WPS7631 Policy Research Working Paper 7631 The Triple Dividend of Resilience Background Paper Investing in Disaster Risk Management in an Uncertain Climate Thomas K.J. McDermott Development Economics Climate Change Cross-Cutting Solutions Area April 2016 Policy Research Working Paper 7631 Abstract Climate change will exacerbate the challenges associated paper offers a simple decision framework that enables with environmental conditions, especially weather variabil- policy makers to identify the particular circumstances under ity and extremes, in developing countries. These challenges which uncertainty about future climate change becomes play important, if as yet poorly understood roles in the critical for disaster risk management investment decisions. development prospects of affected regions. As such, climate Accounting for climate uncertainty is likely to shift the opti- change reinforces the development case for investment in mal balance of disaster risk management strategies toward disaster risk management. Uncertainty about how climate more flexible, low-regret type interventions, especially those change will affect particular locations makes optimal invest- that seek to promote “development first” or “risk-coping” ment planning more difficult. In particular, the inability to objectives. Such investments are likely to confer additional derive meaningful probabilities from climate models limits development dividends, regardless of the climate future that the usefulness of standard project evaluation techniques, materializes in a given location. Importantly, the analysis such as cost-benefit analysis. Although the deep uncertainty here also demonstrates that climate uncertainty does not associated with climate change complicates disaster risk necessarily motivate a “wait and see” approach. Instead, management investment decisions, the analysis presented where opportunities exist to avail of adaptation co-ben- here shows that these considerations are only relevant for efits, climate uncertainty provides additional motivation a relatively limited set of investment circumstances. The for early investment in disaster risk management initiatives. This paper is a product of the Climate Change Cross-Cutting Solutions Area, and a background paper to “The Triple Dividend of Resilience” report, a joint initiative by the Global Facility for Disaster Reduction and Recovery and the Overseas Development Institute. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The author may be contacted at thomas.mcdermott@ucc.ie. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Investing in Disaster Risk Management in an Uncertain Climate Thomas K.J. McDermott School of Economics, UCC Grantham Research Institute on Climate Change and the Environment, LSE Keywords: Disaster risk management, climate change, uncertainty, decision- making JEL codes: Q54, D81, D61, O40 1. Introduction How does climate change alter the business case for disaster risk management (DRM) investments? How should policy makers treat the uncertain information provided by climate models and make efficient long-term climate-sensitive investment decisions when faced with the wide range of possible climate futures that models predict? This paper focuses on the role of climate change – and uncertainty in relation to future climate projections – in the decision making process for DRM investments. It takes the perspective of the individual policy maker (government or agency) at local or national level, attempting to allocate scarce resources among competing alternatives, while balancing disaster risk management objectives with other development goals, including the long-term sustainability of wealth creation, and the need to adapt to evolving climate risks. Many DRM investments made today will have long-term implications; both because some DRM projects are themselves long-lived, but also because they will influence spatial and economic patterns of development that involve a degree of path dependency (lock-in). DRM investments that take account of climate risk could therefore have potentially important adaptation co-benefits, while avoiding maladaptation risks. Disasters worldwide have caused damages of almost US$200bn annually over the past decade, up from US$50bn in the 1980s. 1 In addition to these direct economic losses, disasters can potentially have longer-term effects on welfare and on economic growth, for example via effects on investment and the provision of basic economic and social infrastructure and, perhaps most importantly from a development perspective, via their direct human impacts2 and indirect effects on the formation of human capital. These impacts and their potentially long-term effects, make disaster risk management (DRM) a first-order consideration for development policy. In spite of this, developing country ministries of finance and planning appear reluctant to invest in such initiatives (see further discussion in Tanner and Rentschler, 2015).3 International aid efforts also tend to prioritize disaster relief and recovery efforts over risk management (see e.g. Kellett and Caravani, 2014). One problem is the relatively narrow framing of standard methods for assessing DRM projects in terms of avoided losses. While avoiding direct human and economic costs is clearly the main objective of any DRM strategy, in many cases there may also be wider ‘development dividends’ associated with initiatives primarily aimed at managing disaster 1 Figures quoted in ‘Investing in Resilience’ World Conference on Disaster Risk Reduction, Sendai 2015 (World Bank GFDRR), available at https://www.gfdrr.org/sites/default/files/publication/Investing-in- Resilience_1.pdf (accessed on 26 August 2015). 2 According to data from MunichRe’s NatCat database, disasters kill some 53,000 people annually worldwide (1980-2008). I avoid the misnomer “natural disasters” given the decisive role of human (in)action in the translation of extreme weather events (and other geo-physical shocks) into disastrous events from a human, social, economic or ecological perspective. 3 The apparent lack of interest in DRM within developing country planning or finance ministries might represent a perfectly rational policy stance, on at least two grounds: First, in a context of scarce resources and many pressing needs, DRM projects – even those with large benefit-cost ratios – may have to compete with other essential investments in things like basic infrastructure, education and health, all of which would also be expected to generate high returns. Second, the benefits of DRM initiatives in terms of avoided losses and emergency expenditures do not necessarily accrue directly to those making the investment decision. This might reflect in part the division of responsibilities across government departments, but could also be the result of a form of moral hazard associated with international disaster relief (or insurance) efforts. 2 risk (Tanner and Rentschler, 2015). Flood protection schemes might for example encourage inward investment by reducing the risk premium associated with a particular location or enabling safe development of locations that are inherently high productivity, but vulnerable to disaster. DRM investments, if designed with evolving climate risk in mind, could also help to avoid costly maladaptation, which might otherwise threaten the sustainability of development. On the other hand, some DRM initiatives will entail development trade-offs. For example zoning restrictions on flood plains might constrain the development of desirable locations. There is also a potential ‘third dividend’ from resilience in the form of co-benefits that accrue from DRM initiatives, regardless of the realization of actual disaster events experienced in any given period. Risk-coping initiatives – e.g. social safety nets and better access to financial services – aimed at reducing the welfare impacts of disasters, might simultaneously promote productive risk-taking in the form of increased entrepreneurship, innovation and diversification of economic activity. Similarly, improvements in the dissemination of risk information and community-based disaster preparedness schemes, for example, can bring additional benefits in the form of increased community cohesion and better state-society relations (Tanner and Rentschler, 2015). Climate change will exacerbate disaster risk for many locations. While there remains deep uncertainty about the precise physical and socioeconomic impacts of future climate change for any particular location, we can say with confidence that climate risk (both extremes and variability) will increase for many developing countries under most climate change scenarios. Climate change therefore reinforces the case for DRM investments (see e.g. SREX, 2012). The evolving nature of climate risk also cannot be ignored for any climate-sensitive investment – especially those with decadal investment horizons. Regardless of current and future mitigation efforts, the global climate is already changing in response to anthropogenic warming. Further warming will continue for some time to come in response to past emissions, given the long-lived nature of some greenhouse gases and the degree of irreversibility in the climate system (Solomon et al. 2007). On current trajectories, that warming may be substantial (see e.g. IEA, 2012; as cited in WB, 2013). We know therefore that some adaptation will be required. DRM investments should therefore be designed with evolving climate risk in mind, and appropriately designed DRM strategies would potentially also confer adaptation co-benefits. In short, appropriate DRM strategies should focus on supporting development paths that are robust to a range of possible climate futures. Such strategies would have the dual benefits of maximizing potential co-benefits of DRM investments for development, while minimizing regret under uncertain climate change. At the aggregate level, the implications of climate change for DRM are unambiguous; climate change reinforces the case for DRM investment, which must also take account of evolving climate risk. At the level of deciding whether or not to adopt a particular DRM project, on the other hand, climate change complicates things. The nature of climate uncertainty is such that the best available forecasts for a particular decision-relevant variable – e.g. rainfall intensity or sea-level rise – at an appropriate spatial and temporal scale, are in many cases characterized by a wide range of possible climate futures, each of which is ‘non-discountable’. In other words, from a policy or decision-maker’s perspective, the possibility that the outcome will be anywhere within that range cannot be disregarded. 