WPS5507 Policy Research Working Paper 5507 The Economics of Natural Disasters Concepts and Methods Stéphane Hallegatte Valentin Przyluski The World Bank Sustainable Development Network Office of the Chief Economist December 2010 Policy Research Working Paper 5507 Abstract Large-scale disasters regularly affect societies over the the cost of a disaster, and emphasizes the most important globe, causing large destruction and damage. After mechanisms that explain and determine this cost. It each of these events, media, insurance companies, does so by first explaining why the direct economic and international institutions publish numerous cost, that is, the value of what has been damaged or assessments of the "cost of the disaster." However these destroyed by the disaster, is not a sufficient indicator of assessments are based on different methodologies and disaster seriousness and why estimating indirect losses approaches, and they often reach different results. Besides is crucial to assess the consequences on welfare. The methodological differences, these discrepancies are due paper describes the main indirect consequences of a to the multi-dimensionality in disaster impacts and their disaster and the following reconstruction phase, and large redistributive effects, which make it unclear what discusses the economic mechanisms at play. It proposes is included in the estimates. But most importantly, the a review of available methodologies to assess indirect purpose of these assessments is rarely specified, although economic consequences, illustrated with examples from different purposes correspond to different perimeters of the literature. Finally, it highlights the need for a better analysis and different definitions of what a cost is. To understanding of the economics of natural disasters and clarify this situation, this paper proposes a definition of suggests a few promising areas for research on this topic. This paper is a product of the Chief Economist of the Sustainable Development Network. 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 hallegatte@centre-cired.fr. 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 The Economics of Natural Disasters: Concepts and Methods1 Stéphane Hallegatte (Météo-France, CIRED), Valentin Przyluski (CIRED) 1 A previous and shorter version of this text has been published in CESifo : Hallegatte S., Przyluski V., The economics of natural disaster, CESifo Forum 2 06/2010, 1. Introduction Largescale disasters regularly affect societies over the globe, causing huge destruction and damage. The 2010 earthquake in PortauPrince and hurricane Katrina in 2005 have shown that poor as well as rich countries are vulnerable to these events, which have longlasting consequences on welfare, and on hu man and economic development. After each of these events, media, insurance companies and international institutions publish numerous assessments of the "cost of the disaster." However these various assessments are based on different me thodologies and approaches, and they often reach quite different results. Beside technical problems, these discrepancies are due to the multidimensionality in disaster impacts and their large redistributive effects, which make it unclear what is included or not in disaster cost assessments. But most importantly, the purpose of these assessments is rarely specified, even though different purposes correspond to dif ferent perimeters of analysis and different definitions of what a cost is. This confusion translates into the multiplicity of words to characterize the cost of a disaster in published assessments: direct losses, asset losses, indirect losses, output losses, intangible losses, market and non market losses, welfare losses, or any combination of those. It also makes it almost impossible to compare or aggregate published estimates that are based on so many different assumptions and methods. To clarify the situation, this paper proposes a definition of the cost of a disaster, and emphasizes the most important mechanisms that explain this cost. It does so by first explaining why the direct economic cost, i.e. the value of what has been damaged or destroyed by the disaster, is not a sufficient indicator of disas ter seriousness and why estimating indirect losses is crucial. Then, it describes the main indirect conse quences of a disaster and of the following reconstruction phase, and discusses the methodologies to measure them. Finally, it proposes a review of a few published assessments of indirect economic conse quences, which confirm their importance and the need to take them into account. 2. The indirect cost of natural hazards 2.1 What is a disaster? What is an indirect cost? There is no single definition of a disaster. From an economic perspective, however, a natural disaster can be defined as a natural event that causes a perturbation to the functioning of the economic system, with a significant negative impact on assets, production factors, output, employment, or consumption. Exam ples of such natural event are earthquakes, storms, hurricanes, intense precipitations, droughts, heat waves, cold spells, and thunderstorms and lightning. Disasters affect the economic system in multiple ways, and defining the "cost" of a disaster is tricky. Pel ling et al. (2002), Lindell and Prater (2003), Cochrane (2004), Rose (2004), among others, discuss typolo gies of disaster impacts. These typologies usually distinguish between direct and indirect losses. Direct losses are the immediate consequences of the disaster physical phenomenon: the consequence of high winds, of water inundation, or of ground shaking. Direct losses are often classified into direct market losses and direct nonmarket losses (also sometimes referred to as intangible losses, even though non 2 market losses are not necessarily intangible). Market losses are losses to goods and services that are traded on markets, and for which a price can easily be observed. Even though droughts or heat waves affect directly the economic output (especially in the agriculture sector), direct market losses from most disasters (earthquakes, floods, etc.) are losses of assets, i.e. damages to the built environment and manu factured goods. These losses can be estimated as the repairing or replacement cost of the destroyed or damaged assets. Since building and manufactured goods can be bought on existing markets, their price is known. Direct market losses can thus be estimated using observed prices and inventories of physical losses that can be observed (as recorded, e.g., in the EMDAT database or insuranceindustry databases) or modeled (using, e.g., catastrophe models of the insurance industry or simplified methodologies; see Ranger et al., 2011). Nonmarket direct losses include all damages that cannot be repaired or replaced through purchases on a market. For them, there is no easily observed price that can be used to estimate losses. This is the case, among others, for health impacts, loss of lives, natural asset damages and ecosystem losses, and damag es to historical and cultural assets. Sometimes, a price for nonmarket impacts can be built using indirect methods, but these estimates are rarely consensual (e.g., the statistical value of human life). Indirect losses (also labeled "higherorder losses" in Rose, 2004) include all losses that are not provoked by the disaster itself, but by its consequences. Different hazards communities have different approaches for defining indirect costs (e.g. Meyer and Messner (2005) and FLOODSite (2009) for floods; Wilhite (2000) and Wilhite et al. (2007) for droughts). Contentious issues may emerge around the edge of these definitions across hazard communities: what are the limits between direct and indirect costs categories? In particular discussions often occur around the notion of business interruption, which can be included in direct losses or in indirect losses, or as a standalone category. For capitaldestroying hazards (flood, earthquakes, storms), the term "indirect losses" is often used as a proxy for "output losses," i.e. the re duction in economic production provoked by the disaster. Output losses include the cost of business in terruption caused by disruptions of water or electricity supplies, and longer term consequences of infra structure and capital damages. An obvious illustration of why indirect losses are important is the difference between disaster scenarios with various reconstruction paces. In terms of welfare, there is a large difference between, on the one hand, a scenario in which all direct losses can be repaired in a few months thanks to an efficient recon struction process and, on the other hand, a scenario in which reconstruction is inefficient and takes years. For the same amount of direct losses, welfare impacts are much larger in the latter case, and this should be taken into account. To help identify indirect losses, we propose the following criteria. First, indirect losses are caused by sec ondary effects, not by the hazard itself. Indirect costs can be caused by hazard destructions or by business interruptions. In addition to this obvious criterion, costs are indirect if they are spanning on a longer pe riod of time, a larger spatial scale or in a different economic sector than the disaster itself. Classification of hazards by scale and time effect is done by, e.g., Brown Gaddis et al. (2006), and Jonkman et al. (2008). This definition is consistent with definitions from different hazard communities. It includes business inter ruption in direct losses (since their most classical definition makes them mainly shortterm, during the hazard duration). Also, this definition avoids consistency problems for slowonset hazards such as 3 drought. With this definition, the reduction in agriculture yield, and in farmer income, are considered as direct costs, consistent with intuition, while the impacts on other economic sector trading with the agri cultural sector are indirect costs. Indirect losses can be market or nonmarket losses (see f.i., Government of Queensland, 2002). Some times, nonmonetary indirect consequences of disasters are also included, like the impact on poverty or inequalities, the reduction in collected taxes, or the increase in national debt. In more general terms, several issues are raised by the use of GDP change as an indicator to assess indi rect losses. These issues are, among others, (i) the question of appropriate scale between the scale of the event and the scale of GDP measurement, (ii) the capacity of GDP to be a good proxy for welfare (see, e.g., CMEPSP, 2009; Council and European Parliament, 2009). 