Environmental Modelling & Software 72 (2015) 44e55
Contents lists available at ScienceDirect
Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft
A risk management tool for tackling country-wide contingent disasters: A case study on Madagascar Stefan Hochrainer-Stigler*, Reinhard Mechler, Junko Mochizuki IIASA e International Institute for Applied Systems Analysis, Schlossplatz 1, A-2361 Laxenburg, Austria
a r t i c l e i n f o
a b s t r a c t
Article history: Received 19 February 2015 Received in revised form 9 June 2015 Accepted 10 June 2015 Available online xxx
Governments are key players in managing disaster risks, and fiscal risk management has become an integral part of disaster risk management. However, the ability of governments to implement disaster risk management strategies differs significantly across countries, depending on their capacity and resource constraints. The CatSim model and its most recent version presented in this paper helps to fill part of the information gap regarding the capacity and resources of a government to deal with natural disaster risk. We provide an in-depth example how the model works for the case of managing cyclone risk in Madagascar. In doing so, we provide recommendations as to how some of the more difficult concepts from the disaster risk theory and modelling field can be most easily understood by nontechnically trained stakeholders. Such understanding is beneficial in facilitating consensus-building among various risk bearers from different sectors regarding options for managing risk. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Natural catastrophes Economic impacts Financial vulnerability Disaster risk management Catastrophe modeling CatSim
Software availability Name of software: CatSim Developer: Hochrainer-Stigler, Stefan; Mechler, Reinhard; Pflug, Georg Year first available: June 2005, Latest Version: September 2013 Software required: Microsoft Windows XP, 2007 Programming language: Matlab and Cþþ, Stand alone application available Program size: 1.225 MB Availability: By contacting the developer Cost: Free 1. Introduction The number and losses of natural disasters have been increasing globally due to factors such as increases in wealth, population growth and rural-to-urban migration (Swiss Re, 2012). Whereas developed countries are better able to cope with natural disaster impacts, less developed countries often have a large proportion of the population severely affected and face a substantial strain due to * Corresponding author. E-mail addresses:
[email protected] (S. Hochrainer-Stigler),
[email protected]. at (R. Mechler),
[email protected] (J. Mochizuki). http://dx.doi.org/10.1016/j.envsoft.2015.06.004 1364-8152/© 2015 Elsevier Ltd. All rights reserved.
their limited resources and ability to finance important social and economic programs (World Bank, 2010a,b). Disaster losses have historically been financed by diversions from the budget, allocated loans and donations from the international community (Mechler, 2004). However, gaps in necessary post-disaster financing for recovery and reconstruction have been frequently encountered (Global Assessment Report, 2013). More emphasis is hence placed on financial planning prior to events and on preparing for, and mitigating, potential losses through financial planning and risk management (Lavell et al., 2012). Governments are seen as key players in managing disaster risks (IPCC, 2012) and fiscal risk management has become an integral part of disaster risk management. However, the ability of governments to implement disaster risk management strategies differs significantly across countries, depending on their capacity and resource constraints (Lal et al., 2012). The CatSim (Catastrophe Simulation) model presented in this paper fills the information gap regarding the capacity and resources of a government to deal with natural disaster risk. It evaluates the government's financial disaster risk management strategies by illustrating the tradeoffs and choices it must make to manage disaster risks. The model assesses a government's contingent disaster obligations and the potential financial shortfalls and the costs and benefits of vulnerability-reducing options including financial management strategies. In addition, the model outlines the economic impacts of
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
natural disasters via a simple economic growth framework to account for the macroeconomic impacts due to natural disasters and the costs and benefits of measures for reducing such impacts thereof. It is equipped with a graphical user interface to facilitate decision support in an interactive and iterative moded a feature that is now seen as an essential part within risk management processes (IPCC, 2012). Over the last eight years, the user interfaces as well as the underlying model framework were tested, changed and improved on various levels based on more than 20 high-level stakeholder workshops held in countries such as Turkey, Philippines, India, Nepal, Mexico, the Caribbean countries, and most recently in Madagascar in 2012 and 2015. Furthermore, the CatSim model approach proved to be broad enough to be applied consistently on the global level (Mechler et al., 2009) while enabling the comparison of risk levels across countries too (Hochrainer, 2006; GAR, 2013). This article presents the latest model version and shows an application to Madagascar.1 While the full model requirements and technical details can be found in Hochrainer-Stigler (2014), this article focuses on the case study application and explains how CatSim informs real world policy-making, what implementation issues have been encountered in stakeholder sessions, and what solutions it has provided. In doing so, we illustrate how some of the more difficult concepts used within the disaster risk management field such as probability and path dependency can easily be understood by non-technically trained audience. We also show that unlike other macroeconomic modeling techniques to estimate country level impact of natural disaster such as the InputeOutput (IO) and Computable General Equilibrium (CGE) Modeling (Rose, 2004), CatSim has a non-optimization-based assessment tool for public budget allocation and allows different policy audience to test alternatively scenarios in an easy-to-understand manner. Such understanding is beneficial in facilitating consensus building among stakeholders from different sectors about the underlying nature of the current risk and ways to reduce it. While the concept of iterative risk management has gained academic and policy interest in recent years (IPCC, 2012), the CatSim implementation presented here can be viewed as one practical example of iterative risk management in which country-level risk has been assessed iteratively with a cross-sectoral policy audience when new information has become available. There are numerous starting points to address disaster risk (Pflug €misch, 2007) and we chose to start with Fig. 1 from and Ro Bettencourt et al. (2006) which focuses specifically on natural hazard risks from a topedown approach. Fig. 1 show important steps that are needed to “mainstream” disaster risk into development planning processes, including risk assessment as well as decision to accept or to lower risk to acceptable levels using various means. Risk management strategies, in principle, can and should include topedown as well as bottomeup approaches (Lal et al., 2012; IPCC, 2012); however the CatSim model is usually applied for topedown strategy assessments only as the government is assumed to be a key actor for setting funding priorities (for bottomeup application of the CatSim model see Mechler et al., 2009). Therefore, we limit our discussion to the level of national development plans. The paper is organized as follows. The next section gives a short outline of the CatSim modeling approach and the different steps the model follows. Section three then presents the detailed application of the model for Madagascar. Section four discusses the lessons learned from the stakeholder engagement. Finally, section
1 The overall description of the framework and modeling approach can be found in Hochrainer (2006); Mechler et al. (2006); Hochrainer and Mechler (2009); Hochrainer et al. (2013).