3 This degree of uncertainty presents considerable challenges for the standard project evaluation decision-making tools, such as cost benefit analysis. From economic theory, we know that efficient investment decisions – and best outcomes – are achieved through optimum expected utility techniques; weighing discounted expected benefits of the investment against anticipated costs. However, the application of such techniques requires well-defined and well-understood risks, conditions that are hardly met in the case of uncertain climate change. Where optimizing techniques demand precise probabilities in order to calculate expectations, in reality any confident statements about future climate are likely to be of a qualitative nature. The analysis presented here is complementary to recent research that identifies countries that are already facing disaster- (or climate-) related ‘stress’ (see e.g. Mechler et al. 2014). Given the array of policy options available to alleviate that ‘stress’ – from efforts to reduce exposure via better land use planning or zoning laws (and their enforcement) to improvements in risk-coping capacity – the challenge for policy makers is achieving the appropriate balance between the at times competing objectives of DRM, development and adaptation. This task is already difficult without the added complexity of an uncertain climate future. However, that uncertainty need not lead to policy paralysis. This paper includes a sketch of a decision-making framework that attempts to simplify the process of accounting for the deep uncertainty associated with climate projections and the specific characteristics of different DRM policy options. In particular, this framework enables policy makers to identify the particular circumstances under which uncertainty about future climate change becomes critical for their DRM investment decisions. 2. Adding co-benefits into the ‘mix’ of DRM policy options The full spectrum of DRM policies extends beyond the obvious hard infrastructure investments, such as flood barriers, traditionally associated with disaster mitigation. The various DRM policy options might usefully be divided into two distinct categories: On the one hand are attempts to reduce the amplitude of the stimulus or shock (i.e. hazard management). There are numerous alternative (in some cases complementary) elements to this approach, including attempts to reduce exposure, via changes to planning and zoning laws, building codes etc., and via defensive infrastructure. These are the interventions traditionally associated with disaster risk reduction efforts. The second set of options relate to efforts to improve risk-coping capacity (i.e. risk management) – accepting that some shocks will occur, and attempting to minimize the longer-term welfare impacts of those shocks. The latter channel includes better hazard information, early-warning systems, emergency procedures and response systems, as well as economic shock absorbers including insurance, credit and social safety nets. This categorization of policy options is not intended to suggest an ‘either/or’ binary decision for policy makers. Optimal DRM strategies will no doubt involve a ‘mix’ of policies aimed at both reducing exposure and improving risk-coping capacity.4 How does climate change (uncertainty) affect the optimal balance between reducing exposure and increasing resilience? The appropriate balance might depend on both the type and severity of the risk – what is known as the ‘risk-layering’ approach (see e.g. 4 Ideally these will be designed in such a way as to reinforce each other. 4 Hallegatte et al. 2010; Mechler et al. 2014). For example; frequent low-impact events might be mitigated through improvements in basic infrastructure (e.g. drainage systems to prevent urban flooding); the impacts of rarer events might be minimized through attempts to reduce exposure, e.g. by preventing settlement in hazard zones via better public information and zoning or land-use plans; while for the most exceptional, large-scale events, improved infrastructure and zoning may not be sufficient or economical (e.g. since these might constrain development of productive urban locations). Instead, early warning systems and evacuation plans, combined with support for reconstruction and reinvestment (bearing in mind moral hazard risks) can help to avoid the worst human and longer-term economic costs of such events. This risk layering approach identifies appropriate DRM strategies for dealing with different risk profiles. But of course the degree of risk (return period) and exposure to that risk are evolving, both in ways that are outside of local policy makers’ control (climate change) and in ways that are amenable to policy action (exposure and resilience).5 Climate change has important implications not just for the appropriate adaptive responses to various risk layers, but also for efforts (policies) that will partly determine the precise risk layers faced at a given location, and by what populations. The various policy options mentioned above will differ in their potential to convey development and adaptation dividends as well as other co-benefits, in addition to their primary objective of managing disaster risk. It is also important for policy makers to consider how existing development trends and the need to consider long-term adaptation to a changing climate interact with and shape DRM investments and their likely outcomes. The following section identifies cases where the objectives of DRM, development and adaptation coincide. However, there are also likely to be trade-offs between these at times competing objectives. Such trade-offs also need to be made explicit in any DRM investment decision. In some cases, DRM could also risk both constraining (long-term) development and fostering maladaptation. A simple matrix illustrating development and adaptation implications of various DRM initiatives is presented in Figure 1. The additional development and adaptation dividends identified here provide further motivation for DRM investments, over and above their potential to reduce losses from disasters. Policy makers apparently tend to perceive DRM investments as representing sunk costs, with little benefit in the case that disaster does not strike (Tanner and Rentschler, 2015). The framework developed here emphasizes that DRM investments could also confer ‘sunk benefits’ – for example where DRM initiatives could help to avoid locking-in future exposure to climate risk, avoiding potentially costly maladaptation. In such cases, the uncertainty associated with climate change, far from justifying a ‘wait and see’ approach, instead provides further motivation for early intervention to manage risk and build risk coping capacity (see further discussion in Section 4). 5 It is important to note that while the current paper focuses on uncertainty related to climate change, DRM investment decisions will also face uncertainties in relation to the other determinants of overall disaster risk – i.e. exposure and vulnerability. Accounting for the potential interactions of these multiple risk determinants is crucial to the formulation of optimal DRM policies – as argued elsewhere in this paper. 5 Figure 1: The figure illustrates potential overlaps between development and adaptation dividends (and trade-offs) of various DRM strategies. Development dividends and trade-offs The debate over DRM investment tends to focus on avoiding losses. But from a development perspective some losses may cause more harm than others. Large-scale monetary (or asset) losses do not necessarily translate into important consequences for welfare or economic development (Hallegatte, 2014). In theory, agents (households, businesses and government) should be able to trade risk through insurance and financial markets (Auffret, 2003). While these coping mechanisms tend to be well developed in wealthier countries, underdeveloped financial markets and the absence of other economic ‘shock absorbers’ (Loayza et al. 2007) in many poorer countries means they remain vulnerable to shocks and setbacks of various kinds.6 The human consequences of disasters also have potentially important longer-term effects on development. Disasters clearly have substantial direct effects on health (morbidity) and mortality – especially in developing countries – with potentially large impacts on subsequent development paths. 7 While monetary losses can be recovered through insurance and emergency reconstruction funds, the development (and human) impacts of disasters represent permanent losses to welfare. The first priority of any DRM strategy 6 Formal risk transfer mechanisms might also be lacking even in developed countries – for example in the case of flood insurance in the UK. 7 The health impacts of extreme weather (droughts, floods etc.) are discussed in detail in Hales et al. (2003). The Lancet has characterized climate change as “the biggest global health threat of the 21st century” (Lancet, 2009). An extensive literature looks at the interaction of climate, health and income (e.g. Tol et al. 2007; Pascual et al. 2008; Strulik, 2008; and Tang et al. 2009). 6 should be to minimize direct human and longer-term welfare and growth impacts.8 Given the potential for disasters to disrupt longer-term development trajectories – particularly in poorer countries – all efforts to reduce the long-run welfare and growth effects of disasters could be considered to generate a development dividend (Tanner and Rentschler, 2015).9 But risk reduction should not be pursued for its own sake (Hallegatte, 2013). Indeed, one challenge for DRM is that some development trends involve the accumulation of risk. It is for this reason that risk management is now emphasized over risk reduction. Inevitably there will be trade-offs between some development and DRM objectives. Obvious examples include trends such as urbanization and the accumulation of people and assets in at-risk or vulnerable locations, such as on coasts. Restricting such development – for example through land-use planning and zoning restrictions – might reduce exposure to hazards, but might also constrain the exploitation of potentially high productivity locations. If, on the other hand, further development of risky locations is allowed, the question then arises; at what point does increasing exposure to disaster risk threaten development itself?10 Hard infrastructure investments, such as flood barriers, might confer a development dividend by reducing the risk premium associated with a particular location or enabling safe development of locations that are inherently high productivity, but vulnerable to disaster, resulting in greater inward investment. However, for developing countries, expensive protective infrastructure may not be the most efficient use of scarce resources. Instead, cheaper but possibly more politically challenging improvements in safety could be achieved through better building regulations and settlement policies (and their enforcement), improved dissemination of information, early warning systems, evacuation and emergency planning and training etc. Such initiatives would have the benefit of prioritizing the protection of human life over economic assets (thus helping to minimize welfare impacts), while investing in institutional capacity and community preparedness initiatives might also be expected to have positive spillovers or co-benefits for economic development via the growth enhancing effects of improved institutions and governance capacity (see Tanner and Rentschler, 2015). Economic development also necessarily involves risk-taking at the individual household or firm level, in the form of entrepreneurial activities, including experimenting with new technologies, innovation, diversification away from traditional modes of production etc. The inability of the poor to cope efficiently with risk – and therefore to take on these productive risks – represents one of the essential problems of development (see e.g. Bryan 8 In spite of their large-scale impacts, evidence on the wider economic effects of disasters, particularly their indirect and longer-term effects, has been relatively scarce (e.g. Cavallo and Noy, 2009). One recent study on the growth effects of tropical cyclones finds significant impacts on long-term economic growth, comparable with the effects of other major shocks such as civil wars and banking crises (Hsiang and Jina, 2014). In fact the effects of cyclones on growth are observed up to 20 years after the event – twice as long as the other shocks mentioned. 