2.2 Definition and assessment purpose These possible definitions of indirect losses create specific difficulties. For instance, indirect losses can have "negativecost" components, i.e. gains from additional activity created by the reconstruction. These gains can be realized in the affected region (e.g., in the construction sector) or in another region (e.g. because it captures market shares lost by the affected region). Another difficulty in disaster cost assessment lies in the definition of the baseline scenario. The cost of the disaster has indeed to be calculated by comparing the actual trajectory (with disaster impacts) with a counterfactual baseline trajectory (i.e., a scenario of what would have occurred in the absence of a disas ter). This baseline is not easy to define, and several baselines are often possible. Moreover, in cases where recovery and reconstruction do not lead to a return to the baseline scenario, there are permanent (positive or negative) disaster effects that are difficult to compare with a nondisaster scenario. For instance, a disaster can lead to a permanent extinction of vulnerable economic activities in a region, because these activities are already threatened and cannot recover, or because they can move to less risky locations. In that case, the disaster is not a temporary event, but a permanent negative shock for a region and it is more difficult to define the disaster cost. Also, reconstruction can be used to develop new economic sectors, with larger productivity, and lead to a final situation that can be considered more de sirable than the baseline scenario. This improvement can be seen as a benefit of the disaster. It is howev er difficult to attribute unambiguously this benefit to the disaster, because the same economic shift would have been possible in absence of disaster, making it possible to get the benefits without suffering from the disasterrelated human and welfare losses. Most importantly, defining the cost of a disaster cannot be done independently of the purpose of the assessment. Different economic agents, indeed, are interested in different types of cost. Insurers, for instance, are mainly interested in consequences that can be insured. Practically, this encompasses mainly the cost of damages to insurable assets (e.g., damaged houses and factories), and shortterm business interruption caused by the disaster (e.g., the impossibility to produce until electricity is restored). 4 For affected households, insurable assets are also a major component, but other cost categories are as important. Primarily, loss of lives, health impacts and perturbation to their daily life are crucial. In addi tion, household are concerned about their assets but also about their income, which can be reduced by business interruption or by loss of jobs, and about their ability to consume, i.e. the availability of goods and services. At the society scale, all these aspects are important, but local authorities, governments and international institutions are also interested in other points. First, to manage the recovery and reconstruction period and to scale the necessary amount of international aid, they need information on the aggregated impact on economic production, on unemployment and jobs, on the impact of inequality and poverty, on local businesses marketshares, on commercial balance, on collected taxes, etc. Second, to assess whether investment in prevention measures are desirable, they need the broadest possible assessments of the total disaster cost to the population, i.e. an estimate of welfare losses. Moreover, disaster impacts can have positive or negative rippleeffects at the global scale, as shown by hurricane Katrina, which led to a significant rise in world oil prices. Depending on the purpose and of the decisionmaking spatial scale, the perimeter of the cost analysis will be different. For instance, a country may want to assess the losses in the affected region, disregarding all outoftheregion impacts, to cali brate the financial support it wants to provide to the victims. But it may also want to assess total losses on its territory, including gains and losses outside the affected region, for example to assess the impact on its public finance. Clearly, depending on the purpose of the assessment, some of the cost components have to be included or not in the analysis. In the following, we focus on the economic cost for the affected region, with the aim of informing decisionmakers on postdisaster financial aid and prevention measure desirability. Assessing a loss of welfare is complicated, as it includes many economic and noneconomic components. Here, we focus on the economic component of welfare losses, and we define the economic cost of disas ter as the lost consumption, considered as an important component and a good proxy of economic related welfare losses2. Of course, this paper does not try to be comprehensive, and major cost compo nents are left out of the analysis, like loss of lives, or health consequences. These additional component are important for the population welfare and therefore for prevention measure assessments. 2.3 Consumption losses and output losses This section explains how to assess consumption losses from asset and output losses. More precisely, it explains why the sum of asset losses and output losses is a good proxy for the loss of consumption. To do so, Fig. 1, (a), (b) and (c) show simplified representations of a postdisaster situation. Figure 1(a) depicts the situation in which only output losses are estimated. In this scenario, the disaster leads to a temporary reduction in output during the reconstruction phase. We assume here that reconstruction is a return to the baseline scenario (i.e. a nodisaster counterfactual scenario). As already stated, this is not always the case, but making an assumption on the final state is necessary to define the "cost" of the disaster, and the assumption of a return to the nondisaster baseline scenario is likely to be the most neutral one for this type of assessment. 2 In an utilitarist framework, what matters is not output and production, but consumption. 5 The sum of instantaneous output loss is what is often referred to as the indirect loss. But reconstruction needs in the disaster aftermath mean that a significant share of the remaining production will have to be devoted to reconstruction, as shown in Fig. 1(b). In other terms, the resources used to rebuild damaged houses cannot be used to build new houses, or to maintain existing ones. This reconstruction output is included in total output, and is not a loss of output. But it is a "forced" investment, in addition to the normaltime "investment--consumption" tradeoff. It causes, therefore, a loss of welfare. The value of this forced investment is the replacement value of damaged asset, i.e. what is referred to as the direct losses. This is what is represented in Fig. 1(c): the sum of the output loss and of the reconstruction out put is what cannot be used for consumption and nonreconstruction investment, and what is here re ferred to as "total losses." In this framework, the total cost is the sum of the indirect cost (i.e., the reduction of the total value add ed by the economy due to the disaster), and the direct cost (i.e., the portion of the remaining value added that has to be dedicated to reconstruction instead of normal consumption). Capital and output losses can therefore be simply added to estimate consumption losses. Of course, Fig. 1 shows a simplified situation in which production has no flexibility. In this case, recon struction needs cannot be satisfied through increased production and it has to crowd out other consump tions and investments. Figure 2 depicts a different case, in which there is a limited flexibility in the pro duction process: capital destruction leads to a reduction in output; but unaffected capital can increase its own production to compensate this reduction, for instance through an increase in work hours by workers at unaffected factories and businesses. In practice, there are gross indirect losses, and gross indirect gains (due to the stimulus effect of the reconstruction). But there is still a fraction of the remaining production that is used for reconstruction instead of normal consumption, even though this share is smaller than in absence of production flexibility. 6 7 Figure 1: Direct losses, indirect losses, and "total" losses, i.e. consumption losses. This figure assumes that there is no flexibility in the production process. 