45
Fig. 1. Mainstreaming risk into development planning processes. Source: Bettencourt et al., 2006.
five ends with a conclusion and outlook to the future. 2. The CatSim approach/program The CatSim approach and model presented here consists of five steps as described below and illustrated in Fig. 2. The blue area in Fig. 2 represents the ‘static’ risk assessment (i.e. assessment of financial vulnerability and risk for the next year) while the green area represents the dynamic risk assessment (i.e. assessment of costs and benefits of different risk management strategies over a given time horizon). Step 1 (direct risk estimates) derived from hazard, exposure and (physical) vulnerability information is combined with steps 2 (economic and financial resilience) and 3 (economic vulnerability) to give the estimate of current risk. Steps 4 (probabilistic fiscal and macroeconomic impacts) and 5 (risk management options) calculate the consequences of disasters to future growth paths and investigates risk management options. This dynamic risk assessment evaluates future outlook over 2 to 10 (or more) years. Note, as climate change will likely alter hazard intensities and frequencies in the future, and as global changes will likely affect exposure and vulnerabilities too, such dynamics have to be taken explicitly into account in risk assessments. Table 1 describes the 5 different steps used in the CatSim model in a nutshell. As CatSim focuses on the national level decision-making, neither input nor output are geospatial in nature. Generally, outputs are provided in the form of interactive graphs where end-users may change parameters and perform sensitivity analysis. The consecutive steps adopted in the CatSim framework are most useful in introducing the notion of direct and indirect risk which is one key aspect for selecting appropriate disaster management strategies, e.g. indirect risks can be significant and neglect of it could severely underestimate the full consequences (such as development) due to disasters (Hallegate, 2014; Hochrainer and Mechler, 2009). The graphical user interface is also an important model component, which allows users to change important parameters and assumptions. Through visualization, policy-makers may study the consequences of disaster and risk reduction interventions in the stand-alone software. In terms of programming, the CatSim software is structured around modules to increase flexibility. Each module can be extended separately dependent on country circumstances (see Hochrainer-Stigler, 2014 for more details). As already indicated, there are two main modules in the software program/interface represented by the two areas in Fig. 2: the first module allows for the assessment of financial vulnerability and risk for the next year (static risk assessment in blue), the second studies the costs and benefits of
46
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
Fig. 2. The CatSim approach. Based on Hochrainer and Mechler 2013.
Table 1 CatSim model flow. Phases
Steps
Description
Data source
Static Risk Assessment (assessment of current risk)
Step 1:
The risk of direct asset losses expressed in terms of their probability of occurrence and destruction in monetary terms is modeled as a function of hazard (frequency and intensity), the elements exposed to those hazards and their physical vulnerability (susceptibility to physical damage). The financial preparedness of the public and private sectors to the direct losses is assessed. Financial preparedness is a measure of financial resilience and can be defined as the access of the state or central government to funds for financing reconstruction of public infrastructure and the provision of relief to households and the private sector. Financial preparedness will, in turn, depend on the general economic conditions of the country. Financial vulnerability, measured in terms of the potential resource gap, is assessed by simulating the risks to the public sector and the financial resilience of the government to cover its post-disaster liabilities following disasters of different magnitudes. The consequences of a resource gap on the macroeconomic development of the country are characterized with indicators, such as economic growth or the country's external debt situation. These indicators represent consequences to economic flows as compared to consequences to stocks addressed by the asset risk estimation in step 1. Strategies are developed and illustrated that build financial resilience of the public sector. The development of risk financing strategies has to be understood as an adaptive process, where measures are continuously revised after their impact on reducing financial vulnerability and risk has been assessed within the modeling framework.
Country level probabilistic loss information must be obtained from forward-looking or backward-looking risk assessment*.
Step 2:
Step 3:
Dynamic Risk Assessment (assessment of future DRR management options and economic impacts)
Step 4:
Step 5:
Country level fiscal preparedness data must be gathered through available database such as the World Development Indicators and national level statistical publication.
N/A (this step combines data of Step 1 & 2)
Parameters for the macroeconomic module must be estimated based on available information regarding past economic output, capital input and population. For loan conditions, discount rates and potential conditions for insurance must be gathered from multiple sources including existing literature and local interviews.
Note: *Forward-looking assessment refers to probabilistic risk assessment based on hazard, exposure and vulnerability as performed using available tools such as the CAPRA (Comprehensive Approach to Probabilistic Risk Assessment) (Cardona et al., 2012). Backward-looking assessment uses past disaster data and estimates existing risk using the extreme value statistical modeling. In this article, we used risk data derived from backward-looking assessment (Kull et al., 2013). Source: The authors.
different risk management strategies over a given time horizon (dynamic risk assessment in green). We consider the user interface a crucial component, as it allows stakeholders to understand and deal with the considerable
uncertainty associated with economic and catastrophe parameters. As the problem of government risk financing involves a trade-off between stability and growth (Hochrainer and Mechler, 2009) users therefore should interactively decide which trade-off they are
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
willing to commit to, and which indicators they consider most useful in analyzing such trade-off. Given stakeholders have different opinions regarding which dimensions are important to them, an open dialog can be established based on these diverse considerations. A full CatSim step-by-step example and guidelines for using all the user-interface and stand-alone application software with complete explanations of variables and estimation methods can be found in Hochrainer-Stigler, 2014. The next section sets out a concrete and detailed example for Madagascar where the latest version of CatSim was used. 3. Case study: Madagascar Madagascar has one of the highest cyclone risks worldwide and has been confronted with a series of economic challenges (see Thomas, 2011 for a summary). Poverty remains a major challenge, with approximately 77 percent of the population living below the poverty line in 2010. Additionally, due to a number of factors, including political instability, GDP per capita has effectively been reduced over the last decades and now amounts to about half of its level in 1960 (World Bank, 2011). The current political situation has contributed to the nation's economic challenges (CIA Factbook, 2012). Madagascar's economy is mainly based on subsistenceagriculture, fishery and forestry, with the primary sector employing over 80 percent of the active population. The primary sector contributes to about 30 percent of the total national GDP and 60e65 percent of the national export revenues. Until 2010 when regulatory complications abruptly halted these exports, leading to a sharp decline in production (World Bank 2012; CIA Factbook, 2012), the export of clothing was a rising sector with the secondary sector contributed around 16 percent of total GDP. The tertiary sector represents the largest share of GDP (around 55 percent). Madagascar is in one of the more cyclone active regions as shown in historical world cyclone tracks from 1849 to 2010 (Fig. 3). The east coast especially is prone being located on the path of destructive cyclones coming from the Indian Ocean. In the southern hemisphere, cyclones occur in three principal regions: i) the Indian Ocean near Madagascar, where over 10% of the global total cyclones occur, ii) the oceanic area to the northeeast and northewest of
47
Australia, and iii) Gulf of Carpentaria. Tropical cyclones are generally more frequent in the northern hemisphere (75 percent of the global total) than in the southern hemisphere. Climate change is projected to increase the intensity of cyclones in the future in Madagascar while decreasing its frequency (see GFDRR, 2011 for a summary). Drought is another important issue, especially in the southern regions, where climate change may worsen its impact (CHRR, 2010). By using a newly developed method in Hochrainer-Stigler and Pflug (2012), it is possible to include such multi risks in the new version of CatSim and can be incorporated in the future if necessary risk assessment steps are taken. In the Madagascar workshop however, these additional risks and possible increases in hazard frequency and intensity were not examined in detail due to resource and data constraints. 3.1. CATSIM model implementation The model was introduced in Madagascar during a 4 day workshop in May 2012 to incorporate disaster risk management into fiscal and development planning processes. The model explored the feasibility of disaster risk management options, supporting the ongoing CPGU (Cellule de prevention etgestion des urgencies, a technical public organization in Madagascar which manages projects about disaster risk reduction) led-study on “Mainstreaming Disaster Risk Management and Climate Change in Economic Development”. The World Bank Madagascar Country Office through financing from the Global Facility for Disaster Reduction and Recovery (GFDRR) supported the project. The project plan suggested that the model to be used and implemented on an annual basis in the future for the budget planning process. Also, further updates with new risk information have been conducted in January 2015 as part of the joint United Nations Office for Disaster Risk Reduction (UNISDR)/ISLANDS Programme for Financial Protection against Climatic and Natural Disasters. More than 30 key stakeholders from finance, education, infrastructure and other ministry across different levels of government participated in both 2012 and 2015 policy stakeholder sessions (Hochrainer-Stigler, 2014; UNISDR, 2015). In the 2012 session, an analysis of past losses and initial direct
Fig. 3. Cyclone tracks from 1849 to 2010 worldwide. Source: NOAA, 2012.