9 Although it’s important to note that DRM investments generally, and especially defensive ones, represent a cost to society (a constraint on welfare maximization), and their opportunity costs – the best alternative use of those resources – should not be overlooked (see e.g. Tol and Leek, 1999). 10 The longer-term sustainability of recent economic growth in many developing countries is far from guaranteed. Growth collapses are all too common. In fact, economic ‘takeoff’ – a discrete transition from low or stagnant growth to sustained positive growth – is quite rare (Easterly, 2006). Instead, most growth accelerations are not sustained, with growth spurts followed by collapse more common in poorer countries (Hausmann et al. 2005). This pattern of growth and decline has also been identified in the historical data for pre-industrial Europe (Broadberry and Gardner, 2013). A comparison with these historical episodes suggests that many African countries today remain below the income thresholds associated with sustained economic growth (in the European experience), and therefore remain vulnerable to setbacks and reversals. 7 et al. 2014). Financially constrained households struggle to cope with risk, employing inefficient coping mechanisms both ex-ante and ex-post (e.g. Mobarak and Rosenzweig, 2013). Ex-ante, financially constrained households “seek insurance by investing in safe but less productive assets” (Acemoglu & Zilibotti, 1997, p.710). In other words, they tend to trade off average returns against income variability, for example by holding low-return liquid assets (Rosenzweig and Binswanger, 1993) or diversifying productive activities, sacrificing the efficiency gains from specialisation (Collier et al. 2008; Dercon, 2012). “As a result, poor countries will endogenously have lower productivity” (Acemoglu & Zilibotti, 1997, p.710). Ex-post coping mechanisms – following an income shock – for credit-constrained households include; drawing down savings (Paxson, 1992); selling productive assets (Rosenzweig and Wolpin, 1993; Scoones and Chibudu, 1996); drawing on child labour (Jacoby and Skoufias, 1997); or engaging in expensive informal borrowing (Benson, 1997; Banerjee and Duflo, 2011). 11 These ex-post coping strategies will similarly have consequences for the sustainability of productive (longer-term) investments, including in children’s education, and as a result tend to be associated with increased poverty, lower investment and lower growth (Elbers et al. 2007). A lack of economic infrastructure (including access to financial services) is compounded in many developing countries by weak social safety nets and institutional or governance barriers that limit opportunities for escaping from poverty (see e.g. Collier et al. 2008; Dercon, 2012). The inability of the poor to cope efficiently with risk, and the knock on effects for both the productivity and sustainability of their investments, creates the conditions for persistent poverty or poverty trap dynamics (Azariadis and Stachurski, 2005). These standard concerns from the development literature in relation to coping with risk are further compounded by the expectation that climate change will exacerbate risk and income volatility, particularly for poorer developing countries (see e.g. Samson et al. 2011; and WB, 2013; as discussed further in the next section). DRM investments help to reduce the ‘background risk’ of disaster faced by households, firms and investors (Tanner and Rentschler, 2015). The reduction of this risk, which constrains investment and ultimately development, generates a ‘development dividend’ from DRM. The promotion of risk-coping strategies – enabling greater productive risk taking, while also reducing the welfare impacts of disasters – especially amongst the poor who tend to cope inefficiently with risk, represents a potential ‘third dividend’ from DRM. DRM strategies that focus on improvements in risk-coping capacity will also promote the adaptive capacity and autonomy of the poor (e.g. by increasing access to financial services and basic infrastructure etc.). Adaptation should not just be about responding to changes as they occur, but rather efficient adaptation strategies will involve preparing to adapt (Oreskes et al. 2010), and in particular building adaptive capacity so that individuals, households, communities and nations will have the resources and (economic) flexibility necessary to minimize welfare losses (and maximize welfare gains) from future climate change. Part of such strategies will likely overlap with DRM objectives, potentially leading to improved resilience to disaster and weather risk, greater capacity amongst the poor for productive risk-taking as well as capacity building for future adaptation. Ultimately such strategies reduce the likelihood of a vicious cycle between risk and poverty. 11For further discussion on the role of finance in coping with the impacts of climate change on the poor, see the review by Skoufias et al. (2011). 8 Adaptation dividends and trade-offs Regardless of current and future mitigation efforts, the global climate is already changing in response to anthropogenic warming. Further warming will continue for some time to come in response to past emissions, given the long-lived nature of some greenhouse gases and the degree of irreversibility in the climate system (Solomon et al. 2007). On current trajectories, that warming may be substantial (see e.g. IEA, 2012; as cited in WB, 2013). We know therefore that some adaptation will be required.12 Many DRM initiatives will involve efforts to cope with climate risk. Adaptation to future climate change therefore represents a natural ‘dividend’ of most DRM investments. In particular, initiatives undertaken now to promote sustainable development paths, or to avoid locking-in further vulnerabilities, offer potentially large adaptation dividends. Existing development trends, such as migration to coastal, urban areas, and settlement in vulnerable locations more generally, increase exposure to extreme weather events. To the extent that these trends are irreversible – for example because of the strong degree of path dependency in urban locations (see e.g. Davis and Weinstein, 2002; Michaels and Rauch, 2013) – delaying action to manage those risks now could lock in greater future costs to society. Adaptation dividends, particularly in the form of attempts to avoid locking-in unsustainable patterns of development, provide a strong motivation for early action on DRM. Delaying adaptive DRM investments likely incurs the opportunity cost of missed opportunities for adaptation. Once a new settlement is established in a vulnerable location, for example, it is difficult for that pattern of spatial development to be reversed. In general, adapting to climate change involves trading off opportunities to exploit today’s climatic conditions against the ability to exploit (anticipated) future conditions (Millner, 2012); in other words, sacrificing some resources today in the expectation of improving future welfare. There are therefore likely to be some trade-offs between adaptation and development co-benefits of DRM.13 In a developing country context, one might expect the balance of priorities (justifiably) to favor today’s challenges and opportunities over (uncertain) future ones. However, that does not imply that adaptation co-benefits (and trade-offs) are not relevant factors in the design of appropriate DRM strategies. Increasing the resilience of rural livelihoods – and in the process avoiding weather risk translating into disasters – involves addressing existing vulnerabilities and improving the capacity of the poor to cope with existing, reasonably well known and understood risks. As discussed earlier, such initiatives might generate additional development and adaptation dividends above and beyond their primary purpose of reducing the direct losses associated with disasters. The main concern with such investments is to avoid moral hazard or maladaptation in the form of fostering unsustainable modes of production or constraining the opportunities for (and inherent dynamics of) transformative development. The literature on the resilience of rural livelihoods has tended to focus on in-situ forms of adaptation, such as increasing local food security and self-sufficiency (Dercon, 2012). 12 It is also worth noting that mitigation and adaptation efforts are not substitutes, but are instead complements. While future climate change is uncertain, we know that weaker mitigation effort increases the likelihood of changes for which adaptation will be increasingly difficult (costly, painful and disruptive) if not impossible (Oreskes et al. 2010). 13 One exception to this might be initiatives aimed at increasing the adaptive capacity and autonomy of the poor – as mentioned above. 9 While such risk management initiatives would appear to align with adaptation and development objectives, there are risks. For example, Dercon (2012) points out that “many drought-resistant crops have low returns, leading to more security but also less poverty reduction (Morduch, 1995; Dercon 1998).” Similarly, in-situ adaptation may represent maladaptation where existing activities and locations are already marginal and likely to face deteriorating climatic conditions. Investments in agriculture, such as new seed varieties or irrigation infrastructure might improve resilience to weather risk in the short-term, but could risk locking in forms of production that eventually become unsustainable under climate change (e.g. over-use of ground water). Sensible DRM strategies – taking account of the need to consider long-run adaptation to a changing climate – should also consider existing development trends; whether these are likely to be sustainable under a range of possible climate futures; but also how climate change may affect (amplify or diminish) these existing trends. A good example is rural- urban migration. While much of this migration is driven by the ‘pull’ of economic opportunities in urban areas, climate change might also reinforce the ‘push’ of limited economic opportunities and precarious livelihoods in rural areas (see e.g. Barrios et al. 2006; Henderson et al. 2014). On the other hand, climate change also poses significant threats to urban areas in the form of increased risk of heat-waves and flooding. In the absence of adaptation planning and DRM strategies, such threats could lead to the emergence of urban ‘push’ factors (i.e. a flight from vulnerable urban locations) and the subsequent loss of important development opportunities inherent in