discrepancy. Discounting is important, but in this paper we emphasise that how one approaches the economics of risk and uncertainty, and how one attempts to model the very closely related issue of low-probability/high-damage scenarios (which we connect to the recent discussion of ���dangerous��� climate change), can matter just as much. We demonstrate these arguments empirically, using the same models applied in the Stern Review. Together, the issues of risk and uncertainty on the one hand, and ���dangerous��� climate change on the other, raise very strongly questions about the limits of a welfare-economic approach, where the loss of natural capital might be irreversible and impossible to compensate. Thus we also critically reflect on the state-of-the-art in integrated assessment modelling. There will always be an imperative to carry out integrated assessment modelling, bringing together scientific ���fact��� and value judgement systematically. But we agree with those cautioning against a literal interpretation of current estimates. Ironically, the Stern Review is one of those voices. A fixation with cost-benefit analysis misses the point that arguments for stabilisation should, and are, built on broader foundations. r 2007 Elsevier Ltd. All rights reserved. Keywords: Climate change Cost-benefit analysis Catastrophic climate change Dangerous climate change Integrated assessment Risk Uncertainty 1. Introduction The Stern Review on the Economics of Climate Change concluded that there can be ������no doubt������ the economic risks of business-as-usual (BAU) climate change are ������very severe������ (Stern, 2006, p. 188). The total cost of climate change was estimated to be equivalent to a one-off, permanent 5���20% loss in global mean per-capita con- sumption today. And the marginal damage cost of a tonne of carbon emitted today was estimated to be around $312 (p. 344). These estimates are high in relation to the previous literature, as commentaries on the Review have already pointed out (e.g. Nordhaus, 2006 Tol and Yohe, 2006). For instance, in his review of estimates of the marginal damage cost of a tonne of carbon emitted approximately today, Tol (2005) finds that the mean estimate from the wider literature���peer-reviewed and otherwise���is $122/t C, while considering only the peer-reviewed ARTICLE IN PRESS www.elsevier.com/locate/gloenvcha 0959-3780/$ -see front matter r 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.gloenvcha.2007.05.008 Corresponding author. Department of Geography and Environment and Centre for Environmental Policy and Governance, London School of Economics and Political Science (LSE), Houghton Street, London WC2A 2AE, UK. Tel.: +442079557589 fax: +442079557412. E-mail address: s.dietz@lse.ac.uk (S. Dietz).

literature it is $43/t C. Indeed, since the first wave of cost- benefit analyses (CBAs)1 of climate policy in the early���mid- 1990s (e.g. Cline, 1992 Fankhauser, 1995 Maddison, 1995 Manne et al., 1995 Nordhaus, 1991, 1994a Peck and Teisberg, 1992 Plambeck and Hope, 1996 Tol, 1997), most have yielded low estimates of the cost of climate change and recommended modest optimal greenhouse gas (GHG) emission reductions (with the notable exception of Cline, 1992). Of course, the conclusions of this wider literature have been subject to criticism. Of the many arguments made, we will engage with two in this paper. The first is that most CBA studies are limited to a relatively narrow set of the most measurable impacts, which turn out to be (not entirely by coincidence) marginal to future economic development. They are particularly weak at representing the potential that does seem to exist for costs to escalate rapidly at high levels of global warming. This is linked with a technical economic concern, namely that the most aggregated cost- benefit models implicitly assume perfect substitutability between natural capital and other forms of capital (high- lighted by Neumayer, 1999). But climate damage could, it is argued, deplete and degrade so-called ���critical��� natural capital, which is essential for human development and the loss of which can neither be reversed nor be compensated by increasing production and consumption of other goods and services. In this paper, we connect these concerns to recent discussions about ���dangerous��� climate change (see Schellnhuber et al., 2006). The second is that CBA studies inadequately represent the uncertainty that surrounds the impacts of climate change (e.g. Azar, 1998 Roughgarden and Schneider, 1999 various contributions to Climatic Change, 56(3)). These two criticisms are related, since one of the principal sources of uncertainty is that surrounding the likelihood of catastrophic climate damages and the economic consequences of high levels of warming more generally. There are other criticisms. Traditionally, much of the focus has been on high discount rates (e.g. Azar and Sterner, 1996 Broome, 1992 Cline, 1992). We do not devote this paper to discounting, since it is well known that the monetary cost of future climate change is highly sensitive