48
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
and indirect risk assessment were performed prior to the workshop, so as to enable the model to be used by the stakeholders during the workshop. Additionally, different economic modeling approaches (inputeoutput, general equilibrium, econometric) were considered and selected for the workshop. A full report on assumptions and the modeling approach were sent to the participants one month prior to the workshop to familiarize them with the approach and to come up with any questions. The idea behind these arrangements is that instead of modeling exercise being a ‘black-box’, the users have the possibility to closely review the advantages and limitations of the approach, estimation procedures and assumptions. This eventually should lead to increase trust in the model and the associated process. Furthermore, the stakeholders are given more in-depth information of the current situation along with an opportunity to see what kind of influence the assumptions had on overall outcomes as initial estimates changed in alternative modeling runs. The interactive process allowed different stakeholders to present numbers and choices to others and to show the kind of decisions they would make based on their perception and the modeling results. In combination with groupbased sessions and discussions, such an approach improved the potential for reaching consensus and to identify most important next steps. In 2015 when updated risk information became available, national-level policy-makers on their volition requested their CatSim assessment be updated as part of the IOC/UNISDR public financing project. This speaks of their general trust and familiarity with the CatSim modeling approach. While the IOC/UNISDR project is ongoing, the following section describes in detail CatSim risk assessment conducted in 2012. 3.2. Direct risk assessment 3.2.1. Total direct risk of natural disasters Direct risk is a combination of hazard, exposure and (physical) vulnerability of exposed assets (Fig. 2). Ideally, a catastrophe modeling approach (see Grossi and Kunreuther, 2005) is used to assess the risk in detail. For example, the hazard information should incorporate at least three variables regarding the source parameters of the hazard: the location of future events, their frequency of occurrence and their severity. The exposure assessment should capture the spatial distribution of the assets at risk, including infrastructure assets such as schools, hospitals or roads as well as private sector assets such as factories as well as houses. Physical vulnerability relates the physical impact of the hazard on the exposed elements. For example, it gauges the relationship between the intensity of the hazard and the percentage of house damage, e.g. damage ratio. Because the intensity and the level of damage are uncertain, the damage itself is an uncertain quantity as well. Underlying each damage function is a frequency component and a severity component. The former determines the probability that an exposed element is damaged and the later determines the percentage of property damaged, assuming damage has occurred. Unfortunately, it is often the case, even in developed countries, that not enough information is available to perform such an approach and alternatives have to be sought (Kull et al., 2013, Michel-Kerjan et al., 2013). While for Madagascar already some good information about the hazard was available (GFDRR, 2011), virtual no information on the asset and (physical) vulnerability side existed at the time of 2012 workshop.2 In the workshop, therefore, it was explained why such information is important and the need for
2 As part of the 2015 CatSim update, fiscal risk assessment based on the countrylevel exposure, hazard and vulnerability using a new global dataset is being conducted as of this writing.
these types of information was recognized by the different ministries. However, since initial estimates were required, alternative approaches had to be performed to obtain preliminary estimates for the workshop. Hence, the EM-DAT database (CRED, 2013) was examined and a minimum distance technique (see HochrainerStigler et al., 2014) was applied to assess the fat tails of the loss distribution (i.e. extremes). The technique and data is used as an approximation of direct risk. It should be noted that loss distributions from other sources, if available, can be easily implemented in the CatSim model via the user interface. The minimum input requirements here include public and private sector losses for selected return periods (see Hochrainer-Stigler, 2014). Fig. 4 shows the results of the loss distribution estimation, including the empirical distribution shown in blue, fitted distribution as calculated using a Generalized Pareto and an Extreme value distribution, and estimated parameters. Note, K is a shape parameter, which takes a value above 0 if the distribution is found to have a heavy tail. Sigma and mu are location parameters needed to estimate return periods (see Embrechts et al., 1997 for a discussion). The return period of the largest event recorded in the past, together with the 100 year event loss are shown too (and used for calibration purposes). Additionally, more detailed information was available for some past events, especially after the 2008 cyclone season, when the World Bank led a cyclone damage assessment using the ECLAC methodology (World Bank, 2008). The losses from those detailed damage assessments were combined via frequency and magnitude analysis of past cyclone paths in Madagascar to calibrate a new loss curve. Further, re-assessment of losses due to cyclone events was performed during the mission in Madagascar, with information obtained from the finance ministry and consultants regarding sectorial losses of severe cyclones over the last 20 years. Given different datasets with varying (or unknown) levels of quality, a minimum, baseline as well as maximum loss distribution were estimated for performing sensitivity tests and represent the direct. The estimation of total capital stock exposed was performed based on the method from Sanderson and Striessnig (2009). In more detail, the Penn World tables from 2012 were used to estimate the average of the first five years of the country's investment series to back-project investment until 1900. To aggregate regional physical capital stocks for the entire period, the perpetual inventory method was applied with an assumedannual depreciation rate of 4 percent (which was chosen based on stakeholder feedback). The estimated capital stock was compared to GDP from the World Development Indicators and averages of capital stock to GDP ratios over the time period were calculated. The capital stock to GDP was found to range between 8 and 10, with an average of 9.5 in the last 20 years. We therefore estimated total capital stock to be around 54.8 billion USD (in constant 2000 prices). To select which part of total capital belongs to the public sector we used first pre-estimates from other studies (Freeman et al., 2002a,b; Hochrainer, 2006) and assumed that approximately 30% of the total stock is public (this is in line with global averages and also the detailed cyclone event loss assessment in 2008) (World Bank, 2008). Since one third of the population of Madagascar is very poor, the government is assumed to absorb a large extra burden in the case of a cyclone. Consistent with global average figures (Hochrainer and Mechler 2009) it is further assumed that the government will have to spend an amount equivalent to 20% of the total stock losses to provide relief. For an estimated total capital stock of 54.8 billion USD, the maximum contingent liabilities for the government of Madagascar (this is the theoretical maximum assuming all capital stock is destroyed) are therefore around 27.3 billion USD if both public assets and private sector relief are assumed. If the government decides to exclude any implicit liabilities then their liability would go down to 16.4 billion
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
49
Fig. 4. Estimated loss distribution (left hand GEV, right hand Pareto type) based on calibrated EM-DAT (2012) data using the minimum distance technique based on HochrainerStigler and Pflug, 2012. Source: Own calculations.