high productivity urban locations. Climate uncertainty reinforces the need for flexibility – as discussed in greater detail in the subsequent sections. DRM policies should therefore aim to increase the autonomy of poor and vulnerable groups – e.g. by enabling people to move on their own terms (which means moving in some cases, and remaining in place in others).14 This is likely best achieved through standard development initiatives such as improvements in access to finance, access to markets (including labor markets), investments in health and education and basic economic infrastructure (transport, energy and sanitation). In a rural context then, DRM and development objectives would appear to be reasonably well aligned. DRM considerations, especially under uncertain climate change, reinforce the case for policies that seek to improve the economic flexibility of the rural poor (generating both development dividends and additional co-benefits), but also emphasize the need to consider the long-term sustainability of rural development initiatives. In an urban context, on the other hand, there is a greater tension between the objectives of DRM and development. Existing development trends (and also potentially rural DRM strategies) will tend to exacerbate urban disaster risk and hazard exposure. Appropriate DRM strategies therefore need to consider both how to support economic transformations (which are an integral part of long term development)15, but also how to manage the additional risks of large-scale migration to urban areas, which are themselves often vulnerable to disasters, especially urban flooding (Hallegatte et al. 2010). Defensive investments, such as the construction of flood barriers, may reduce the incentives to adapt – a form of moral hazard – or even lead to maladaptation behaviors (see e.g. Collier et al. 2008). The presence of a flood barrier will presumably lead either de jure or de facto to greater development of the protected (but risky) area. In the case of failure, 14 For example, see Government Office for Science [UK] (2011). 15 See e.g. Dercon, 2012; Lewis, 1954; Harris and Todaro, 1970. 10 losses would then be exacerbated relative to some baseline scenario. Many hard infrastructure DRM projects (e.g. flood barriers) also face the challenge of providing either complete protection or complete failure – i.e. binary outcomes. This places an even greater burden on designers and policy makers to get the level of protection right, and raises again the challenge of dealing with uncertain future risks, due to climate change. Characterizing this uncertainty, and how policy makers should cope with it, is the subject of the remaining sections. 3. A changing climate for development Many developing countries already face challenging climatic conditions; in general they are hotter and experience more variable rainfall patterns than their richer counterparts (Stern, 2007). It is widely anticipated that the effects of future climate change will exacerbate these climatic challenges in poorer parts of the world (Solomon et al. 2007; IPCC 2013/2014; World Bank, 2010, 2013). The expectation that climate change will have its most damaging effects in poorer countries is based partly on projections of where future changes in climate will be most negative from a socioeconomic perspective (e.g. Samson et al. 2011; World Bank, 2013), and partly on the observation that poorer countries are more vulnerable to changing climatic conditions, given their exposure (existing climate and reliance on agricultural output) and lack of adaptive capacity (see e.g. Fankhauser and McDermott, 2014). The economic and development impacts of gradual changes in climate can be illustrated using historical data. For example, changes in moisture availability have been shown to have notable effects on agricultural productivity, rural-urban migration patterns (Henderson et al. 2014) and economic growth (Barrios et al. 2010; Brown et al. 2013). Observed declines in moisture availability to date have been most pronounced in already arid areas, exacerbating existing vulnerabilities (Henderson et al. 2014). Anthropogenic forcing is expected to result in global warming (i.e. increases in average temperatures) of anywhere between 2°C and as much as 5-6°C, by 2100, under different emissions scenarios.16 Such warming, however, will not be distributed evenly around the planet. Some locations will experience significantly more warming for any given global average change. For example, land warms faster than oceans. Similarly, there are well- understood physical mechanisms, which indicate that warming at high latitudes will be greater than at lower latitudes. Thus, for any given target for the increase in global mean temperatures (e.g. +2 degrees), the implication is that most if not all land areas would warm by more than this, and higher latitude land areas by substantially more. However, it is not just changes in mean temperatures or precipitation that matter for development. Climate change represents a change in the distribution of future weather (Daron and Stainforth, 2013); investment decisions and economic activity more generally will be sensitive to more than the mean of that distribution (Stainforth et al. 2007a). From a DRM perspective, it is the frequency of extremes that is most relevant, since extreme temperatures, precipitation (both abundance and scarcity) and winds are generally the 16For example, the IEA’s 2012 assessment suggests that in the absence of further mitigation there is a 40 percent chance of warming exceeding 4°C and a 10 percent chance of it exceeding 5°C by 2100 (as cited in WB, 2013). 11 triggers of climatic disasters.17 Making projections about the future distribution (frequency) of extreme weather events - storms (tropical cyclones), floods, droughts, heat-waves, cold- waves etc. – at a scale relevant to policy makers and investment decisions is even more challenging than predicting average changes (see e.g. SREX, 2012).18 Again, however, we can make some qualitative predictions, based on physical principles. Higher average temperatures would change the shape of local temperature distributions (see e.g. Stainforth et al. 2013). A first order expectation is that the shift towards higher average temperatures would involve more frequent extremes of heat and less frequent extremes of cold. Temperatures currently considered extremely hot may become the norm. The frequency of heavy rainfall events or the proportion of total rainfall from heavy falls is also expected to increase for many areas (SREX, 2012). This follows from the basic principle that a warmer atmosphere can hold more water. More intense rainfall episodes might lead to an increase in flooding risk. The intensity of rainfall events might even increase in locations where total rainfall is anticipated to decline, leading to higher variability (SREX, 2012). Drought intensity is also expected to increase, for some areas and seasons, although there is only medium confidence in these expectations (SREX, 2012). For tropical cyclones there is an expectation for fewer cyclones, but maximum wind speed is likely to increase, at least in some ocean basins – potentially leading to greater loss or damages (SREX, 2012). Higher global temperatures could also lead to increased flooding risk via sea-level rise, which is (obviously) particularly threatening to low-lying, coastal or deltaic areas, which happen to include some of the most densely populated regions of the planet, many of them in developing countries (e.g. Bangladesh). These qualitative assessments about what climate change will imply for future weather provide useful guidance for policy-makers in a fairly general sense. In particular, it is confidently expected that climate risk (both variability and extremes) will increase for many developing countries under most climate change scenarios. This expectation therefore reinforces the case for DRM investment, and highlights adaptation to climate change as an important potential co-benefit of DRM initiatives. Translating anticipated global changes into projections at the kind of temporal and spatial scales that are relevant for project evaluation type decisions is considerably more challenging. The problem of uncertainty in climate change projections, and the implications for DRM investment strategies are discussed in greater detail in the next sections. The focus here is on uncertainty related to climate change. However, it should be noted of course that DRM investments are also subject to uncertainty in relation to future exposure and vulnerability – both of which are at least partly dependent on development trajectories. As argued elsewhere in this paper, DRM strategies also need to take account of the interaction of development trends with (evolving) climate risks. Without anthropogenic forcing, weather is often considered to be chaotic; “under changing concentrations of atmospheric GHGs, the behaviour is not chaotic but pandemonium (Spiegel 1987)” (Stainforth et al. 2007a, p.2147). Making (long-term, in some cases irreversible) investment decisions under climate change presents the challenge of dealing with deep uncertainty. Even a ‘perfect’ climate model would produce a 17 The translation of an extreme weather event into a disaster depends on socioeconomic factors (see e.g. Sen, 1983; Tol and Leek, 1999; Kahn, 2005; and discussion elsewhere in this document). 18 For example the SREX report (2012, p.130) states “it is not possible in this chapter to provide assessments of projected changes in extremes at spatial scales smaller than for large regions”, while “for many regions of the world, no downscaled information exists at all and regional projections rely only on information from GCMs” (p.131). 