to it (most recently demonstrated by Guo et al., 2006). However, we will estimate the sensitivity of the Stern Review calculations to increases in the components of the social or consumption discount rate, in order to put other factors in context. A further set of criticisms has been levelled at the welfarist ethical framework underpinning CBA. Along this line of reasoning, it may be unethical to impose environmental damage on future generations, as compensation by way of increased consumption of non- environmental goods is inadequate in principle (e.g. Barry, 1991 Spash, 1994). This also amounts to postulating that natural resources are non-substitutable, but on an a priori basis. Legitimate as alternative ethical frameworks are, in this paper we adopt an avowedly welfarist perspective. The primary purpose of this paper is to show that the Stern Review���s estimates depend heavily on its relatively comprehensive treatment of risk and uncertainty on the one hand, and on its attempts to represent low-probability/ high-impact damage scenarios on the other. Moving from the least to the most encompassing modelling approaches in these respects makes a very substantial difference to the overall estimate of climate damage. Moreover, this difference is of a comparable magnitude to the differences caused by alternative discounting assumptions, with the obvious and important proviso that, in most of this analysis as in most others, a high (utility) discount rate renders largely irrelevant in present-value terms any serious consequences of climate change in the far-off future. That is to say, a low (utility) discount rate is almost always needed for uncertain, dangerous climate change in the far- off future to matter.2 In Section 2, we discuss in more detail how CBA studies can model ���dangerous��� climate change. We also survey concerns about how CBA studies treat risk and uncer- tainty. This inevitably means that we must mention discounting, since the estimation of social welfare losses due to climate change usually involves the application of the same utility function to the distribution of consumption across time, space and risks or states of the world. Throughout Section 2, we clarify the assumptions of the Stern Review���s IAM and welfare valuation and in Section 3, we provide additional details about them. Section 4 reports our results and Section 5 provides a discussion. 2. Dangerous and uncertain climate change 2.1. Dangerous climate change in integrated assessment models (IAMs) It has often been observed that the IAMs used in cost- benefit studies are limited to a relatively narrow set of the most measurable impacts, a point systematically made by Downing et al. (2005). In particular, very few IAMs have been extended to cover large-scale, discontinuous changes to the climate system. And none have yet been explicitly extended to cover so-called ���socially contingent��� impacts, which are large-scale, ���second-round��� socio-economic responses to climate change like conflict and migration. Since there appears to be a correlation between the measurability of impacts and their potential magnitude, such that many of those we understand least have the ARTICLE IN PRESS 1We use the term CBA quite loosely in this paper to capture any studies, which estimate the monetary costs of climate change/benefits of climate- change mitigation, the costs and benefits of mitigation together, or other closely related issues. 2On the other hand, we present some scenarios in this paper, in which the combination of the risk of very high climate damages and high risk aversion leads to high overall estimates, in present-value terms, even with a higher utility discount rate. More generally, if it could be demonstrated that the impacts of climate change in the short to medium term are very high, then the influence of the utility discount rate would be reduced. S. Dietz et al. / Global Environmental Change 17 (2007) 311���325 312

potential to be highly damaging, we might reasonably conclude that most IAM studies are restricted in their capacity to simulate the potentially rapid escalation of climate damages as warming proceeds. While perhaps not the most likely scenario, such an escalation is plausible and thus its omission is serious, which is the basic source of a criticism levelled by Neumayer (1999). For him, CBA studies deny the possibility that GHG emissions are depleting and degrading natural resources, which are essential for human development and the loss of which can neither be reversed nor be compensated by increasing production and consumption of other goods and services. Technically, these are essential, non-substitutable natural resources, or ���critical��� natural capital assets. These arguments provide a point of contact between the cost-benefit tradition and the recent framing of climate impacts in terms of ���dangerous��� climate change. Article 2 of the United Nations Framework Convention on Climate Change (UNFCCC) calls for stabilisation of atmospheric GHG concentrations such that ���dangerous anthropogenic