USD (i.e. explicit liability). Different assumptions about what kind of liabilities the government is responsible for can be tested within CatSim iteratively and used for discussion. Based on the information above, the probabilistic losses due to cyclones in terms of percent of capital stock loss are estimated for Madagascar. For example, total potential losses for the government, with an annual exceedance probability of five percent (a 20 year event), is assessed at approximately 114.3 million USD. The 50-, 100- and 500-year events would cause losses of about 409, 1047 and 8954 million USD, respectively. An important summary measure is the annual expected losses, or the losses to be expected on average every year. It is the sum of all losses weighted by the probability of occurrence. Graphically, the expected losses are represented by the area above the loss distribution (Fig. 4). For the public sector, modeled annual losses are on average around 10 million USD with a standard deviation of 39 million USD. For the baseline model, annual average losses are around 55 million USD with a standard deviation of 253 million USD. The minimum baseline case gives the annual expected losses of around 26 million USD with a standard deviation of 98 million USD. Integrating the curve for the maximum case, losses are very high, with average annual losses estimated to be around 213 million USD and a standard deviation of about 787 million USD. Table 2 summarizes the results. It should be noted that disasters are not average, but extreme, events occurring very rarely. Over a longer time period, like 100 or 500 years, catastrophe losses suffered will be close to the sum of annual expected losses over these years. Large shocks could however occur anytime and therefore necessitate the modeling of
Table 2 Annual expected losses and standard deviations for selected models. Model
Annual expected loss (million USD)
Standard deviation
Public sector only Baseline case Minimum case Maximum case
10 55 26 213
39 253 98 787
Note: Public sector only includes the government's explicit liability. The baseline, minimum and maximum cases are representing the uncertainty within the estimation procedure to produce loss distributions (based on the backward-looking approach explained above). Source: Own calculations.
potential consequences of these shocks. 3.2.2. Fiscal resources availability to cope with disasters As indicated before, ad-hoc as well as proactive financial sources to cope with the event may differ widely between countries and dependent partly on the general socio-economic situation as well as market access (e.g. to insurance) of the respective country (Linnerooth-Bayer et al., 2005; Lal et al., 2012). Financial resources including costs and constraints must be sufficiently understood for planning an effective disaster risk management strategy, and Madagascar is no exception. As Madagascar is constrained by its fiscal inflexibility and low revenue base it is assumed that domestic credit is available up to approximately USD 50 million (this is a very optimistic assumption, see also World Bank, 2011). Diversion from the budget is not feasible to a large extent and the study assumed that 1 percent of the budget can be diverted based on the stakeholder input during the workshop. The study further assumed that 10 percent of the total losses will be financed by outside assistance (a rather optimistic assumption, averages from past events are showing assistance around 3e5 percent of total losses, see Becerra et al., 2012). Borrowing from multilateral and international sources is assumed to be possible due to low debt (in the worst case scenario, there were some Chinese Banks who actually would gave a credit for Madagascar). We estimated that post-disaster loans were possible up to approximately 40 million USD with an equal split from multilateral and international sources, at different interest rates and conditions. Central Bank credits (e.g. printing money) were assumed not feasible due to risk of high inflation. These are generally very optimistic assumptions and therefore the results should be treated with caution. The initial parameters merely served as the baseline for discussion and these parameters were subsequently re-assessed in the workshop again. During the workshop, a fruitful discussion took place regarding different options, needs as well as consequences of the assumptions and changes were assessed in an iterative manner. 3.2.3. Fiscal resources gap analysis In a next step, the government's economic vulnerability in terms of potential resource gap due to cyclone exposure was estimated. This step is the most crucial of CatSim modeling as complex aspects such as risk and resources are combined. This is done by examining the direct risk and potential resources for given events. For the
50
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
technical details of this step, we refer to the Hochrainer-Stigler (2014), and next discuss the main results and implications for the baseline case. Fig. 5 proved to be successful in getting this idea of risk and resources across (others are shown in Hochrainer-Stigler, 2014). The large graph on the middle right hand side shows the return periods for each loss event on the x-axis and the corresponding losses in monetary terms. The upper line gives the total losses while the colors show the different financing instruments that could be used in case of a disaster. Keeping in mind the data limitations and restrictive assumptions, the analysis shows that the government of Madagascar has sufficient financing up to the 23 year cyclonerelated loss event. However, for less frequent and more severe events a gap would occur. For example, the gap would amount to 0.95 billion USD for a 100 year event in the baseline case. In Table 3, the financing gap year event as well as the corresponding gap for a 100 year event loss is shown for some of the estimates used during the workshop. The financing gap year event is similar for the “public sector only” and the “baseline” model runs. However, as the former does not exhibit a fat tail, the losses for more extreme events are significantly lower as can be seen by the lowest figure for the 100 year event gap (Table 3). Even with very optimistic loss financing assumptions, one can expect higher losses for more extreme events. On the other hand, the maximum bounds model gauges losses to be very high. It also should be noted that while for the
Table 3 Financing gap return period and 100 year gap (billion USD). Model
Financing gap year event 100 year gap
Maximum
Baseline
Minimum
Public
1 5.71
23 0.95
4 0.54
24 0.09
Source: Own calculations.
minimum model the financing gap year event is lower than for the baseline case (due to higher loss estimates for more frequent events) the losses for extremes is larger for the baseline case. There is only limited information available to validate the results but the PDNA (Post disaster needs assessment) done for the 2008 cyclone season in Madagascar (GFDRR, 2008) estimated direct losses to be around 180 million USD (approximately a 15e20 year event according to the baseline case) and indicated as in this analysis severe problems to finance the losses with the limited resources at hand. The consequences of these alternative assumptions were discussed during the workshop, including the possibilities to limit the responsibility of the government to finance its own losses as well as to increase loan packages for large scale events. After discussions, the stakeholders decided to use the baseline case scenario to assess the indirect and possible long-terms risks in the next step (dynamic assessment shown in Fig. 2).