12 distribution of possible future weather trajectories, only one of which will ever be realized. If that distribution contains a large range of possible values for decision-relevant variables (e.g. precipitation quantity and timing, temperature averages and extremes, wind speeds etc.), simply taking expectations (i.e. the mean) of that distribution, as would be common in economic analyses (or standard project evaluation techniques such as cost-benefit analysis), may be seriously misleading, particularly given the possibility of important thresholds or tipping points which may lie within the range of uncertainty (Kemp, 2005). Of course, there is no such thing as a ‘perfect’ climate model. Uncertainties in relation to modeling future climate change derive from several distinct sources (Stainforth et al. 2007a); (anthropogenic) forcing, initial conditions, and model imperfections (both model uncertainty and model inadequacy). These uncertainties are challenging enough for global climate models, which are relatively well understood. Global models can provide information relevant to mitigation decisions. However, adaptive (and DRM) investment decisions require climate information at a much finer spatial resolution. Attempts to ‘down-scale’ global climate projections to spatial and temporal scales relevant for adaptive investment decisions involve additional layers of uncertainty associated with “local physics, topography, and an incomplete understanding of how downscaling techniques interact with uncertainties already present in GCMs” (Millner, 2012, p.144). Uncertainty – specifically the disagreement between climate models – has been shown to increase as the spatial scale of climate projections is reduced (Mason and Knutti, 2011; as discussed in Heal and Millner, 2014). One common approach to dealing with uncertainty over future climate change has been to produce a range of climate change projections, for example based on ‘ensembles’ of different climate models, with varying initial conditions and forcings. The outputs of these various model runs are often combined into a single probability density function (PDF), by applying a weighting scheme (Tebaldi and Knutti, 2007). This single PDF might be thought to characterize climate risk for a given variable, location, time period etc., and the standard optimizing techniques of economic analysis could then be applied to make investment decisions, taking account of risk-return trade-offs. However, such weighted combinations of model outputs are likely to be highly misleading in that they imply a degree of confidence that is not justified by the underlying assumptions and the known inadequacy of our current models. In particular, Stainforth et al. (2007a, p.2155) emphasize that “the lack of any ability to produce useful model weights, and to even define the space of possible models, rules out the possibility of producing meaningful PDFs for future climate based simply on combining the results from multi-model or perturbed physics ensembles” (see also Smith, 2007; and Tebaldi and Knutti, 2007; cited in Millner, 2012). 19 The inability to calculate probabilities with any confidence renders standard economic approaches to project evaluation (e.g. cost benefit analysis) inadequate. Where DRM investments are sensitive to climate uncertainty, cost-benefit analyses need to be supplemented by additional screening devices, as discussed further in the next section. Where standard economic analyses may demand precise probabilities, in reality any confident statements about future climate are likely to be of a qualitative nature, and reliant 19 Furthermore, they caution against holding out hope of such uncertainty being reduced in the near future; “Perturbed physics experiments have demonstrated the large range of responses possible from the current class of models. There is no reason to expect this range to be substantially smaller in the next generation of models; increased physical realism may well increase it.” (Stainforth et al. 2007a, p.2159). 13 on a number of significant assumptions.20 However, Stainforth et al. (2007a) note that models can still provide insight, and even qualitative guidance can be valuable in informing adaptation decisions. In interpreting the output from climate models, these authors encourage users to view them as providing “a range of possibilities which need to be considered” (Stainforth et al. 2007a, p.2159). They characterize the best information that climate models or ensembles can currently provide as a “lower bound on the maximum range of uncertainty”, or what they refer to as the ‘climate envelope’ (Stainforth et al. 2007a, p.2155).21 They also stress that this range is ‘non-discountable’ in the sense that “we should not disregard the possibility that the response could be anywhere within the envelope”. This last point is crucial for DRM investment decisions. From a policy or decision-maker’s perspective, the possibility that the outcome will be anywhere within that range cannot be disregarded. Standard evaluation techniques, such as cost-benefit analysis, where risks are discounted according to known or expected probability of their occurrence over a particular time period, might therefore produce very misleading policy recommendations. Stainforth et al. (2007b) instead advise using the boundaries of the ‘climate envelope’ as an initial screening for adaptive investment decisions. For example, for a decision-relevant climate variable, one should consider the maximum and minimum projected values for that variable and whether either of these ‘extremes’ might alter the planned investment. Would the most benign future climate scenario render the project unnecessary, or economically unviable in terms of avoided losses relative to costs? Would the ‘worst case’ future climate scenario render the project inadequate, in terms of providing the desired level of protection? More generally, risk assessments need to consider many combinations of values for key parameters, including those related to climate, exposure and vulnerabilities. The nature of uncertainty in relation to climate projections means that the range of climate-related values in any such exercise needs to include the boundaries at either end of the ‘climate envelope’. We return to this idea of using the ‘climate envelope’ as a screening device as part of a heuristic guide to DRM investments under uncertain climate change in the next section. 4. A decision-framework for DRM investment under climate uncertainty How should policy makers treat the highly uncertain information available from climate models and associated analyses of socioeconomic impacts? One seemingly reasonable approach would be to plan for a central (most likely) scenario. However, such an approach is problematic (and risky) on a number of levels. For one thing, in a non-linear climate system, taking a central or mean expectation might under-represent the possibility of a 20 Translating physical changes in the distribution of future weather into socioeconomic impacts adds a further layer of uncertainty (Heal and Millner, 2014).20 For example, combining a full range of climate models with impact estimates, Burke et al (2011) find the 95% confidence intervals around outcomes are increased by a factor of five. Henderson et al. (2014) discuss the literature that attempts to project the impacts of anticipated climate change on African agriculture. They find that most studies predict yield losses for staple crops of between 8-15 percent by 2050 and 20-47 percent by 2090, depending on the climate scenario, although under more optimistic scenarios, and depending on the assumed degree of successful adaptation, some studies find more modest or even positive impacts. 21 “‘Lower bound’ because further uncertainty exploration is likely to increase it; ‘Maximum range of uncertainty’ because methods to assess a model’s ability to inform us about real-world variables, e.g. shadowing techniques (Judd et al. 2004), could potentially constrain the ensemble and reduce the range.” (Stainforth et al. 2007a). 14 threshold or tipping point being breached, resulting in some extreme or catastrophic scenario. 22 One must also consider whether projects approved under a ‘most likely’ scenario risk locking in development paths that would be vulnerable to more extreme scenarios (or equally, whether such projects would ultimately prove to be unnecessary in the case that realized changes are less than anticipated). A classic example of this dilemma is flood defenses. They must be built to some specification. But what is a reasonable (or optimal) level of protection? Deciding whether or not to invest in expensive infrastructure projects requires some form of project evaluation. The standard approach to project evaluation for investment decisions is to apply a cost-benefit analysis that compares the (discounted) expected value of future benefits (e.g. the value of avoided losses) with anticipated costs of the investment. With well-defined and well-understood risks and uncertainties, optimum expected utility techniques demonstrably produce the best outcomes (Lempert and Collins, 2007). However, under uncertain climate change, these conditions may not hold (Arrow et al., 1996; Jaeger et al., 1998; Dessai et al., 2004; all as cited in Lempert and Collins, 2007).23 Furthermore, it is now well established in the environmental economics literature that investments related to environmental problems generally have important additional characteristics that are neglected by these standard evaluation frameworks, in particular; uncertainty over future costs and benefits of the project; irreversibilities once the policy or investment has been approved; and the option of delaying action until more information becomes available (see e.g. Pindyck 2002). Alternative decision rules, for example based on the principles of minimizing regrets (see e.g. Heal and Millner, 2014) or ‘robust’ decision making (e.g. Lempert and Collins, 2007; Hallegatte, 2009) offer the promise of formalized quantitative analysis under deep uncertainty. However, such techniques require substantial analytical or computational resources, for example in order to calculate expected costs and benefits under a potentially large numbers of ‘plausible’ climate futures (Lempert and Collins, 2007; Hallegatte et al. 2010). Given that decision-making capacity is itself a scarce resource in many developing countries, this section offers a simpler decision-making framework, incorporating Stainforth et al.’s idea of screening investment decisions against the boundaries (maximum and minimum) of the ‘climate envelope’. The deep uncertainty over future climate change, and its physical and socioeconomic impacts, would appear to underscore the need for greater flexibility to be incorporated in the design of DRM strategies. While the need for flexibility in adapting to climate change is incontrovertible (see e.g. Fankhauser et al. 1999), its relevance for DRM investments will be greater in some circumstances (locations and projects) than others, and in particular will depend on the climate sensitivity of the proposed investment and the range of possible climate futures (the breadth of the ‘climate envelope’) for decision-relevant variables. So for example, where screening the proposed investment against the climate envelope indicates that even the most extreme climate change projections (at either end of the 22 The possibility of ‘extreme’ outcomes has increasingly been viewed as an important element in appropriate climate policy design (e.g. Millner, 2013; Weitzman, 2009; as cited in Heal and Millner, 2014; see also Cavallo and Noy, 2009; and Hallegatte et al. 2007). While the emphasis in those papers is predominantly on physical tipping points, there may also be important adaptive (socioeconomic) tipping points or thresholds, as discussed in Swart et al. (2013). 23 The economic calculus requires estimation of expected benefits of the investment (to be weighed against anticipated costs). In the absence of known probabilities, expectations cannot be calculated. The inability to produce probabilities from climate projections is therefore problematic for optimizing techniques. 