8 Figure 2: Direct losses, indirect losses, and "total" losses, i.e. consumption losses. This figure assumes that there is a limited flexibility in the production process. 9 In this situation also, the consumption loss is still the sum of direct (asset) and indirect (output) losses (Figure 2.c), making it necessary to estimate output losses. But output losses are not only the lost pro duction from the affected capital, but also the output gains and losses from unaffected capital, in the rest of the economy. It makes the assessment of output losses more complicated, since it depends on com plex economic mechanisms and tradeoffs. Moreover, in practice, the reduction in consumption can be mitigated or amplified by (i) changes in pric es; (ii) flexibility in the production process; (iii) changes in the savingconsumption tradeoff for the re maining production and, (iv) the fact that the rebuilt capital will be more recent than before the disaster, with potential benefits. The following section will describe methodologies to assess output losses, and highlight the most important processes to take into account. 3. Processes and determinants of indirect losses If output losses represent an important component of total losses, it becomes essential to develop me thodologies to assess them. To do so, we propose to start by assessing the lost output from the directly affected capital. In a second subsection, we investigate the systemic impacts of disasters, including the effect on the capital that is not directly affected by the disaster, and we will see that these mechanisms lead to several open research questions. 3.1 From asset losses to output losses The first step in an assessment of output losses is to estimate how much output is lost because of direct asset losses. Economic theory states that, at the economic equilibrium and under certain conditions, the value of an asset is the net present value of its expected future production, and this equality has been widely used to assess disaster output losses. Assuming this equality is always verified, the output loss caused by capital loss is simply equal to the value of the damages, capital losses and output losses are simply equal, and the sum of asset and output losses is the double of asset losses. The assumption that output losses are equal to capital losses is however based on strong assumptions, which are not always verified. In estimates of disaster consequences, what is referred to as "asset loss" is the replacement value of the capital. To have the equality of asset loss and output loss, a double equality needs to be verified: re placement value has to be equal to market value; and market value has to be equal to the net present value of expected output. In an optimal economy at equilibrium, these two equalities are valid. First, if the market value of an asset is lower (resp. larger) than the net value of its output, then investors will buy more (resp. sell) more of this asset to capture the difference in value, making asset price increase (resp. decrease). Second, if market value were higher (resp. lower) than replacement value, then investors would increase (resp. decrease) the amount of physical capital to restore the equality between market and replacement value (assuming decreasing returns). Therefore, in theory, there is no difference between capital losses and the reduction in output from this capital. But the required assumptions are questionable. 10 First, for the replacement value and the market value to be equal, the economy needs to be at its opti mum, i.e. the amount of capital is such that its return is equal to the (unique) interest rate. This is unlikely for the capital that is affected by natural disasters, especially as infrastructure and public assets are heavi ly affected. Since these assets are not exchanged on markets, they have no market prices. Moreover, they are not financed by private investors, but decided about through a political process taking into account multiple criteria (e.g., landuse planning objectives), and there is no reason for their purelyfinancial re turn to be equal to the (private) interest rate. Practically, some assets may have an output value lower than their replacement value (e.g., a secondary road that is redundant and does not provide a significant gain of time or distance), while some may have an output value much larger than their replacement val ue (e.g., a bridge that cannot be closed without large consequences for users). Second, for market values to be equal to net present value of expected output, expectations have to be unbiased and markets need to be perfect. This is not always the case especially in sectors affected by disasters, where expectations can be heavily biased (e.g., in housing market). Also, output losses are most of the time estimated from a social pointofview. The equality between market value (for the owner) and expected output (for the society) is valid only in absence of externali ties. Some assets that are destroyed by disasters may exhibit positive externality. It means that their val ue to the society is larger than the value of the owner's expected output. Public goods have this charac teristic, among which most infrastructures, health services, education services. An example is provided by the San Francisco Oakland Bay Bridge, which is essential to the economic ac tivity in San Francisco and had to be closed for one month after the Loma Prieta earthquake in 1989. Its replacement value has no reason to be equal to the loss in activity caused by the bridge closure, because the bridge production is not sold on a market, the bridge has no market value, and the social return on capital of the bridge is unlikely to exhibit decreasing returns and is likely to be much higher than the in terest rate. Another example is the health care system in New Orleans. Beyond the immediate economic value of the service it provided, a functioning health care is necessary for a region to attract workers. After Katrina landfall on the city, the absence of health care services made it more difficult to reconstruct, and the cost for the region was much larger than the economic direct value of this service. Finally, financial issues may also lead to a difference between losses in asset market value and reductions in net present value of expected outputs. Indeed, a natural disaster may lead to the bankruptcy of some businesses, even though their assets are not fully wiped out. In that case, if the remaining assets are not "recycled" by other businesses, then the affected businesses production is reduced by 100%, while their asset values were only partially reduced. 3.2 The systemic impact of natural disasters and open research questions The equality between output losses and asset losses is questionable for any economic shock, small or large. The most important issues appear when considering very large shocks, or systemic events, which are the events that perturb the functioning of the entire economic system and affect relative prices. In this case, output losses may be damped or amplified by several mechanisms. 11 Changes in prices Fig. 1 and 2 show output in real terms, i.e. with no monetary effects. But output losses can be estimated assuming unchanged (predisaster) prices or taking into account the impact of the disaster on prices. Both assumptions lead to the same result if the disaster has only a marginal impact on the economy, with little impact of prices, but can be very different in the opposite situation. In other terms, one can assume that if a house is destroyed, the family who owns the house will just have to rent another house at the predisaster price. But this assumption is unrealistic if the disaster causes more than a marginal shock. In postdisaster situations, indeed, a significant fraction of houses may be destroyed, leading to changes in the relative price structure. In this case, the price of alternative housing can be much higher than the pre disaster price, as a consequence of the disasterrelated scarcity in the housing market. Estimating the value of lost housing service should then be done using this higher cost instead of the predisaster one, which can lead to a significant increase in the assessed disaster cost. Unfortunately, it is difficult to pre dict ex ante the change in prices that would be caused by a disaster, making loss assessment more com plicated. The same reasoning is possible in all other sectors, including transportation, energy, water, health, etc. In extreme cases, there may be rationing. This is because markets are not at equilibrium in disaster after math (i.e., price is not such that demand equals production). The « If I can pay it, I can get it » assumption is not valid in postdisaster situations (e.g., there is no available house for rent at any price, there is no qualified worker to repair a roof). In these situations, therefore, the welfare impact of lost production cannot be estimated as the product of lost produced quantity and predisaster prices. Providing an un biased estimate requires an assessment of the disaster impact on prices and taking into account ration ing. Often considered as resulting from unethical behavior from businesses, which are thought to benefit from the disaster, postdisaster price inflation (also referred to as "demand surge") can also have positive consequences. This inflation, indeed, helps attract qualified workers where they are most needed and creates an incentive for all workers to work longer hours, therefore compensating for damaged assets and accelerating reconstruction. It is likely, for instance, that higher prices after hurricane landfalls are useful to make roofers from neighboring unaffected regions move to the landfall region, therefore in creasing the local production capacity and reducing the reconstruction duration. Demand surge, as a consequence, may also reduce the total economic cost of a disaster, even though it increases its burden on the affected population. Length of the reconstruction phase Importantly, there is a large difference between losing a home for one day (in this case the total loss is the reconstruction value, i.e. the direct loss) and losing a home for one year (in this case the total loss is the reconstruction value, i.e. the direct loss, plus the value of one year of housing services, i.e. the output loss). Of course, the longer the reconstruction period, the larger the total cost of the disaster. The reconstruction phase, and the economic recovery pace, will ultimately determine the final cost of the natural disasters. The reconstruction pace is linked to the constraints to the reconstruction phase, which are of two types. First, they can be financial. This concerns situations in which households and businesses can simply not finance the reconstruction. This is of particular importance in countries with limited re sources (Freeman et al., 2002; Mechler et al., 2006). 