Fig. 5. Financial vulnerability and resource gap for Madagascar using the baseline direct risk estimates. Note: The middle figure shows return periods on the x-axes and resources needed/supplied in the y-axes. The bottom figure is an interactive graph with which users can explore the details of alternative fiscal resource allocation at different return periods.
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
3.3. Dynamic economic risk assessment Information on the resource gap is helpful for determining the current level of risk. However, more important ultimately are the implications of this gap on economic development and other “flow variables,” which allows for mainstreaming disaster risks into development planning and fiscal management. For this, direct risk, financial vulnerability and the prevalent economic conditions in Madagascar are combined to derive an estimate of macroeconomic impacts. Economic modeling approaches such as inputeoutput modeling and general equilibrium modeling are available (for cons and pros of these approaches see Rose, 2004); but most of them could not be used due to data scarcity. Though some inputeoutput tables were available, they were outdated and deemed unusable due to Madagascar's economic situation that has drastically changed during the last years. Hence, it was decided to estimate an aggregate CobbeDouglas function, using past capital stock, GDP rates as well as labor as inputs. The parameters can be changed in the user interface and default growth estimates without a disaster event can be calibrated based on available models such as one used by the World Bank (Mechler, 2004). In this way, added flexibility can account for fast changes in the political and economic structure. Fig. 6 is the second key interface for dynamic risk assessment, which is showing long term consequences of potential disasters on future growth, i.e. a selection of trajectories for Madagascar's future GDP (in constant 2000 billion USD). The economy grows over time
51
as investment adds to the capital stock, but in a number of cases disasters causes a loss of assets and income. Depending on the financial resilience of the government, these events put the economy on a lower growth trajectory. In some cases, there is a dramatic decrease in economic activity. We have found, based on discussion with the stakeholders, that the graphical representation of path dependency and risk was the most informative way to convey the idea of disaster risk, including the importance of probabilities and considering path dependencies. Note, the occurrence of natural disasters, and growth trajectories thereafter, are stochastic and depend on the probability distribution of financial losses. Normally, 5000 or 10,000 trajectories are calculated for policy assessment, but the Figure only summarizes 500 for illustration purposes. These trajectories do not have equal probability: the cases with economic growth proceeding as planned (the trajectories in the upper part) have a higher probability of occurrence than the catastrophic cases appearing at the bottom. Such an assessment illustrates the worst outcomes compared to the planned business-as-usual cases of economic development. Table 4 presents the resource gap probability for the next 5 years and the credit buffer drop for the baseline, maximum and minimum models. The credit buffer drop is defined as the average drop in external savings (credit) available over the next 5 years from the starting level (defined here to be 40 million USD). A decrease of the credit buffer indicates an expected increase in indebtedness and no
Fig. 6. Potential GDP impacts due to disaster events. Note: each line represents the result of stochastic GDP trajectories estimated using a Monte-Carlo simulation.
52
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
Table 4 Financing gap probability and credit buffer drop (million USD). Model
Probability of financing gap (%) Credit buffer drop (million USD)
Maximum
Baseline
Minimum
93 39
48 32
89 36
Source: Own calculations.
possibilities to borrow in the future. The public sector only model is excluded from the analysis, since private sector losses must be incorporated for macroeconomic projections. For the baseline model, the probability of encountering a financing gap over the next 5 years is 48 percent. For the remaining models, it is between 89 and 93 percent, i.e. financing problems will emerge in the future with near certainty. Additionally, indebtedness will likely increase as well. While it is estimated that around 40 million USD of credit could currently be obtained from international markets, this would decrease to less than 8 million USD (in the baseline model) or near zero (in the maximum and minimum models). Hence, even in the case that losses can be financed sporadically in the short term via credits, the availability of this kind of option is decreasing over the long term, making it difficult to keep creditworthiness in the future. This leads to the question as to what kind of instruments or options could be chosen to decrease this type of risk.
The final step of the CatSim model evaluates possible risk reduction and risk transfer measures. Investing in the risk financing instruments can be viewed as a trade-off between economic growth and stability: while economic growth is higher on average if the government does not allocate its resources to risk instruments, the economy has smaller risks of sever extremes under proactive public sector financing instruments and is therefore more stable. Budgetary resources allocated to catastrophe reserve funds, insurance and contingent credit (as well as to preventive loss-reduction measures) reduce the potential resource gap, and thus can ensure a more stable development path. At the same time, ex ante financing and prevention measures come at a price in terms of other investments foregone and will inevitably cause an adverse impact on the growth path of an economy. The CatSim model assesses this trade-off by comparing the costs of selected ex-ante measures with their benefits in terms of decreased possibility of a resource gap. Fig. 7 shows the decrease in the probability of a financing gap for the next 5 years under different options. The expenditure as percent of available budget is shown from 0 to 80 percent (x-axis); however, one should recognize that only 1 to 2 percent could be actually used in the most optimistic settings. For example, using 2 percent of the discretionary budget over the next 5 years, such risk management option would decrease the probability of a resource gap from 43 to around 22 and 31 percent. The credit buffer drop would not change significantly, indicating
Fig. 7. Decrease of the financing gap probability using different instruments. Note: Mitigation refers to disaster risk reduction investment. Insurance refers to catastrophe insurance. Reserve fund refers to ex-ante budget allocation in the form of an accumulation fund. Contingent credit refers to ex-ante arrangement for preferential loans.