15 spectrum) would not alter the investment decision, building in flexibility may be redundant, and would likely incur unnecessary additional costs. Uncertainty therefore has qualitatively different implications for different types of DRM investment projects. Figure 2 and the following discussion provide a heuristic decision framework that enables policy makers to determine the extent to which issues related to investing under uncertainty should influence DRM investment decisions. Figure 2: Outline of decision-framework for determining the role of uncertainty in project evaluation decisions. In the absence of any uncertainty in relation to future climate, DRM investments should be guided by standard cost-benefit type analysis – comparing the expected benefits of the investment in terms of avoided losses with anticipated costs and the best alternative use of the funds (standard investment opportunity cost) – supplemented by due consideration of potential development and adaptation dividends and trade-offs, as discussed in section 2. Under uncertain future climate change, the investment decision becomes more complex. However, that uncertainty is only relevant under certain conditions. A full illustration of how climate uncertainty can be incorporated into a project evaluation framework is provided in Figure 3. First, we should consider if the investment itself is sensitive to changes in climate risk. In other words, are the expected benefits, in terms of avoided losses, dependent on the severity of some climate-related hazard? Flood defenses and zoning or planning restrictions are clearly highly climate sensitive in this sense. On the other hand, some DRM initiatives such as disaster preparedness, emergency planning procedures, and various risk-coping strategies are likely to be less sensitive to changing 16 climate risk and therefore require less detailed consideration of climate uncertainty before they can be safely adopted or rejected (postponed). Figure 3: Full ‘decision tree’ for project evaluation under climate uncertainty If the proposed project is climate-sensitive, and assuming that projections for decision- relevant variables are available at the appropriate temporal and spatial scales, the project should be screened against the full range of possible climate futures.24 Such a screening process in practice involves just two additional calculations – one at each of the boundaries (max and min) of the ‘climate envelope’ for the decision-relevant variable. If the project survives the full range of the climate envelope then it can be accepted without further consideration of issues related to uncertainty. Rejection of the project at either boundary of the climate envelope will have distinct implications for that investment decision. If under the most benign climate future the project is no longer required (i.e. becomes uneconomical), then the risk of redundancy needs to be considered and alternative uses of scarce resources may be preferred. However, such a project may still be worth pursuing, particularly where there are (large) anticipated development or adaptation dividends (or other social, environmental or economic co- benefits – see Tanner and Rentschler, 2015). On the other hand, if under a ‘worst case’ climate scenario the project would be rendered inadequate in terms of the level of protection provided, a more complete consideration of uncertainty and how it relates to the specific characteristics of the proposed investment would be required. This would also apply to the situation of a climate-sensitive investment 24In the absence of such projections, further consideration of issues related to uncertainty would be required, as discussed in further detail below. 17 project where no decision-relevant projections are available. A first-order consideration for projects that fail the ‘worst case’ climate scenario stress test (or where decision-relevant projections are unavailable) would be the risk of locking-in further vulnerability, such as in the case of flood defenses, where the concentration of population and economic assets exposed might be increased as a result of the protection provided (moral hazard risk). In other words, could the proposed investment actually make things worse under some climate futures? If so, the project is in danger of exacerbating (rather than mitigating) disaster risk. Such scenarios are obviously highly undesirable and would therefore suggest a rejection of the proposal. If on the other hand, these (moral hazard) risks are not present the project could be allowed to proceed, provided it meets two further criteria. First, could the project be scaled in response to evolving risk (or risk evaluations)? In other words, is there flexibility in the design to increase or decrease the level of protection over time? Secondly, are there partial benefits from different degrees of intervention, or does the project need to achieve a certain scale (or be fully completed) before any benefits are conferred (time-to-build concerns – see e.g. Millner, 2012; Roberts and Weitzman, 1981; Majd and Pindyck, 1987).25 If the proposed project has both partial benefits (i.e. doesn’t face ‘time-to-build’ concerns) and flexibility in design (i.e. the level of protection is scalable in response to changing risk profiles) then investment might still be justified, even where the project has failed the ‘worst case’ climate scenario stress test. If these characteristics are not (both) present, however, this suggests the project should be rejected (or postponed, pending new information). The optimal timing of investment and the value of waiting for new information The literature on investing under uncertainty emphasizes that where investments are at least partly ‘irreversible’, there is an opportunity cost to investing today in the form of the foregone opportunity to wait and learn from new information (e.g. Dixit and Pindyck, 1994; Pindyck, 2002 etc.). The investment decision is then not just whether or not to invest, but also whether it is optimal to invest today or wait for new information. The potential value of waiting for new information will depend on the degree of irreversibility of the investment and the associated value of avoided ‘regret’ in the case of its adoption. Uncertainty does not automatically imply a ‘wait and see’ approach to DRM. In the first instance, the value of waiting for new information (in terms of avoided regret) must be weighed against the likelihood of its timely arrival, which is not guaranteed in the case of climate prediction. Under climate change, the prospects for significant improvements in our ability to make reliable forecasts at decision-relevant scales appear pretty dim (Stainforth et al. 2007a; Masson and Knutti, 2011; Knutti and Sedlacek, 2013; Heal and Millner, 2014). 26 The various sources of uncertainty in climate projections become relevant at different time scales and also differ in the extent to which we can expect improvements in our 25 Some adaptive investment projects – e.g. large infrastructure projects or research and development – confer no benefits until completion. Such investments require greater certainty over their robustness to future conditions before approval should be granted, given the sunk costs (and delayed benefits) associated with their adoption today. 26 The dim prospects of improvement in the predictive accuracy or precision of climate models need not suggest that efforts at improving information should be overlooked. In particular, large improvements in risk awareness could be made based on better information about current exposure and risk patterns. 18 understanding and predictive capacity (as discussed in Millner, 2012). Over shorter time horizons, total prediction uncertainty is dominated by internal variability (initial conditions) and model imperfections, with internal variability increasingly important at smaller spatial scales and shorter time horizons. The latter, in particular, may be amenable to gradual improvements through a better understanding of initial conditions, based on improvements in observations (Smith et al. 2007). The uncertainty over longer-term projections, on the other hand, is dominated by uncertainty over future emissions, which is essentially unknowable (Millner, 2012).27 Adaptive investments with high adjustment costs (i.e. those that involve longer-term commitments, less flexibility or elements of irreversibility) require the greatest precision in (longer-term) forecasts, before one can safely (unreservedly) recommend adoption or rejection (postponement) of that investment (Millner, 2012). Under uncertain climate change, this finding would suggest that DRM investments favor more flexible initiatives over those with large sunk costs and elements of irreversibility. However, this analysis does not suggest that hard defensive infrastructure investments should never be undertaken, but rather that the uncertainty over future climate change increases the risks associated with such projects and therefore places a greater burden of proof on their advocates to ensure that the benefits in the absence of climate change are sufficiently large, and unlikely to be reversed under a large range of possible climate futures. There are two kinds of irreversibility and associated potential for ‘regret’ related to uncertain DRM investment decisions (Pindyck, 2002). These work in opposite directions. One is the sunk cost associated with an investment – an obvious example is the case of hard infrastructure projects (once built they are essentially fixed), but sunk costs may also be in the form of political constraints making a DRM policy (e.g. land zoning) difficult to reverse once implemented. These sunk costs make adopting a policy today more risky than would be implied by a standard cost-benefit analysis. For example, the construction of flood defenses might encourage greater investment and higher density settlement in the protected area. In the case that flood risk turns out to be worse than anticipated, and the defenses are breached, the costs to society would then be greater than in the absence of that investment (moral hazard risk). The second type of irreversibility is a ‘sunk benefit’ or negative opportunity cost, of adopting the policy now rather than waiting; this relates to foregone adaptive opportunities (Fankhauser et al. 1999), which are missed while one waits for new information. Such adaptive opportunities might be most strongly associated with initiatives that attempt to guide development trends towards more sustainable trajectories – as discussed earlier. These sunk benefits (or adaptation dividends) strengthen the case for adopting a policy today, relative to what would be implied by a standard cost-benefit analysis. DRM policies that reduce future exposure to climate risk convey a future ‘development dividend’ by improving the resilience of development gains. Even if we were to adopt a ‘safety first’ policy (precautionary principle), this would still not necessarily favor a ‘wait and see’ approach to DRM policy. Inaction (postponement of the 27 Some uncertainty over impacts will be reduced through better understanding of the historical links between climatic variability and socioeconomic response. However, unmitigated climate change will likely lead to conditions not seen in human history, potentially making historical evidence redundant. Warming of 5 or 6°C has not been experienced since before the ice ages, some 3 million years ago. This also relates to the point made by Stainforth et al. (2007a) that climate models face the additional problem of extrapolation – they are attempting to simulate a never before experienced state of the climate system. 