12 Constraints are also technical. Technical limits to the ability to increase production are obvious in the construction sector, which experience a dramatic increase in demand after the disaster. In spite of this demand, production does not follow, because there are strong constraints on reconstruction. Many households are able to pay for reconstruction, but cannot find workers and contractors to carry out the work. For instance, the limited availability of glaziers increased the cost of reconstruction and lowered the reconstruction pace after the 2001 chemical explosion in Toulouse (France), despite glaziers coming from all the country to carry out the work. The same is true for businesses and factories. This explains why reconstruction often takes several years, even for limited damages (e.g., the 2004 hurricane season in Florida; see McCarty and Smith, 2005). Output gains and losses from the nonaffected capital Damages in crucial intermediate sectors may lead to negative "network effects" in the economy, leading to production losses even for businesses that are not directly affected by the disaster. Water, electricity, gas and transportation are the most critical sectors, and their production interruption has immediate consequences on the entire economic system. In past cases, it has been shown that the indirect conse quences of utility services had larger consequences than direct asset losses, both on households (McCar ty and Smith, 2005) and on businesses (e.g., Tierney, 1997). Of course, when capital cannot produce be cause of a lack of input (e.g., electricity, water), input substitution, production rescheduling, and longer work hours can compensate for a significant fraction of the losses (see Rose et al., 2007). These mechan isms can damp output losses, and can especially reduce the crowdingout effects of reconstruction on normal consumption and investment (see Fig. 2). But their ability to do so is limited, especially when losses are large. There are many sources of flexibility in the economic system. First, production capacity is not fully used in normal times, and idle production capacity can be mobilized in disaster aftermath to compensate for lost production from lost assets. Second, behaviors can change in disaster aftermath, and workers can in crease their work hours in unaffected businesses to help society cope with disaster consequences (and sometimes benefit from increased prices). As a consequence, unaffected capital can often increase pro duction to compensate for output loss from affected capital. After mild disasters, net output gains can even be observed, explained by the nonzero price elasticity of production, and by the underoptimality of the predisaster situation that leaves some room for increased production (e.g., Hallegatte and Ghil, 2008). In an economy that fully uses all resources and cannot increase its production over the shortterm (whatever the price level), such a gain would be impossible. In a more realistic economy that does not use efficiently all resources (with underemployment, and imperfect allocation of capital), additional de mand does not lead only to inflation, but also to increased output. The stimulus effect of disasters Disasters lead to a reduction of production capacity, but also to an increase in the demand for the recon struction sector and goods. Thus, the reconstruction acts in theory as a stimulus. However, as any stimu lus, its consequences depend on the preexisting economic situation, or the phase of the business cycles. If the economy is in a phase of high growth, in which all resources are fully used, the net effect of a sti mulus on the economy will be negative, for instance through diverted resources, production capacity scarcity, and accelerated inflation. If the predisaster economy is depressed, on the other hand, the sti mulus effect can yield benefits to the economy by mobilizing idle capacities. This complex interplay be 13 tween business cycles and natural disasters economics is analyzed in detail in Hallegatte and Ghil (2008), who support the counterintuitive result that economies in recession are more resilient to the effects of natural disasters. This result appears consistent with empirical evidence. For instance, the 1999 earth quake in Turkey caused destructions amounting to 1.5 to 3% of Turkey's GDP, but consequences on growth remained limited, probably because the economy had significant unused resources at that time (the Turkish GDP contracted by 7% in the year preceding the earthquake). In this case, therefore, the earthquake may have acted as a stimulus, and have increased economic activities in spite of its terrible human consequences. In 1992 also, when hurricane Andrew made landfall on south Florida, the econo my was depressed and only 50% of the construction workers were employed (West and Lenze, 1994). The reconstruction needs had a stimulus effects on the construction sectors, which would have been imposs ible in a better economic situation. The productivity effect When a disaster occurs, it has been suggested that destructions can foster a more rapid turnover of capi tal, which could yield positive outcomes through the more rapid embodiment of new technologies. This effect, hereafter referred to as the "productivity effect" (also labeled "Schumpeterian creative destruc tion effect"), has been mentioned for instance by AlbalaBertrand (1993), Stewart and Fitzgerald (2001), Okuyama (2003) and Benson and Clay (2004). Indeed, when a natural disaster damages productive capi tal (e.g., production plants, houses, bridges), the destroyed capital can be replaced using the most recent technologies, which have higher productivities. Examples of such upgrading of capital are: (a) for house holds, the reconstruction of houses with better insulation technologies and better heating systems, al lowing for energy conservation and savings; (b) for companies, the replacement of old production tech nologies by new ones, like the replacement of paperbased management files by computerbased sys tems; (c) for government and public agencies, the adaptation of public infrastructure to new needs, like the reconstruction of larger or smaller schools when demographic evolutions justify it. Capital losses can, therefore, be compensated by a higher productivity of the economy in the event aftermath, with asso ciated welfare benefits that could compensate for the disaster's direct consequences. This process, if present, could increase the pace of technical change and accelerate economic growth, and could there fore represent a positive consequence of disasters. As an empirical support for this idea, AlbalaBertrand (1993) examined the consequences of 28 natural disasters on 26 countries between 1960 and 1979 and found that, in most cases, GDP growth increases after a disaster and he attributed this observation, at least partly, to the replacement of the destroyed capital by more efficient one. Using econometric ana lyses, Skidmore and Tuya (2002) find an increase in growth after disasters and suggest the productivity effect can explain this result.3 However, the productivity effect is probably not fully effective, for several reasons. First, when a disaster occurs, producers have to restore their production as soon as possible. This is especially true for small businesses, which cannot afford long production interruptions (see Kroll et al., 1991; Tierney, 1997), and in poor countries, in which people have no mean of subsistence while production is interrupted. Replac ing the destroyed capital by the most recent type of capital implies in most cases to adapt company or ganization and worker training, which takes time. Producers have thus a strong incentive to replace the destroyed capital by the same capital, in order to restore production as quickly as possible, even at the price of a lower productivity. In extreme case, one may even imagine that reconstruction is carried out 3 Interestingly, this result from Skidmore and Tuya (2002) is contradicted by other findings (e.g., Noy and Nualsri, 2007; Noy, 2009; Hochrainer, 2009; Jaramillo, 2009; Raddatz, 2009; see Section 4). 