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
that most of the losses have to be financed through credits. This is further reflected in the yellow line, which represents the contingent credit option. It is beneficial to obtain easy and inexpensive credits after a disaster event; however, one still has to repay the loan and therefore become more financially vulnerable during the payment period as compared to other financing options. Furthermore, investing 2 percent of the discretionary budget into risk management options would decrease GDP by half a billion over the next 5 years as government budget is diverted to more economically productive investment options. Which options finally fit better is a question which cannot be addressed via looking at the model results but has to incorporate other probably nonquantifiable dimensions (such as education and health issues) too, which where embedded in the presentation of the sub-groups during the workshop. 4. Discussion and outcomes The policy workshop using the latest version of the CatSim model provided a forum to discuss disaster risk mainstreaming into sustainable development within a broader context. In addition to technical assessment questions, the Madagascar stakeholders expressed that additional information, institutions and capacities are needed in the future for effectively mainstreaming risk. The specific necessary next steps need to be taken were found to be similar compared to previous workshops done in other countries and are summarized next. Regarding the direct risk assessment part, it was agreed by all participants that more detailed exposure and (physical) vulnerability data is necessary and has to be gathered in the future. Frequent updates and sharing of information among different government ministries are found to be key features for managing successfully future risk. Especially the spatial distribution of critical infrastructures and other system relevant elements at risk as well as corresponding vulnerabilities should be targeted first. In that way, hot spot areas could be identified which could serve as test cases for implementing disaster risk management strategies. In the long run an inventory of different kind of (public and private sector) assets as well as corresponding vulnerabilities should be built which could serve as input parameters for more advanced direct risk estimation approaches which could then used as input parameters for CatSim. Additionally, we discussed that it will be important, for example for model calibration purposes as well as direct risk assessments, to measure future cyclone events and its effects on society with as much details as possible (as the study of the 2008 Cyclone season, World Bank, 2008). Such analysis could also give indications what kind of emergency steps should be taken first and what kind of recovery processes are most needed in the medium and long run. Furthermore, participants realized that financial vulnerability of the public sector presents only one, albeit an import one, aspect of vulnerability to natural hazards. Evaluation of other quantitative and qualitative indicators for sustainable development under catastrophe risks is necessary to complement this concept. Disaster risk management has to be understood within a general integrated approach to enhance socio-economic development. It is especially important to include qualitative data and objectives such as education (see Lutz et al., 2008) and quality of infrastructure (World Bank, 1994). The collection and evaluation of such information necessitates the improved interaction of different ministries using both quantitative and qualitative approaches to reach consensus. Clearly, cyclone risk is only one part of the puzzle and needs to be integrated within a systems based perspective. Furthermore, as disasters risks and financial vulnerability estimates involve substantial degree of uncertainty, it is important that wider groups of
53
stakeholders/potential users have full participation in the design, estimation and use of instruments in the future. The stand-alone application of CatSim with the accompanied manual should help in that regard to keep the dialog open. During the workshop, stakeholders agreed that risk reduction is more appropriate for frequent events and risk financing (such as insurance or a reserve fund) seems more effective for less frequent events. As the resource gap is estimated to occur already for very frequent events, it therefore seemed better to focus on risk reduction options first (such as the GFDRR supported development of cyclone-resistant national building codes). As for concrete policy options for disaster risk, the Mexico disaster fund for the public sector as well as the Caribbean insurance facility were seen as potential ways forward. In Mexico, following the devastating earthquake of 1985, the national government created a budgetary program called FONDEN (Fund for Natural Disasters) to enhance the country's financial preparedness for natural disaster losses. FONDEN provides last-resort funding for uninsurable losses, such as emergency response and disaster relief. The Fund has helped over the years to better understand the risks as well as possible ways to reduce them (for example via the reinsurance of the Fund, see Cardenas et al., 2007). In the Caribbean case the Catastrophe Risk Insurance Facility (CCRIF) has been operating since June 2007 with 16 Caribbean countries. The fund covers up to 20% of the estimated infrastructure loss, and claims are paid depending on an index for hurricanes (wind speed) and earthquakes (ground shaking). The pool gives a substantial reduction in premium cost of about 45e50% for the participating countries. Hence, there were clear advantages of pooling their risks, and thus diversifying across island states (World Bank, 2007). Such disaster fund and/or risk pooling options together with other African states could be also a possibility for Madagascar. Indeed, due to the tight financial situation in Madagascar it is likely the case that opportunity costs will be high and therefore such options have to be assessed in the general context of economic development strategies for future economic growth and CatSim could help shape such discussions (as in the Mexican case, see Cardenas et al., 2007, and the CCRIF, see Hochrainer and Mechler, 2009) illustrating advantages as well as disadvantages of alternative risk instruments given the current and future financial resilience. However, there are also clear limitations in using our approach which have to be kept in mind. First of all, it does not replace an extensive catastrophe modeling analysis which is needed for setting up real-world insurance instruments. While the model can give indications of the insurance costs, the uncertainties around such premium calculations are too high to be used right away for market based applications. Hence, for specific insurance arrangements more/extensive modeling would be necessary. Furthermore, the economic model used for calculating indirect effects of disasters need to be coupled with the models used in the respective (finance) ministries for projecting future economic performance. This is not an easy task and would require further careful considerations in respect of the specific models used in the country which usually include considerations down to the project level (see Fig. 1). As such, various additional research, modeling and stakeholder workshops are needed for implementation of specific instruments. While these limitations are noteworthy, nevertheless the model can estimate the level of risk the country is exposed to and therefore can provide information which options may be the most promising ones. As the approach is broad enough, risk levels can be compared for groups of countries making it possible to determine at which level of risk each of these countries may think to behave risk averse and implement risk financing options (Mechler and Hochrainer-Stigler, 2014). This can also be beneficial on the global level for a possible Climate Adaptation Fund which could decrease
54
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55
individual country costs considerably (see Hochrainer-Stigler, 2014). While obvious, it is still important to note that data scarcity is one of the most important limitations for disaster risk assessment and the management of it. As such information is usually distributed among different government institutions not aware of the importance of their data for the whole process the workshops can increase this awareness and stimulate data gathering and interactions between them. Last, the idea of an open rather than closed dialogue around how to best decrease future disaster risk increases the possibility of a consensus between different stakeholders and their primary focus, may it be financial, human or social related.
between different stakeholders and only after understanding the need for collaboration based on an integrated perspective and iterative discussions sustainable solutions may be found (IPCC, 2012). Consequently, the CatSim model is essentially only one part within a broader process based approach to tackle the question how disaster risk can be planned for. Concurrent with recent global risk modeling efforts (see GAR 2015) more and more risk assessment information is now available on the country scale, which can be easily implemented in CatSim and further support the decision making process.