19 investment) is also an active policy choice (or at least should be treated as such) and one that carries its own set of risks and potential for regret. These are most pronounced in situations that involve a degree of irreversibility in the investment decision. Where the irreversibility is on the side of the proposed investment potentially locking in exposure to future risk (sunk cost and moral hazard risk) – for example a flood defens system that carries moral hazard risk in the event of its failure – uncertainty implies that adoption of the proposal may not be safe. In other words, a safety first approach in this case would favor postponement of the investment. On the other hand, where the irreversibility is on the side of the proposed investment potentially avoiding locking-in future risk (sunk benefit or adaptation dividend) – for example, land use or zoning restrictions that attempt to guide existing development trends to avoid creating long-term, irreversible exposure to climate hazards – uncertainty implies that it may not be safe to postpone the investment. In other words, a safety first approach in this case would favor the early adoption of the proposed DRM initiative. There are relatively limited circumstances in which a full consideration of uncertainty and how it interacts with the specific characteristics of a proposed DRM investment project would be required. Specifically, this is only the case for projects that are in the first instance (highly) climate-sensitive and additionally where either; (1) no decision-relevant projections are available, or (2) under the ‘worst case’ climate scenario, the proposed project would be rendered inadequate in terms of the level of protection provided. Where climate uncertainty is most relevant, it would appear to shift the balance of appropriate investment strategies away from those with a large component of commitment or irreversibility, towards more flexible, low regret interventions. However, as noted in this section, a safety first approach is not equivalent to advocating a ‘wait and see’ attitude to DRM investments. In some cases, particularly where DRM could help to avoid locking-in future exposure to climate risk, uncertainty provides additional motivation for the early adoption of DRM policies. 5. Conclusions The expectation that climate risk (both variability and extremes) will increase for many developing countries under climate change reinforces the case for disaster risk management investments and highlights adaptation to climate change as an important potential co-benefit of DRM initiatives. Uncertainty over the precise climate risk that will be faced at any given location represents a ‘known-unknown’ for DRM strategies. The nature of climate projections is such that they offer a very uncertain picture of the future at the kind of spatial and temporal scales that are relevant for project evaluation decisions. However, ignoring uncertain future climate risk could result in exposing DRM investments to large costs in the form of maladaptation and missed opportunities for adaptation or development dividends. While the deep uncertainty associated with climate change complicates DRM investment decisions, the analysis presented here shows that these considerations are only relevant for a relatively limited set of investment circumstances. This paper offers a decision-making framework that attempts to simplify the process of accounting for the deep uncertainty associated with climate projections and the specific characteristics of different DRM policy options. In particular, this framework enables policy makers to identify the particular 20 circumstances under which uncertainty about future climate change becomes critical for DRM investment decisions. This decision framework also emphasizes two important elements of successful DRM strategies; the first is to give careful consideration to alternative uses of scarce resources – which is particularly crucial in a development context – and potential development or adaptation co-benefits. The second is to stress test DRM policies against the boundaries of the ‘climate envelope’ – in other words to consider not just ‘likely’ or expected scenarios, but to consider if the proposed intervention remains worthwhile even under the most benign or ‘worst case’ projected climate outcomes. Uncertainty related to future climate change does not necessarily motivate a ‘wait and see’ approach to DRM investments. Instead, the analysis here has demonstrated that where opportunities exist to avail of adaptation dividends – for example where DRM initiatives could help to avoid locking-in future exposure to climate risk – climate uncertainty provides additional motivation for early investment in DRM initiatives. An optimistic message also emerges from this analysis, which has identified substantial overlap between the flexible, low-regret type interventions favored under uncertain climate change, and the risk-coping initiatives that are likely to have the greatest co-benefits for economic development. Such policies deliver economic and social benefits (development dividends), regardless of the climate future that materializes in a given location. These flexible, low-regrets type DRM policies are also less likely to incur moral hazard risk, which may be associated with hard defensive infrastructure investments, such as flood defenses. This paper therefore presents a case for greater investment in disaster risk management initiatives, conditional on these being designed with an explicit development-first approach and due consideration of uncertainty over future climatic conditions. In short, appropriate DRM strategies should focus on supporting development paths that are robust to a range of possible climate futures. Such strategies would have the dual benefits of maximizing potential co-benefits of DRM investments for development, while minimizing regret under uncertain climate change. 21 References Acemoglu, Daron, & Zilibotti, Fabrizio. 1997. Was Prometheus Unbound by Chance? Risk, Diversification, and Growth. Journal of Political Economy, 105(4), pp. 709–751. Adger, WN 1999. Social vulnerability to climate change and extremes in coastal Vietnam. World Development 27(2): 249-269. Adger W, Dessai S, Goulden M, Hulme M, Lorenzoni I, Nelson D, Naess L, Wolf J, Wreford A (2009) Are there social limits to adaptation to climate change? Climatic Change 93(3):335–354 Albala-Bertrand, J.M. 1993. Political economy of large natural disasters: With special reference to developing countries. Oxford University Press. Arrow, K. J., et al. (1996). In J. J. Houghton, et al. (Eds.), Climate Change 1995, Economic and Social Dimensions of Climate Change (pp. 53–77). New York: Cambridge University Press. Auffret, Philippe. 2003. High consumption volatility: the impact of natural disasters? Policy Research Working Paper Series 2962. The World Bank. Azariadis, Costas, & Stachurski, John. 2005. Poverty Traps. Chap. 5 of: Aghion, Philippe, & Durlauf, Steven (eds), Handbook of Economic Growth, vol. 1. Elsevier. Banerjee, AV and E Duflo, 2011. Poor Economics: A radical rethinking of the way to fight global poverty. New York: Public Affairs. Barr, R., S. Fankhauser and K. Hamilton (2010). “Adaptation Investments: A Resource Allocation Framework”, in: Mitigation and Adaptation Strategies for Global Change, 15(8): 843- 858. Barrios, S. Bertinelli, L., and Strobl, E. (2006). ‘Climate change and rural-urban migration: The case of Sub Saharan Africa,’ Journal of Urban Economics 60: 357-371 Benson, C. (1997) The economic impact of natural disasters in Vietnam, Working Paper 98, Overseas Development Institute, London. Bernanke, B.S., 1983. Irreversibility, uncertainty, and cyclical investment. Quarterly Journal of Economics 98, 85–106. Broadberry, Stephen and Leigh Gardner (2013). “Africa’s growth prospects in a European mirror: A historical perspective”. CAGE-Chatham House Series, No.5, February 2013. Brooks, N., N. Adger and M. Kelly (2005). “The determinants of vulnerability and adaptive capacity at the national level and the implications for adaptation”, in: Global Environmental Change 15: 151-163. Brown, C., R. Meeks, Y Ghile and K. Hunu 2013. Is water security necessary? An empirical analysis of the effects of climate hazard on national-level economic growth. Phil Trans R Soc A 371. Bryan, Gharad, Shyamal Chowdhury and Ahmed Mushfiq Mobarak (2014). 22 Underinvestment In A Profitable Technology: The Case Of Seasonal Migration In Bangladesh. Econometrica Vol. 82, No. 5 (September, 2014), 1671–1748 Burke, M, J Dykema, D Lobell, E Miguel and S Satyanath 2011. Incorporating climate uncertainty into estimates of climate change impacts, with applications to U.S. and African agriculture. NBER Working Paper 17092. Burton, I. (2009). “Climate Change and the Adaptation Deficit”, in: E.L.F. Schipper and I. Burton, eds. The Earthscan Reader on Adaptation to Climate Change, London: Earthscan Cavallo, Eduardo Alfredo, & Noy, Ilan. 2009. The economics of natural disasters: A survey. RES Working Papers 4649. Inter-American Development Bank, Research Department. Collier, P., Conway, G., and Venables, T. (2008). ‘Climate change and Africa,’ Oxford Review of Economic Policy 24(2): 337-353. Daron, JD & DA Stainforth (2013) On predicting climate under climate change. Environ Res Letters 8. Davis, Donald R., and David E. Weinstein. “Bones, Bombs, And Break Points: The Ge- ography of Economic Activity.” American Economic Review 92.5 (2002): 1269. Dercon, S. (1998), “Risk, Crop Choice and Savings : Evidence from Tanzania”, (1996), Economic Development and Cultural Change, vol. 44, no. 3, April, pp. 485-514. Dercon, Stefan (2012), “Is Green Growth Good for the Poor?” World Bank Policy Research Working Paper 6231. Washington DC: World Bank. Dessai S, Hulme M (2007) Assessing the robustness of adaptation decisions to climate change uncertainties: A case study on water resources management in the East of England. Global Environmental Change 17(1):59–72 Dessai, S., Adger, W. N., Hulme, M., Turnpenny, J., Koler, J., & War- ren, R. (2004). Defining and experiencing dangerous climate change—An editorial essay. Climatic Change, 64, 11–15. Dessai S, Hulme M, Lempert R, Pielke Jr R (2009) Climate prediction: A limit to adaptation? In: WN Adger, I Lorenzoni, K O’Brien (eds) Adapting to Climate Change: Thresholds, Values, Governance, Cambridge University Press, pp 64–78 Dixit AK, Pindyck RS (1994) Investment under uncertainty. Princeton University Press, Princeton, N.J. Easterly, William. 2006. Reliving the 1950s: the big push, poverty traps, and takeoffs in economic development. Journal of Economic Growth, 11, 289–318. Easterly, William, & Levine, Ross. 2003. Tropics, germs, and crops: how endowments influence economic development. Journal of Monetary Economics, 50(1), 3 – 39. Elbers, C, JW Gunning and B Kinsey, 2007. Growth and risk: Methodology and Micro Evidence. The World Bank Economic Review 21(1): 1-20. 