14 with lower productivity, to make reconstruction as fast as possible, with a negative impact on total prod uctivity. Second, even when destructions are quite extensive, they are never complete. Some part of the capital can, in most cases, still be used, or repaired at lower costs than replacement cost. In such a situa tion, it is possible to save a part of the capital if, and only if, the production system is reconstructed iden tical to what it was before the disaster. This technological "inheritance" acts as a major constraint to pre vent a reconstruction based on the most recent technologies and needs, especially in the infrastructure sector. This effect is investigated in Hallegatte and Dumas (2008) using a simple economic model with embodied technical change. In this framework, disasters are found to influence the production level but cannot influence the economic growth rate, in the same way than the saving ratio in a Solow growth model. Depending on how reconstruction is carried out (with more or less improvement in technologies and capital), moreover, accounting for the productivity effect can either decrease or increase disaster costs, but this effect is never able to turn disasters into positive events. Poverty traps It is crucial to also take into account the possibility that natural disasters increase poverty. In particular, because they destroy assets and wipe out savings, they can throw households into "poverty traps", i.e. situation in which their productivity is reduced, making it impossible for them to rebuild their savings and assets. These microlevel poverty traps can also be created by health and social impacts of natural disas ters: it has been shown that disasters can have longlasting consequences on psychological health (Norris, 2005), and on children development (from reducing in schooling and diminished cognitive abilities; see for instance Santos, 2007; Alderman et al., 2006). These poverty traps at the microlevel (i.e. the household level) could even lead to macrolevel poverty traps, in which entire regions could be stuck. Such poverty traps could be explained by the amplifying feedback reproduced in Fig. 3: poor regions have a limited capacity to rebuild after disasters; if they are regularly affected by disasters, they do not have enough time to rebuild between two events, and they end up into a state of permanent reconstruction, with all resources devoted to repairs instead of addition of new infrastructure and equipments; this obstacle to capital accumulation and infrastructure develop ment lead to a permanent disasterrelated underdevelopment. This effect has been analyzed by Halle gatte et al. (2007) with a reducedform model that shows that the average GDP impact of natural disas ters can be either close to zero if reconstruction capacity is large enough, or very large if reconstruction capacity is too limited (which may be the case in some least developed countries). 15 Limited reconstruction capacity Reduced economic Long reconstruction development Amplifying period after disasters feedback loop Reduced accumulation of Large economic cost of capital and infrastructure natural disasters Figure 3: Amplifying feedback loop that illustrates how natural disasters could become responsible for macrolevel poverty traps. This type of feedback can be amplified by other long term mechanisms, like changes in risk perception that reduces investments in the affected regions or reduced services that make qualified workers leave the regions. Because of these mechanisms, the consequences of a disaster can last much longer than what is considered the recovery and reconstruction period. 4. Assessment methodologies From this description of mechanisms and determinants of indirect losses, many scholars have attempted to assess these losses for real or potential events. To do so, they have mobilized various methodologies. Higherorder losses are sometimes measured using firm or householdlevel surveys (e.g., Kroll et al., 1991; Tierney, 1997; and Boarnet, 1998; Smith and McCarty, 2006). More frequently, they are estimated using economic models, including (i) microeconomic models at the household level (e.g., Dercon, 2004); (ii) econometrics models at the local level (e.g., Strobl, 2008) or the national level (AlbalaBertrand, 1993; Skidmore and Toya; 2002; Noy and Nualsri 2007); (iii) inputoutput (IO) models at the regional or national level (e.g., Gordon et al., 1998; Okuyama and Chang, 2004; Hallegatte, 2008); (iv) Computable General Equilibrium (CGE) models at the regional or national level (e.g., Rose et al., 1997; Rose and Liao, 2005, 2007; Tsuchiya et al., 2007); or (v) networkproduction system model, even though this line of research has not been operationalized yet (Haimes and Jiang, 2001; Henriet and Hallegatte, 2008). Available source of data for loss assessment are summarized in Annex 1. Data collection on past events A first line of assessment consists of data collection on past events. This approach considers a single event in a single location. For instance, using firmlevel surveys, Kroll et al. (1991), Tierney (1997), and Boarnet (1998) investigate the consequences of lifeline and transportation interruption of firm activity and survival for the Loma Prieta earthquake in 1989 and the Northridge earthquake in 1994. This line of 16 research is often concerned with the impact of network disruption and in particular transport disruption (commuting in particular) using transport model based on criticality assessment.4They found that the consequences of infrastructurerelated indirect impacts are often larger than the direct impact on firms. West and Lenze (1994) summarize the impact of hurricane Andrew on Florida, including job market con sequences. The Bureau of Labor Statistics (2006) of the US also provides a detailed analysis of the large labor market consequences of hurricane Katrina. Using household survey in three counties and 16 cities after the 2004 hurricane landfalls in Florida, Smith and McCarty (2006) show that households are more often forced to move outside the affected area by infrastructure problems than by structural damages to their home. The ECLAC methodology (UN ECLAC, 2003) assesses direct indirect and secondary effects of single events by a collection of data and information of various types (physical, monetary, and expert judgment) in each sector. Indirect costs are thus estimated through collection of information from economic agents, governments and experts, taking in consideration particular aspects such as transport disturbances cost, loss of opportunities etc. Of course, the aggregation of data from very different sources should be done with care to avoid gaps and double counting. Econometric approaches Econometric approaches are based on statistics and do not investigate a single event. They focus on se ries of events and investigate the "mean" indirect cost of these events, like the average impact on long term economic growth. Econometrics analyses at national scale have reached different conclusions on the impact of disasters on growth. AlabalaBertrand (1993) and Skidmore and Toya (2002) suggest that natural disasters have a pos itive influence on longterm economic growth, probably thanks to both the stimulus effect of reconstruc tion and the productivity effect (also labeled "Schumpeterian creative destruction effect" or productivity effect, described earlier in Section 4.b.). Others, like Noy and Nualsri (2007), Noy (2009), Hochrainer (2009), Jaramillo (2009), and Raddatz (2009), suggest exactly the opposite conclusion, i.e. that the overall impact on growth is negative. As suggested by Cavallo and Noy (2010) and Loayza et al. (2009), the dif ference between both conclusions may arise from different impacts from small and large disasters, the latter having a negative impact on growth while the former enhance growth. There are also examples of localscale econometric approaches. For instance, Strobl (2008) investigates the impact of hurricane landfall on countylevel economic growth in the US. This analysis shows that a county that is stuck by at least one hurricane over a year sees its economic growth reduced on average by 0.79 percentage point, and increased by only 0.22 percentage point the following year. On Vietnam, Noy and Vu (2009) investigate the impact of disasters on economic growth at the province level, and found that lethal disasters decrease economic production while costly disasters increase shortterm growth. GDP is not the only relevant indicator of disaster economic consequences, and RodriguezOreggia et al. (2009) focus instead on poverty and the World Bank's Human Development Index at the municipality level in Mexico. They show that municipalities affected by disasters see an increase in poverty by 1.5 to 3.6 percentage point. Looking at different economic variables, Hallstrom and Smith (2005) assess the 4 The Weather project is a 7FAP of the EU concerned with weather extremes and transport working more precisely on these issues. 