5. Conclusions
Becerra, O., Noy, I., Cavallo, E., 2012. A., Foreign Aid in the Aftermath of Large Natural Disasters.IDB. Working Paper No.IDB-WP-333. Bettencourt, S., Croad, R., Freeman, P., Hay, J., Jones, R., King, P., Lal, P., Mearns, A., Miller, G., Pswarayi- Riddihough, I., Simpson, A., Teuatabo, N., Trotz, U., Van Aalst, M., 2006. Not if but When e Adapting to Natural Hazards in the Pacific Islands Region: a Policy Note. East Asia and Pacific Region. Pacific Islands Country Management Unit; The World Bank, Washington D.C. Cardenas, V., Hochrainer, S., Mechler, R., Pflug, G., Linnerooth-Bayer, J., 2007. Sovereign financial disaster risk management: the case of Mexico. Environ. Hazards 7 (1), 40e53. Cardona, O.D., et al., 2012. CAPRA e comprehensive approach to probabilistic risk assessment: international initiative for risk management effectiveness. In: A: World Conference on Earthquake Engineering. “15th World Conference on Earthquake Engineering”. Lisboa, pp. 1e10. CHRR, 2010, Columbia University Center for Hazards and Risk Research (CHRR) and Columbia University Center for International Earth Science Information Network (CIESIN). CRED, 2013. EM-DAT: International Disaster Database, Centre for Research on the Catholique de Louvain, Belgium. Epidemiology of Disasters. Universite Delgado-Galvan, X., Izquierdo, H., Benitez, J., Perez-Garcia, R., 2014. Joint stakelholder decision making on the management of the Silao-Romita aquifer using AHP. Environ. Model. Softw. 51, 310e322. Embrechts, P., Klüppelberg, C., Mikosch, T., 1997. Modellingextremal Events for Insurance and Finance. Springer, Berlin. CIA Factbook, 2012. The World Factbook. https://www.cia.gov/library/publications/ the-world-factbook/geos/ma.html. Freeman, P.K., Martin, L.A., Linnerooth-Bayer, J., Mechler, R., Saldana, S., Warner, K., Pflug, G., 2002a. Financing Reconstruction. Phase II Background Study for the Inter-American Development Bank Regional Policy Dialogue on National Systems for Comprehensive Disaster Management. Inter-American Development Bank, Washington DC. Freeman, P.K., Martin, L., Mechler, R., Warner, K., Hausman, P., 2002b. Catastrophes and Development, Integrating Natural Catastrophes into Development Planning. Disaster Risk Management Working Paper Series No.4. Worldbank, Washington DC. GFDRR, 2008. Damage, Loss, and Needs Assessment for Disaster Recovery and Reconstruction after the 2008 Cyclones Season in Madagascar. Government of Madagascar, UN and World Bank. Available at: https://www.gfdrr.org/ madagascarpdna2008. GFDRR, 2011. Vulnerability, Risk Reduction and Adaptation to Climate Change. Madagascar Climate Risk and Adaptation Country Profile. World Bank, Washington, D.C. Global Assessment Report, 2013. GAR e Global Assessment Report. UNISDR. Available at: http://www.unisdr.org/we/inform/gar. Global Assessment Report, 2015. GAR e Global Assessment Report. UNISDR. Available at: http://www.preventionweb.net/english/hyogo/gar/2015/en/home/ index.html. Grossi, P., Kunreuther, H. (Eds.), 2005. Catastrophe Modeling: a New Approach to Managing Risk. Springer, New York. Hallegatte, S., 2014. Modeling the role of inventories and heterogeneity in the assessment of the economic costs of natural disasters. Risk Anal. 34, 152e167. http://dx.doi.org/10.1111/risa.12090. Hewitt, R., van Delden, H., Escobar, F., 2014. Participatory land use modelling, pathways to an integrated approach. Environ. Model. Softw. 52, 149e165. Hochrainer, S., 2006. Macreoeconomic Risk Management against Natural Disasters. DeutscherUniversitaetsverlag (DUV), Wiesbaden. Hochrainer, S., Mechler, R., 2009. Assessing financial and economic vulnerability to natural hazards: bridging the gap between scientific assessment and the implementation of disaster risk management with the CatSim model. In: € ter, D., Klein, R., de la Vega-Leinert, A. (Eds.), Assessing Patt, A., Schro Vulnerability to Global Environmental Change.London, Earthscan, pp. 173e194. Hochrainer-Stigler, S., 2014. User Interface of the CatSim Model and Practical Guidelines. IIASA, Laxenburg, Austria. Available at: http://www.iiasa.ac.at/ publication/more_XO-14-004.php. Hochrainer-Stigler, S., Pflug, G., 2012. Risk management against extremes in a changing environment: a risk-layer approach using copulas. Environmetrics 23 (8), 663e673.
Governments can face post-event deficits in financing response and relief, and reconstruction. Such deficits can have adverse ripple effects on a country's long-term development or fiscal stability as well as the ability to finance social and economic programs. As illustrated in this article, investing in ex-ante disaster risk financing and mitigation options can be seen as a trade-off between growth and stability. Budgetary resources allocated to loss mitigation, catastrophe reserve funds, insurance and contingent credit improve the stability of economic performance, but come at a price in terms of other investments foregone. The presented CatSim model assesses this trade-off by estimating current risk levels and comparing the costs of selected ex-ante measures with their benefits in terms of the decrease of the probability of resource gaps and the improvement in the ability to access loans. CatSim is equipped with a flexible stand-alone software package and tested user interfaces to increase the potential of consensus for different stakeholders during the process of mainstreaming disaster risk into development planning policy processes. In this article the latest version of CatSim was presented and applied as part of an in-depth case study example for Madagascar, which has one of the highest cyclone risks worldwide. During a workshop in 2012 finance ministry and disaster management officials used CatSim to identify risk and financial vulnerability in terms of resource gaps or post-disaster liquidity shortfalls for financing reconstruction and relief. The model informed participants regarding the pros and cons, costs and benefits of risk financing instruments reserve fund, (re-) insurance or contingent credit. Generally, there was great interest in the exercise and the CatSim model was seen as a successful tool to bridge the gap between science, policy-making and implementation, especially owing to the user interfaces created in the last years. While the stakeholders were eager to know the optimal mix of strategies at the beginning of the workshop; at the end of the workshop, they understood, after some heavy discussions that they would need more information and importantly collaboration (see also Hewitt, van Delden and Escobar, 2014). Based on this successful experience, the CATSIM model is now being updated with new risk information as part of the ongoing joint UNISDR/ISLANDS Programme. Through these policy-stakeholders sessions, participants also learnt that better link between different institutions is needed and other non-quantifiable dimensions must be incorporated too. Participants also had learnt that not all risk can be reduced. Those risks that be reduced come at cost, and not everything can be solved immediately, but improved evaluation and management options can be incorporated in a step by step manner as more information becomes available. The importance of establishing an open dialogue with as much participation as possible should not be underestimated as perception biases (Thompson, 1997; Verweij and Thompson, 2011; Delgado-Galvan et al., 2014) can be large