23 Fankhauser, Samuel and Thomas K.J. McDermott. “Understanding the adaptation deficit: Why are poor countries more vulnerable to climate events than rich countries?” Global Environmental Change 27 (2014): 9-18. Fankhauser S, Smith J, Tol R (1999) Weathering climate change: some simple rules to guide adaptation decisions. Ecological Economics 30 Government Office for Science [UK] 2011. Foresight: Migration and Global Environmental Change – Final project report. Hales, S., Edwards, S.J., & Kovats, R.S. 2003. Impact on health of climate extremes. In: McMichael et al., A.J. (ed), Climate change and human health: Risks and responses. Geneva: World Health Organization. Hallegatte, S. 2009 Strategies to adapt to an uncertain climate change. Global Environmental Change 19: 240-247. Hallegatte, S (2013) A normative exploration of the link between development, economic growth, and natural risk. Mimeo. Hallegatte, S (2014), “Economic Resilience”. World Bank Policy Research Working Paper 6852. Hallegatte, S., Hourcade, J-C. & Dumas, P. (2007). Why economic dynamics matter in assessing climate change damages: Illustration on extreme events. Ecological Economics, 62, 330-340. Hallegatte, S. et al. (2010), “Flood Risks, Climate Change Impacts and Adaptation Benefits in Mumbai: An Initial Assessment of Socio-Economic Consequences of Present and Climate Change Induced Flood Risks and of Possible Adaptation Options”, OECD Environment Working Papers, No. 27, OECD Publishing. Harris, J. and Todaro, M. (1970). ‘Migration, unemployment and development: a two- sector analysis,’ The American Economic Review 60(1): 126-142. Hausmann, Ricardo, Pritchett, Lant, & Rodrik, Dani. 2005. Growth Accelerations. Journal of Economic Growth, 10(4), pp. 303–329. Heal, G. and A. Millner (2014). Uncertainty and decision making in climate change economics. Review of Environmental Economics and Policy 8(1):120-137. Henderson, J.V., Storeygard, A., and Deichmann, U. (2014). ‘50 years of urbanization in Africa,’ World Bank Policy Research Working Paper 6925 Hsiang, SM and AS Jina 2014 . The Causal Effect Of Environmental Catastrophe On Long-Run Economic Growth: Evidence From 6,700 Cyclones. NBER Working Paper 20352. IPCC (2013). Climate change 2013: The physical science basis. Contribution of Working Group 1 to the “Fifth Assessment Report” of the Intergovernmental Panel on Climate. Cambridge: Cambridge University Press. Jaeger, C., Renn, O., Rosa, E., & Webler, T. (1998). Decision anal- ysis and rational action. 24 In S. Rayner & E. L. Malone (Eds.), Human Choice and Climate Change, volume 3 (pp. 141– 216). Columbus, OH: Battelle Press. Jacoby, Hanan G., & Skoufias, Emmanuel. 1997. Risk, Financial Markets, and Human Capital in a Developing Country. The Review of Economic Studies, 64(3), 311–335. Janowiak, John E. 1988. An Investigation of Interannual Rainfall Variability in Africa. Journal of Climate, 1(Mar.), 240–255. Judd, K., Smith, L.A. & Weisheimer, A. 2004. Gradient free descent: Shadowing and state estimation using limited derivative information. Physica D 190: 153-166. Jury, Mark R. 2002. Economic Impacts of Climate Variability in South Africa and Devel- opment of Resource Prediction Models. Journal of Applied Meteorology, 41(1), 46–55. Kahn, Matthew E. 2005. The Death Toll from Natural Disasters: The Role of Income, Geography, and Institutions. Review of Economics and Statistics, 87(2), 271–284. Kellett, J. and Caravani, A. 2014. Financing Disaster Risk Reduction: A 20 year story of international aid. London: Overseas Development Institute. Kemp, M. 2005. Science in culture: Inventing an icon. Nature 437, 1238. Knutti R. 2008. Should we believe model predictions of future climate change? Philo- sophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 366(1885):4647 –4664 Knutti, R., and J. Sedlacek, 2013. Robustness and uncertainties in the CMIP5 climate model projections. Nature Climate Change 3: 369-73. Lancet, The. 2009. Managing the health effects of climate change. Lancet, 373, 1693–733. Lempert, R.J. & Collins, M.T., (2007). “Managing the Risk of Uncertain Threshold Responses: Comparison of Robust, Optimum, and Precautionary Approaches”. Risk Analysis, 27(4), 1009-1026. Lewis, W. A. 1954. ‘Economic Development with Unlimited Supplies of Labour,’ The Manchester School 22: 139-191 Loayza, Norman V., Ranciere, Romain, Serven, Luis, & Ventura, Jaume. 2007. Macroeconomic Volatility and Welfare in Developing Countries: An Introduction. The World Bank Economic Review, 21(3), 343–357. Majd S, Pindyck RS (1987) Time to build, option value, and investment decisions. Journal of Financial Economics 18(1):7–27 Marx, B., Stoker, T., and Tavneet, S. (2013). ‘The economics of slums in the developing world,’ Journal of Economic Perspectives 27(4): 187-21 Masson, D. and R. Knutti, 2011. Spatial scale dependence of climate model performance in the CMIP3 ensemble. Journal of Climate 24(11): 2680-92. Mechler, Reinhard, Laurens M. Bouwer, Joanne Linnerooth-Bayer, Stefan Hochrainer- 25 Stigler, Jeroen C. J. H. Aerts, Swenja Surminski and Keith Williges (2014). Managing unnatural disaster risk from climate extremes. Nature Climate Change 4: 235-7. Michaels, Guy, and Ferdinand Rauch (2013). Resetting the urban network: 117-2012. Bureau for Research and Economic Analysis of Development (BREAD), 2013. Millner, Antony (2012), “Climate Prediction for Adaptation: Who needs what?” Climatic Change 110: 143-167. Millner, Antony (2013), “On welfare frameworks and catastrophic climate risks”, Journal of Environmental Economics and Management 65(2): 310-25. Mobarak, AM and MR Rosenzweig, 2013. “Informal risk sharing, index insurance, and risk taking in developing countries”, American Economic Review 103(3):375-80. Morduch, J. (1995), “Income smoothing and consumption smoothing”, Journal of Economic Perspectives. 9(3):103-114. Nicholson, Sharon E. 2000. The nature of rainfall variability over Africa on time scales of decades to millenia. Global and Planetary Change, 26(1-3), 137 – 158. Oreskes, N, DA Stainforth and LA Smith 2010. Adaptation to Global Warming: Do Climate Models Tell us What we Need to Know? Philosophy of Science 77: 1012-1028. Pascual, M, Cazelles, B, Bouma, M.J, Chaves, L.F, & Koelle, K. 2008. Shifting patterns: malaria dynamics and rainfall variability in an African highland. Proceedings of the Royal Society B: Biological Sciences, 275(1631), 123–132. Paxson, C. (1992). Using weather variability to estimate the response of savings to transitory income in Thailand, The American Economic Review 82(1): 15-33. Pindyck, RS 2002. Optimal timing problems in environmental economics. Journal of Economic Dynamics and Control 26: 1677–1697. Roberts K, Weitzman ML (1981) Funding criteria for research, development, and exploration projects. Econometrica 49(5):1261–1288 Rodrik (1999) Where Did All the Growth Go? External Shocks, Social Conflict, and Growth Collapses. Journal of Economic Growth, 4, 385–412. Rosenzweig, M., and Binswanger, H. (1993).Wealth, weather risk and the composition and profitability of agricultural investments, The Economic Journal 103(416): 56-78. Rosenzweig, Mark R., & Wolpin, Kenneth I. 1993. Credit market constraints, consumption smoothing, and the accumulation of durable production assets in low- income countries: Investments in bullocks in India. Journal of Political Economy, 101(2), pp. 223–244. Samson, J., Berteaux, D., McGill, B.J., & Humphries, M.M. 2011. Geographic disparities and moral hazards in the predicted impacts of climate change on human populations. Global Ecology and Biogeography, 20(4), 532–44. Schelling, T. (1992), “Some Economics of Global Warming”, American Economic Review 82: 1-14. 26 Schelling, T. (1997). The Cost of Combating Global Warming: Facing the Tradeoffs, in: Foreign Affairs, 76(6): 8-14. Scoones, Ian, & Chibudu, Chinaniso. 1996. Hazards and opportunities: Farming livelihoods in dryland Africa, lessons from Zimbabwe. London and New Jersey: Zed Books. Sen, Amartya. 1983. Poverty and Famines: An Essay on Entitlement and Deprivation. Oxford University Press, USA. Skoufias, E., Rabassa, M., and Olivieri, S. (2011). The poverty impacts of climate change: A review of the evidence. World Bank Policy Research Working Paper 5622. The World Bank. Smith DM, Cusack S, Colman AW, Folland CK, Harris GR, Murphy JM (2007) Improved surface temperature prediction for the coming decade from a global climate model. Science 317(5839):796–799 Smith LA (2007) Chaos : a very short introduction, vol 159. Oxford University Press, Oxford Solomon, Susan et al., eds. 2007. Climate change 2007: The physical science basis. Contribution of Working Group 1 to the “Fourth Assessment Report” of the Intergovernmental Panel on Climate. Cambridge: Cambridge University Press. Spiegel, EA (1987). Chaos – a mixed metaphor for turbulence. Proc R Soc A 413: 87-95. SREX (2012): Special report of the IPCC: Managing the risks of extreme events and disasters to advance climate change adaptation. Cambridge: Cambridge University Press Stainforth D, Allen M, Tredger E, Smith L (2007a) Confidence, uncertainty and decision- support relevance in climate predictions. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 365(1857):2145–2161 Stainforth, DA, TE Downing, R Washington and A Lopez 2007b. Issues in the interpretation of climate model ensembles to inform decisions. Phil Trans R Soc A 365. Stainforth, D.A., S.C. Chapman, W.A. Watkins (2013) “Mapping climate change in European temperature distributions”, Environmental Research Letters 8. Stern, Nicholas H. 2007. The economics of climate change: the Stern review. Cambridge University Press. Strulik, Holger. 2008. Geography, health, and the pace of demo-economic development. Journal of Development Economics, 86(1), 61 – 75. Swart, Rob, Sabine Fuss, 
 Michael Obersteiner, Paolo Ruti, 
 Claas Teichmann and Robert Vautard (2013). Beyond vulnerability assessment. Nature Climate Change 3: 942-3. Tang, Kam Ki, Petrie, Dennis, & Prasada Rao, D.S. 2009. The income-climate trap of health development: A comparative analysis of African and Non-African countries. Social Science and Medicine, 69(7), 1099 – 1106. 27 Tanner, Tom and Jun Rentschler (2015). “Unlocking the triple dividend of resilience”, Interim policy note. London: Overseas Development Institute. Tebaldi C, Knutti R (2007) The use of the multi-model ensemble in probabilistic cli- mate projections. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 365(1857):2053–2075 Tol, Richard S.J., & Leek, Frank P.M. 1999. Economic analysis of natural disasters. In: Downing, Thomas E., Olsthoorn, Alexander A., & Tol, Richard S.J. (eds), Climate, change and risk. London: Routledge. Tol, Richard S.J., Ebi, Kristie L., & Yohe, Gary W. 2007. Infectious disease, development, and climate change: a scenario analysis. Environment and Development Economics, 12, pp 687–706. Tol, R.S.J. and G.W. Yohe (2007), 'The Weakest Link Hypothesis for Adaptive Capacity: An Empirical Test’, Global Environmental Change, 17: 218-227. Verschuren, Dirk, Laird, Kathleen R., & Cumming, Brian F. 2000. Rainfall and drought in equatorial east Africa during the past 1,100 years. Nature, 403(6768), pp. 410–414. Weitzman, ML 2009. On modeling and interpreting the economics of catastrophic climate change. Review of Economics and Statistics 91(1): 1-19. World Bank (2010), World Development Report: Development and Climate Change. Washington DC: World Bank. World Bank (2011), Building Resilience to Disasters – Delivering Results. Global Facility for Disaster Reduction and Recovery Annual Report 2011. World Bank (2013), Turn down the heat. Climate extremes, regional impacts and the case for resilience. A report for the World Bank by the Potsdam Institute for Climate Impact Research and Climate Analytics. Washington DC: World Bank. 28