17 impact of hurricane risk perception on housing values in Florida, and find that hurricane risks reduces property values by 19 percent. Modelbased approaches designed for cost assessment: inputoutput models, computable general equili brium models, and hybrid models The "adaptability" and "flexibility" of the production system and its ability to compensate for unavailable inputs is largely unknown and largely depend on the considered timescale. Over the very short term, the production system is largely fixed, and the lack of one input can make it impossible to produce. Moreo ver, over short timescales, local production capacity is likely to be highly constrained by existing capaci ties, equipments and infrastructure. Only imports from outside the affected region and postponement of some nonurgent tasks (e.g., maintenance) can create a limited flexibility over the shortterm. This is what is represented in economic InputOutput model (e.g., Rose and Miernyk, 1989; Haimes and Jiang, 2001; Okuyama, 2004; Rose and Liao, 2005; Haimes et al., 2005; Hallegatte, 2008; Hallegatte et al., 2011), in which producing one unit of output requires a fixed amount of all input categories. Over the longer term and the entire reconstruction period, which can stretch over years for largescale events, the flexibility is much higher: relative prices change, incentivizing production in scarce sectors; equipments and qualified workers move into the affected region, accelerating reconstruction and replac ing lost capacities; and different technologies and production strategies can be implemented to cope with longlasting scarcities. The production system organization can also be adjusted to the new situation: one supplier that cannot produce or cannot deliver its production (because of transportation issues, for in stance) can be replaced by another suppliers; new clients can be found to replace bankrupt ones; slightly different processes can be introduced to reduce the need for scarce inputs (e.g., oilrunning backup gene rator can be installed if electricity availability remains problematic). These types of substitution are represented in Calculable General Equilibrium models (e.g., Rose et al., 2007), in which the scarcity of one input translates into higher price, and reduced consumption of this input, compensated by larger consumption of other inputs. IO models are often considered too pessimistic, since they assume that no substitution can take place in the production system. CGE models are on the opposite considered as too optimistic, because realworld prices have stickiness and cannot adjust perfectly, and because substitution has technical limits that are not always adequately represented in production functions.5 The reality probably lies somewhere in between these two extremes, prompting the work on interme diate models. These intermediate models are either IO models with flexibility like in Hallegatte (2008), CGE models with reduced substitution elasticity like in Rose et al. (2007), or I/OCGE hybrid with bottom characteristics such as Hottidge et al. (2005) TERM Model for analyzing Australian drought 20002003. These models emphasize the importance of infrastructure (see, e.g., Haimes and Jiang, 2001). For in stance, transportation is crucial and losses in this sector have large ripple effects on the rest of the econ omy, as shown by Gordon et al. (1998). In their analysis of the Northridge earthquake, they used the Southern California Planning Model (SCPM) and found that a substantial share of business interruption 5 Price elasticity in CGE can also be seen as an artificial way to model input substitution, model prices being proxies for scarcity in each sector. The fact that real-world prices do not react like model prices does not mean that this "trick" is not useful to model substitution in scarcity situation. 18 were due to offsite problems, such as disruptions in the transportation system that restricted the movement of goods and employees. Tsuchiya et al. (2007) reached the same conclusion in simulated TokaiTonankai earthquakes in Japan, applying a spatial CGE. The water and electricity sectors also play a significant role. Rose et al. (2007) analyzed the impact of a potential twoweek total blackout due to a potential terrorist attack in Los Angeles, i.e. an approximate loss of production worth $250 million, and finds that the total cost would be about $13 billion, decreas ing to $2.8 billion if extensive production rescheduling is possible at low cost. For the landfall of Katrina on New Orleans, the availability of a large amount of data allowed many mod eling analyses. Hallegatte (2008), for instance, estimated using a regional inputoutput model that indi rect economic losses in Louisiana after Katrina amounted to $42 billion compared to $107 billion of direct economic losses. More generally, this analysis concludes that regional indirect losses increase nonlinearly with direct losses, suggesting the existence of threshold in the coping capacity of economic systems. In this analysis of Louisiana, indirect losses remain negligible (or even negative) for direct losses below $50 billion, and then increase nonlinearly to reach $200 billion for direct losses of the same amount (see Fig. 4). When direct losses are lower than $50bn, aggregated indirect losses are close to zero. The aggregation may however hide important disparities among sectors and among social categories. They do not de scribe welfare change for some sectors, geographical zones or segments of the population. Some of these categories can be heavily affected, as is illustrated by much evidence from the landfall of Katrina in New Orleans. This aggregation risk highlights the importance of precisely defining the scope and purpose of the assessment, as stated in Section 2.2. In Hallegatte (2008), indirect losses decrease rapidly if it is possible to "import" reconstruction means (workers, equipment, finance) from outside the affected region. This result highlights the importance of taking into account interregional interactions. This result is moreover reinforced by Okuyama (2004), who applies a Miyazawa's extension to the conventional inputoutput framework and the sequential inter industry model (SIM), introduced by Romanoff and Levine (1977), to assess the regional indirect cost of the Great Hanshin Earthquake in the Kinki region in Japan and the interregional impacts. He finds that most of the cost arises from income losses in the rest of Japan, due to reduced export to the Kinki region, showing that impacts outside the directly affected region cannot be disregarded. 19 Figure 4: The direct losses ­ indirect (output) losses as a function of direct (asset) losses, in Louisiana for Katrinalike disasters. (source: Hallegatte, 2008) Other modelbased approaches: idealized models, hybrid models, public finance coping capacity Idealized models aim at emphasizing mechanisms at play in natural disasters aftermath. Even though their aim is not directly to assess the costs of extreme events, they help identify important mechanisms and investigate their role (e.g., Hallegatte and Dumas, 2008 for the role of endogenous technical change; Hallegatte and Ghil, 2008, for the interaction with business cycles). In some hazards community, e.g. in the drought literature, the physical impacts and economic impacts are so closely linked that hybrid models combine physical aspects and economics. This is the case in the hydrological ­ economic model in Booker (1995) and the Biophysicalagroeconomic model in Holden and Shiferaw (2004). This methodology of cost assessment is probably not easily replicable outside the aca demic community but they provide an important scoping aspect and help identify the most important mechanisms. Another important line of research aims at assessing the impact of natural disasters on public finances. It aims at assessing the capacity of governments to cope with large expenses due to disasters and their subsequent ability to deliver basic services. Example of this approach is the IIASA CATSIM model devel oped in Mechler et al. (2006). 5. Conclusion and knowledge gaps This paper highlights the main difficulties in defining, measuring, and predicting the total cost of disas ters. It focuses on indirect (or output) losses, considered as a major component of the total loss of wel fare. This paper supports the view that the economic assessment of indirect losses is central to the sys temic appreciation of welfare. There are several methodologies to assess these indirect losses, they are all based on questionable assumptions and modeling choices, and they can lead to very different results. The main conclusion is twofold. 20 First, it is impossible to define "the cost" of a disaster, as the relevant cost depends largely on the pur pose of the assessment. The best definition and method obviously depend on whether the assessment is supposed to inform insurers, prevention measure costbenefit analyses, or international aid providers. A first lesson from this article is that any disaster cost assessment should start by stating clearly the pur pose of the assessment and the cost definition that is used. Following this recommendation would avoid misleading use of assessments, and improper comparison and aggregation of results. Depending on the purpose of the assessment, the relevant definition of the indirect cost is different, and the most ade quate methodology may also change. It suggests then that different approaches are not only necessary but also interconnected. Indeed, for instance, a better understanding of the roles of transport disruption in indirect costs may be crucial to refine modeling approaches, and to understand substitution and inter modality issues. Second, there are large uncertainties on indirect disaster costs, and these uncertainties arise both from insufficient data and inadequate methodologies. Considering the importance of unbiased estimates of disaster cost, for instance to assess the desirability of prevention measures, progress in this domain would be welcome and useful. To do so, much more research should be devoted to this underworked problem. Four main issues for future research are suggested: The understanding