References
S. Hochrainer-Stigler et al. / Environmental Modelling & Software 72 (2015) 44e55 Hochrainer-Stigler, S., Timonina, A., Williges, K., Pflug, G., Mechler, R., 2013. Modelling the indirect and fiscal risks from natural disasters using the CatSimModel. In: Background Paper Prepared for the 2013 Global Assessment Report on Disaster Risk Reducftion. UNISDR, Geneva, Switzerland. http://www. preventionweb.net/gar. Hochrainer-Stigler, S., Mechler, R., Pflug, G., Williges, K., 2014. Funding public adaptation to climate-related disasters. Estimates for a global fund. Glob. Environ. Change 25, 87e96. IPCC, 2012. Summary for policymakers. In: Field, C.B., et al. (Eds.), Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, UK, and New York, NY, USA, pp. 1e19. Kull, D., Mechler, R., Hochrainer-Stigler, S., 2013. Probabilistic cost-benefit analysis of disaster risk management in the development context. Disasters 37 (3), 374e400. http://dx.doi.org/10.1111/disa.12002. Lal, P.N., et al., 2012. National systems for managing the risks form climate extremes and disasters. In: Field, C.B., et al. (Eds.), Managing the Risks of Extreme Events and Disasters to Advance Climate Change. Cambridge University Press, Cambridge, UK, and New York, NY, USA, pp. 339e392. Lavell, A., Oppenheimer, M., Diop, C., Hess, J., Lempert, R., Li, J., Muir-Wood, R., Myeong, S., 2012. Climate change: new dimensions in disaster risk, exposure, vulnerability, and resilience. Chapter 1. In: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L., Mastrandrea, M.D., Mach, K.J., Plattner, G.-K., Allen, S.K., Tignor, M., Midgley, P.M. (Eds.), Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation, A Special Report of Working Groups I and II of the Intergovernmental Panel on Climate Change (IPCC). Cambridge University Press, Cambridge, UK, and New York, NY, USA, pp. 25e64. Linnerooth-Bayer, J., Mechler, R., Pflug, G., 2005. Refocusing disaster aid. Science 309, 1044e1046. Lutz, W., Crespo Cuaresma, J., Sanderson, W.C., 2008. The demography of educational attainment and economic growth. Science 319 (5866), 1047e1048. Mechler, R., 2004. Natural Disaster Risk Management and Financing Disaster Losses in Developing Countries. VerlagfuerVersicherungswissenschaft, Karlsruhe. Mechler, R., Hochrainer-Stigler, S., 2014. Revisiting arrow-lind: managing sovereign disaster risk. J. Nat. Resour. Policy Res. 6 (1), 93e100. Mechler, R., Linnerooth-Bayer, J., Hochrainer, S., Pflug, G., 2006. Assessing financial vulnerability and coping capacity: the IIASA CatSim model. In: Birkmann, J. (Ed.), Measuring Vulnerability and Coping Capacity to Hazards of Natural Origin. Concepts and Methods. United Nations University Press. Mechler, R., Hochrainer, S., Pflug, G., Lotsch, A., Williges, K., 2009. Assessing the Financial Vulnerability to Climate-related Natural Hazards. Policy Research Working Paper. 5232. World Bank, Washington, DC. Michel-Kerjan, E., Hochrainer-Stigler, S., Kunreuther, H., Linnerooth-Bayer, J., Mechler, R., Muir-Wood, R., Ranger, N., Vaziri, P., Young, M., 2013. Catastrophe risk models for evaluating disaster risk reduction investments in developing countries. Risk Anal. 33 (6), 984e999.
55
NOAA (2012) National Oceanic and Atmospheric Administration. Available at: http://www.nhc.noaa.gov/data/. € misch, W., 2007. Modeling, Measuring and Managing Risk. World SciPflug, G., Ro entific, Singapore. Rose, A., 2004. Economic principles, issues, and research priorities in hazard loss estimation. In: Okuyama, Y., Chang, S.E. (Eds.), Modeling Spatial and Economic Impacts of Disasters. Springer, New York. Sanderson, W.C., Striessnig, E., 2009. Demography, Education, and the Future of Total Factor Productivity Growth. IIASA Interim Report IR-09e002 (Laxenburg, Austria). Swiss Re, 2012. Natural Catastrophes and Man-made Disasters in 2011. Sigma, Zurich. Nr. 2/2012. Thomas, Anne-Claire, 2011. Poverty, Risk and Insurance in Rural Madagascar. PhD catholique de Louvain a Louvain-la-Neuve (Belgique). Thesis. l'Universite Thompson, M., 1997. Cultural theory and integrated assessment. Environ. Model. Assess. 2, 139e150. UNISDR (the United Nations Office for Disaster Risk Reduction), 2015. Review of Madagascar. UNISDR Working Papers on Public Investment Planning and Financing Strategy for Disaster Risk Reduction. Verweij, M., Thompson, M. (Eds.), 2011. Clumsy Solutions for a Complex World. Palgrave Macmillan, Basingstoke, Hampshire, UK. World Bank, 1994. Infrastructure for Development. World Bank and Oxford University Press, New York. World Bank, 2007. The Caribbean Catastrophe Risk Insurance Initiative; Results of Preparation Work on the Design of a Caribbean Catastrophe Risk Insurance Facility. World Bank, Washington, DC. World Bank, Government of Madagascar and UN, 2008. Damage, Loss, and Needs Assessment for Disaster Recovery and Reconstruction after the 2008 Cyclone Season in Madagascar. Cyclone Fame, Ivan and Jokwe in Madagascar. World Bank. http://www.3adi.org/tl_files/3ADIDocuments/Country%20information/ Madagascar/Madagascar_gov_2008_recovery_plan.pdf. World Bank, 2010a. A Brief Publication of the World Bank's Africa Region Sustainable Development Department: “Madagascar Disaster Risk Reduction Plan: Moving from Disaster Response to Prevention”. World Bank, 2010b. Natural Hazards, Unnatural Disasters. World Bank, Washington, DC. World Bank, 2011. Madagascar Economic Update: Fiscal Policy e Managing the Present with a Look at the Future. World Bank. February 7, 2011.
Further reading Mechler, R., Hochrainer, S., Linnerooth-Bayer, J., Pflug, G., 2013. Public sector financial vulnerability to disasters. The IIASA CatSim model. In: Birkmann, J. (Ed.), Measuring Vulnerability to Natural Hazards. Towards Disaster Resilient Societies. Revised and Extended Second Edition. United Nations University Press, Tokyo, pp. 380e398.