interference��� with the climate system is prevented (United Nations, 1992), so the term ���dangerous��� is of both academic and political importance. Nevertheless, over 15 years of debate on the meaning of dangerous climate change has resulted in many definitions (Dessai et al., 2004 Schneider and Lane, 2006). This is unsurprising, since what is dangerous is ultimately a value judgement. Thus global CBA of climate policy makes just one of many possible interpretations of danger. The value judgements it makes are based on preference-satisfaction utilitarianism, which is almost always made operational by aggregating all of the effects of climate change on to a single metric, consump- tion-equivalent welfare or utility. This is calculated for a representative consumer worldwide (i.e. one global, very long-lived consumer) and discounted across time-periods. Different consumers may represent different world-regions and different states of the world. Dangerous climate change might then be defined as any process that produces losses in consumption-equivalent welfare, which are both rapid and large-scale relative to global trend consumption. What constitutes rapid and large-scale is arbitrary: there is no a priori basis for drawing the line between dangerous and ���non-dangerous��� at any particular rate or level of welfare loss. Nevertheless, we can provide a little more clarity by focussing on the processes by which they are brought about. These are the very same large-scale, discontinuous changes to the climate system and rapid, large-scale socio-economic responses to climate change, which most IAM studies fail to incorporate. Without them, losses in welfare-equivalent consumption over the next two centuries are predicted to reach just a few percent of global output, relative to estimates without climate change. Given that trend growth in global consumption per capita is often projected to average around 1% per annum or more over the same time period, this is at most a very small ���blip��� on the global growth path. With them, comparable losses can run into many tens of percent of global output, impacts which we might label ���dangerous���. This demands a close examination of the pathways along which IAMs can simulate dangerous climate change, as we have defined it. Evaluating the Stern Review further requires that we understand to what extent its models did so. It is expositionally convenient (i.e. in terms of how they are actually represented by IAMs) to distinguish between three such pathways: 1. Rapid, large-scale impacts of gradual climate change 2. Abrupt, discontinuous and large-scale positive natural feedbacks in the climate system that accelerate global warming 3. Other abrupt, discontinuous and large-scale changes to the climate system that have more direct economic impacts. The first comprises rapid, large-scale impacts of what we might call ���gradual��� climate change. In the most aggregated IAMs like DICE (Nordhaus, 1994a) and PAGE (Hope, 2006), the latter of which was used by the Stern Review, this pathway is represented by a damage function that relates overall impacts, on consumption- or income-equivalent welfare,3 to an index of global mean temperature. Generally, this function is calibrated first through an estimate of overall impacts at 2.5 or 3 1C warming and second through an estimate of the functional form that traces impacts from zero warming, through the estimate for 2.5 1C/31C warming, and beyond to higher global mean temperatures. A very simple specification of the functional form, in this case with respect to 2.5 1C warming, is as follows: d��t�� �� b T��t�� 2:5 g , (1) where T is warming at time t, in terms of global mean temperature above pre-industrial, d(t) is the economic damage caused by climate change expressed as a fraction of consumption or income, b is the consumption loss accompanying 2.5 1C warming and g is the damage function exponent. IAM studies are in broad agreement that the impacts of climate change at 2.5���3 1C warming are small in relation to trend growth in global consumption (i.e. growth forecast in ARTICLE IN PRESS 3 The relationship between consumption and income/output depends on the marginal propensity to save, which may be endogenous to the model or set exogenously. In the Stern Review modelling, the saving rate is a constant 0.2, so consumption-equivalent losses can simply be scaled to income-equivalent losses. However, the role of endogenous saving, and in turn investment, is potentially important. The response of gross saving could dampen the negative impacts of climate change, if agents wish to compensate for the faster depreciation of the capital stock due to climate change. On the other hand, it could amplify them, in the main because capital is less productive due to climate change and thus the rate of return on saving is lower. The effect on net saving, however, is unambiguously negative, because the positive incentive to offset capital depreciation nets out (Fankhauser and Tol, 2005). S. Dietz et al. / Global Environmental Change 17 (2007) 311���325 313