of the economic response to external shocks, i.e. how the economic system can react and adapt in the recovery and reconstruction phase. This research would in particular include a better understanding of how markets function outside equilibrium, and of how agent expectations are formed in situations of high uncertainty. Indeed, natural disasters are situation in which abnormal solidarity and assistance, different governance and political processes take place. These are not taken into account in current literature. A next step may be to introduce progress made in economics about the definition of the preferences of an economic agent and their evolution according to circumstances, such as work put forward by Fehr (1998) and pre sented in survey by Bowles and Gintis (2008). The understanding of interactions between the economic intrinsic dynamics (e.g., business cycles and financial crises) and external shocks (e.g., natural disasters), as analyzed in Hallegatte and Ghil (2008). The coexistence of these two dynamics explains why it is so difficult to "extract" the effect of natural disasters from macroeconomic data series (Strobl, 2008; Noy, 2009). A better understanding of their interactions would allow for a better measurement of disaster cost and for a better understanding of relevant processes. The role of networks has been highlighted in the literature but requires additional work: specific networkshaped economic sectors (e.g., electric system, water distribution, transportation) are especially important, but other sectors also involve network through the organization of supply chains. It is crucial to understand how failure in one business or production unit translates into operational problems for its clients (because of rupture in production input) and its suppliers (because of the reduction in demand). Network structures may play a role in the vulnerability of the economic system (e.g., having fewer suppliers may increase the vulnerability of a business) and analysis at the sectorscale may reveal insufficient to understand it (e.g., Haimes and Jiang, 2001; Henriet and Hallegatte, 2008). 21 Financial aspects also play a role: households and company may delay (or give up) reconstruction because of insufficient financial resources, households may reduce final demand because of lost income and assets, company may go bankrupt even if their production capital is still partly opera tional (e.g., Benson and Clay, 2004; Tierney, 1997). These problems are especially important in developing countries. In general, this research line should also touch upon the role of the insur ance industry (and of its regulation) and its capacity to help fund reconstruction and reduce natu ral disaster indirect cost. 6. Acknowledgment The authors want to thank Celine Guivarch, Julie Rozenberg, and partners from the ConHaz project for their useful comments and suggestions on an earlier version of the paper. This work benefited from fi nancial support from the ConHaz project, a 7th Framework research project, funded by the European Union. This work benefited from many exchanges during an author's visit at, and funded by, the World Bank in February 2010. 7. References AlbalaBertrand, J. M. (1993), The Political Economy of Large Natural Disasters with Special Reference to Developing Countries, Oxford: Clarendon Press. Alderman, H., J. Hodditnott and B. Kinsey (2006), Longterm Consequences of Early Childhood Malnutri tion, Oxford Economic Papers 58, 450­474. Boarnet, M. G., 1998. Business losses, transportation damages, and the Northridge earthquake. Journal of Transportation and Statistics 1, 49­63. Booker, J.F. (1995). Hydrologic and economic impacts of drought under alternative policy responses. Wa ter Resources Bulletin, 31: 889906. Benson, C., & Clay, E. (2004). Understanding the economic and financial impacts of natural disasters. Dis aster risk management series no. 4. Bowles S. et Gintis H., L'idéal d'égalité appartientil au passé ?. Homo reciprocans versus Homo oecono micus, Revue du Mauss 2008/1, n° 31, p. 209241. Brown Gaddis, E., Miles B., Morse S., Lewis D.. (2006). Fullcost accounting of coastal disasters in the United States: Implications for planning and preparedness. Ecological Economics 63 (2007) 307318. Cavallo, E., & Noy, I. (2009). The economics of natural disasters ­ a survey. InterAmerican Development Bank Working Paper #124. Cochrane, H. (2004), "Economic Loss: Myth and Measurement", Disaster Prevention and Management 13, 290­296. 22 Council and European Parliament, 2009, GDP and Beyond. Measuring Progress in a Changing World, COM(2009) 433, Brussels, 20.8.2009 CMEPSP, 2009. Report of the Commission on the Measurement of Economic Performance and Social Progress, September, 14, 2009 Dercon, S. (2004), "Growth and Shocks: Evidence from Rural Ethiopia", Journal of Development Econom ics 74, 309­329. Fehr, E (1998), « Reciprocity and economics : the economic implications of "Homo reciprocans" », Euro pean Economic Review, n° 42, p. 845859. Freeman, P. K., L. A. Martin, J. LinneroothBayer, R. Mechler, S. Saldana, K.Warner and G. Pflug (2002), Financing Reconstruction: Phase II Background Study for the InterAmerican Development Bank Regional Policy Dialogue on National Systems for Comprehensive Disaster Management, Washington DC: Inter American Development Bank. Flegg A.T. and C.D. Webber, 1997. On the appropriate use of location quotients in generating regional inputoutput tables: reply. Regional Studies 31, pp. 795805 Flegg A.T. and T. Tohmo, 2010. Regional InputOutput models and the FLQ formula: a case study of Fin land, Regional Studies (in revision) FLOODsite, 2009. Flood risk assessment and flood risk management. An introduction and guidance based on experiences and findings of FLOODsite (an EUfunded Integrated Project). Deltares | Delft Hydraulics, Delft, the Netherlands Gordon, P., Richardson, H., & Davis, B. (1998). Transportrelated impacts of the northridge earthquake. Journal of Transportation and Statistics, 1, 21­36 Government of Queensland, Guidance on the Assessment of Tangible Flood Damages, Sep 2002, The State of Queensland (Department of Natural Resources and Mines) 2002 Haimes, Y., Horowitz, B., Lambert, J., Santos, J., Lian, C., & Crowther, K. (2005). Inoperability inputoutput model for interdependent infrastructure sectors. I: Theory and methodology. Journal of Infrastructure Systems, 11, 67­79. Haimes, Y., & Jiang, P. (2001). Leontiefbased model of risk in complex interconnected infrastructures. Journal of Infrastructure Systems, 7, 1­12. Hallegatte, S. (2008), "An Adaptive Regional InputOutput Model and Its Application to the Assessment of the Economic Cost of Katrina", Risk Analysis 28, 779­799. Hallegatte S., J. C. Hourcade and P. Dumas (2007), Why Economic Dynamics Matter in Assessing Climate Change Damages: Illustration on Extreme Events", Ecological Economics 62, 330­340. 23 Hallegatte, S. and P. Dumas (2008), "Can Natural Disasters Have Positive Consequences? Investigating the Role of Embodied Technical Change", Ecological Economics 68, 777­786. Hallegatte, S. and M. Ghil (2008), "Natural Disasters Impacting a Macroeconomic Model with Endogen ous Dynamics", Ecological Economics 68, 582­592. Hallegatte S, Ranger N, Mestre O, Dumas P, CorfeeMorlot J, Herweijer C, MuirWood R (2011), Assessing climate change impacts, sea level rise and storm surge risk in port cities: a case study on Copenhagen. Climatic Change, in press. doi:10.1007/s10584-010-9978-3 432 Hallstrom, D.G., and V. K. Smith, 2005. Market responses to hurricanes, Journal of Environmental Eco nomics and Management 50 (2005) 541­561 Henriet, F. and S. Hallegatte, 2008. Assessing the Consequences of Natural Disasters on Production Net works: a Disaggregated Approach, FEEM Working Paper No. 100.2008 Hochrainer, S. (2009), Assessing the Macroeconomic Impacts of Natural Disasters ­ Are There Any?, World Bank Policy Research Working Paper 4968. Holden, S., and Shiferaw, B. (2004). Land degradation, drought and food security in a lessfavoured area in the Ethiopian highlands: a bioeconomic model with market imperfections. Agricultural Economics, 30: 3149 Horridge, M., Madden, J., Wittwer, G., 2005. The impact of the 2002­2003 drought on Australia. J. Policy Modeling 27, 285­308. Jaramillo, C. R. H. (2009), Do Natural Disasters Have LongTerm Effects on Growth?, Universidad de los Andes, mimeo. Jonkman S.N., Bockarjova M., Kok M., Bernardini P., 2008. Integrated hydrodynamic and economic mod elling of flood damage in the Netherlands, Ecological Economics Volume 66, Issue 1, 15 May 2008, Pages 7790 Kroll, C. A., J. D. Landis, Q. Shen and S. Stryker (1991), Economic Impacts of the Loma Prieta Earthquake:A Focus on Small Business, Studies on the Loma Prieta Earthquake, University of California, Transportation Center. Lindell, M. and Prater, C. (2003) "Assessing Community Impacts of Natural Disasters" Natural Hazards Review 4(4) pp. 176185 Loayza, N., E. Olaberria, J. Rigolini and L. Christiansen (2009), Natural Disasters and GrowthGoing beyond the Averages, World Bank Policy Research Working Paper 4980. McCarty,C and S. K. Smith (2005), Florida's 2004 Hurricane Season: Local Effects, Florida Focus, University of Florida, http://www.bebr.ufl.edu/system/files/FloridaFocus1_3_2005_0.pdf. 24 Mechler, R., J. LinneroothBayer and D. Peppiatt (2006), Microinsurance for Natural Disasters in Develop ing Countries: Benefits, Limitations and Viability, ProVention Consortium, Geneva, http://www.proventionconsortium.org/themes/default/pdfs/Microinsurance_study_July06.pdf. Meyer V. & Messner, F., 2005. National Flood Damage Evaluation Methods. A Review of Applied Methods in England, the Netherlands, the Czech Republic and Germany. Discussion Papers, Umweltforschungsze trum Leipzig­Halle, Department Ökonomie, Leipzig: 47 pp. Norris, F. H. (2005), Range, Magnitude, and Duration of the Effects of Disasters on Mental Health: Review Update 2005, Dartmouth Medical School and National Center for PTSD, Hanover and Boston. Noy, I., & Nualsri, A. (2007). What do exogenous shocks tell us about growth theories? University of Ha waii Working Paper 0728. Noy, I. (2009), The Macroeconomic Consequences of Disasters, Journal of Development Economics 88, 221­231. Noy, Ilan, and Tam Vu, 2009. The Economics of Natural Disasters in Vietnam. University of Hawaii Work ing paper 0903. NZ Climate Change Office, 2003. Climate Change Case Study: Assessment of the impacts of sea level rise on floodplain management planning for the Avon river, Harris Consulting Ltd in conjunction with Christchurch City Council Okuyama, Y. (2004). Modeling spatial economic impacts of an earthquake: Inputoutput approaches. Dis aster Prevention and Management, 13 (4), 297­306. Okuyama, Y., Hewings, G., & Sonis,M. (2004). Measuring the economic impacts of disasters: Interregional inputoutput analysis using the sequential interindustry model. In Y. Okuyama & S. Chang (Eds.), Model ing Spatial and Economic Impacts of Disasters. Heidelberg: Springer. Pelling, M., A. Özerdem and S. Barakat (2002), "The Macroeconomic Impact of Disasters", Progress in Development Studies 2, 283­305. Raddatz C. (2009), The Wrath of God: Macroeconomic Costs of Natural Disasters,World Bank Policy Re search Working Paper 5039. Ranger N., S. Hallegatte, S. Bhattacharya, M. Bachu, S. Priya, K. Dhore, F. Rafique, P. Mathur, N. Naville, F. Henriet, C. Herweijer, S. Pohit, J. CorfeeMorlot, 2011. A Preliminary Assessment of the Potential Impact of Climate Change on Flood Risk in Mumbai, Climatic Change, in press, doi:10.1007/s10584-010-9979-2 RodriguezOreggia, Eduardo, Alejandro de la Fuente, Rodolfo de la Torre, Hector Moreno, and Cristina Rodriguez (2009). The Impact of Natural Disasters on Human Development and Poverty at the Municipal Level in Mexico. Manuscript 25 Romanoff, E. and Levine, S.H. (1977), "Interregional sequential interindustry modeling: a preliminary analysis of regional growth and decline in a two region case", Northeast Regional Science Review, Vol. 7, pp. 87101 Rose, A. and Miernyk, W., 1989. Inputoutput analysis: The first fifty years. Economic System Research 1, pp. 229­271. Rose, A. (2004), "Economic Principles, Issues, and Research Priorities in Hazard Loss Estimation", in: Okuyama,Y. and S. Chang (eds.), Modeling Spatial and Economic Impacts of Disasters, Berlin: Springer, 14­ 36. Rose, A., Benavides, J., Chang, S. E., Szczesniak, P., & Lim., D. (1997).The regional economic impact of an earthquake: Direct and indirect effects of electricity lifeline disruptions. Journal of Regional Science, 37, 437­458. Rose, A., & Liao, S.Y. (2005). Modeling regional economic resilience to disasters: A computable general equilibrium analysis of water service disruptions. Journal of Regional Science, 45, 75­112. Rose, A., G. Oladosu and S. Y. Liao (2007), "Business Interruption Impacts of a Terrorist Attack on the Elec tric Power System of Los Angeles: Customer Resilience to a Total Blackout", Risk Analysis 27, 513­531. Santos, I. (2007) Disentangling the Effects of Natural Disasters on Children: 2001 Earthquakes in El Salva dor, Doctoral Dissertation, Kennedy School of Government, Harvard University. Skidmore, M. and H. Toya (2002), Do natural disasters promote longrun growth?, Economic Inquiry 40, 664­687. Stewart, F., Fitzgerald, E.V.K., 2001. War and Underdevelopment. Oxford University Press, Oxford. Strobl, E. (2008), The economic growth impact of hurricanes: evidence from U.S. coastal counties, IZA Discussion Papers 3619. Tierney, K., (1997), "Business Impacts of the Northridge Earthquake", Journal of Continencies and Crisis Management 5, 87­97. Tsuchiya, S., Tatano, H., & Okada, N. (2007). Economic loss assessment due to railroad and highway dis ruptions. Economic Systems Research, 19 (2), 147­162. United Nations Economic Commission for Latin America and the Caribbean (UN ECLAC, 2003), Handbook for Estimating the SocioEconomic and Environmental Effects of Disasters, Santiago, Chile. West, C. T. and D. G. Lenze (1994), "Modeling the Regional Impact of Natural Disasters and Recovery: A General Framework and an Application to Hurricane Andrew", International Regional ScienceReview 17, 121­150. Wilhite, D.A., Svoboda, M.D., Hayes, M.J. (2007). Understanding the complex impacts of drought: a key to enhancing drought mitigation and preparedness. Water Resources Management, 21: 76374 26 Wilhite, D.A., Vanyarkho, O. (2000). Pervasive impacts of a creeping phenomenon. In Wilhite, D.A. (ed.). Drought: A Global Assessment, 1: 24555. London/New York: Routledge 27 Annex 1 Data available for c the assessment indirect impacts Available data on the economic cost of disasters The emergency Events Database (EMDAT) maintained by the Center for Research on the Epidemiology of Disasters (CRED) at the Catholic University of Louvain, Belgium (http://www.emdat.be) is an important source of publicly available data on natural disasters. This database is compiled from diverse sources such as UN agencies, NGOs, insurance companies, research institutions and press agencies. EMDAT defines a disaster as a natural situation or event which overwhelms local capacity and/or necessi tates a request for external assistance. For a disaster to be listed in the EMDAT database, at least one of the following criteria should be met : 10 or more people are reported killed 100 people are reported affected a state of emergency is declared a call for international assistance is issued Disasters can be hydrometeorological, including floods, wave surges, storms, droughts, landslides and avalanches; geophysical, including earthquakes, tsunamis and volcanic eruptions; and biological, covering epidemics and insect infestations (these are much more infrequent in this database). The amount of damage reported in the database consists only of direct damages (e.g., damage to infra structure, crops, and housing). The data report the number of people killed, the number of people af fected, and the dollar amount of direct damages in each disaster. Reinsurance companies surveys are also an extensive source of data but two main issues are met : - the data are not publicly available, or only in a really aggregated fashion - reinsurance companies are collecting data based on the losses insured, so biased towards coun tries in which insurance is well developed. Data needs for econometric analysis For econometric analysis, long time series with sufficient numbers of events are required to have an ade quate sample. Three main problems are met. First collection of data on natural disasters started to be precise and of large scope in the 90s. Earlier data collection is mainly available for large events, and data are less precise. Second, large natural disasters are, fortunately, scarce. This does not allow for a large sample necessary for econometric analysis. Third, the socioeconomic baseline may be an important difficulty to distinguish effects of natural disas ters from effects due to socio economic evolution. For instance, effects of natural disasters on economic growth may be difficult to disentangle from macroeconomic shocks, change in population, trends in ur banization, development of utilities networks etc. Data needs for modelbased approaches Modelbased approaches require large set of data (e.g., inputoutput tables) on the affected economy, and detailed data on which sectors are affected by the disaster. At regional scale, it is particularly rare to 28 have IO tables (e.g., in the US, the BAE provides data at the state level), and table reconstruction using simple rules is sometimes necessary (see, e.g., Flegg et al., 1995; Hallegatte, 2008; Flegg et Tohmo, 2010). The indirect impacts of a disaster largely depend on direct losses, and data on direct losses are often broadly aggregated and rarely disaggregated by sector. Also, data are often based on insurance industry data, and focus therefore on insurable goods. Noninsurable goods, and especially infrastructures, are often absent from these estimates. Again, simple methods may be used to recreate this data, but these data reconstruction methods add an additional large source of uncertainty in the assessment. Examples of methods are: Sector disaggregation can be carried out using the sector values added, assuming that each sector is af fected proportionally: if a sector is twice as large (in terms of VA or productive capital) than another, then it would have suffered from losses that are twice as large (Hallegatte, 2008). If sectorscale data on realevent losses are not available, models can be used to assess direct losses. Examples of models are catastrophe models from risk modelling companies (e.g., RMS, EQECat, Air Worldwide). This method can be applied only for events and regions in which risk modelling is available, i.e. mainly rich countries (e.g., hurricanes in the U.S.). Infrastructure losses can be estimated assuming that the ratio between insurable and noninsurable losses is the same for all events. Using cases in which this information is know (e.g., Katrina in New Or leans), infrastructure losses can be reconstructed for other events (see Hallegatte et al., 2010, on coastal floods in Copenhagen). 29