Chapter 9 Building Resilience to Natural Disaster in Vulnerable States: Savings from Ex Ante Interventions

Gerd Schwartz, Manal Fouad, Torben Hansen, and Genevieve Verdier
Published Date:
September 2020
  • ShareShare
Show Summary Details
Wei Guo and Saad Quayyum


The frequency of natural disasters is expected to rise with climate change, along with the damage they do. This leaves many countries, especially small states, highly vulnerable. The number of Category 4 and 5 storms in the North Atlantic is expected to increase by 45 to 87 percent over the course of 21st century (Knutson and others 2013), and weather events such as floods, coastal inundation, drought, and cyclones are expected to intensify in the Pacific (World Bank 2017). These patterns will exacerbate economic challenges for small island nations in the Caribbean and the Pacific, where average annual damage from these disasters as a percentage of GDP are typically four to five times higher than in other countries (IMF 2019). For example, Dominica was devastated by a hurricane in 2017, with damage done equal to more than 200 percent of GDP, only two years after being by hit by a hurricane that cost the country nearly 100 percent of GDP.

These natural disasters not only destroy lives and livelihoods, but also do significant harm to economic growth and national debt. Noy and Nualsri (2007), Noy (2009), Raddatz (2009), Loayza and others (2012), and Bayoumi, Quayyum, and Das (forthcoming) documented the adverse effect of natural disaster on growth. Lee, Zhang, and Nguyen (2018) found that large disasters a have significant negative effect on growth and fiscal and trade balances among small Pacific island nations. Strobl (2012) explored the impact of hurricanes in the Central America and Caribbean regions and found that on average they lead to reduction in growth of 0.83 percent in the year of impact.

These disasters are associated with large recovery costs as significant stocks of public and private infrastructure have to be rebuilt (IMF 2018b). Given the large size of these shocks and limited fiscal space in disaster-vulnerable countries, much of the recovery costs are often financed by official development assistance from the international community (IMF 2019).

A significant amount of the damage and associated output lost could be avoided through investment in building resilient infrastructure in vulnerable areas before the next disaster strikes (in other words, an ex ante intervention). However, financing for this is insufficient. United Nations Environment Programme (2016) reported that adaptation needs are at least two to three times the available international public financing. Donor support for vulnerable countries is heavily skewed toward postdisaster recovery.1 Domestic financing for resilience building is also limited, as many of the vulnerable countries have high public debt or high-priority development needs.

This chapter explores whether building resilience is cost effective. In other words, whether the benefits are sufficient to justify the upfront costs. It uses a dynamic stochastic general equilibrium model to explore intertemporal trade-offs and the benefits of building resilience, focusing on six countries especially vulnerable to natural disasters—Antigua and Barbuda, Dominica, Fiji, Haiti, St. Lucia, and St. Vincent and the Grenadines. The following two policy options for policymakers over a 20-year span are explored:

1. Take no resilience actions before a disaster occurs.

2. Spend a constant fraction of GDP in building resilient infrastructure in non-disaster years.

The exercise assumes that countries are hit by disasters of various sizes over 20 years based on the historical frequency of these shocks. It also studies a scenario in which the frequency of disaster increases because of climate change.2

The cost of rebuilding public infrastructure after a disaster is found to be larger in the first scenario than in the second, as the stock of infrastructure is less resilient. Policymakers can save in net present value terms by investing in resilience before a disaster and so avoid large rebuilding costs. According to the model used, the average savings for the six island nations considered (net of additional cost of investing in resilience) in the baseline is 10 percent of initial-year GDP over 20 years and increases up to 14 percent of GDP if the frequency of disasters rises. In addition, countries benefit from lower output losses in the event of a disaster, which averages to about 4 percent of initial-year GDP in net present value terms. The average increases to about 6 percent of initial-year GDP in the scenario in which the frequency of disaster increases.

The findings underscore the importance of mobilizing more resources toward building ex ante resilience. As noted in Chapters 3 and 4, for many vulnerable states, financing for such investments will be limited by available fiscal space. Countries will not only need to mobilize domestic revenue and prioritize spending, but also spend better and increase the efficiency of capital spending. The international community can also play a role. By changing the pattern of support toward building resilience, donors can not only increase welfare in disaster-vulnerable countries but can also expect to save in the long term from lower outlays on recovery efforts in disaster-vulnerable countries.

This work is related to the literature exploring the impact of public investment on growth such as Barro (1990), Barro and Sala-i-Martin (1992), and Futagami, Morita, and Shibata (1993). It is also related to papers on the macroeconomic and fiscal impact of natural disaster. Such papers include Cavallo and Noy (2011), Cavallo and others (2013), and Bevan and Cook (2015). Finally, the analysis here is closely related to Marto, Papageorgiou, and Klyuev (2018), which employed a similar model to study how small developing states could build resilience to and recover from natural disasters while maintaining debt sustainability. In particular, Marto, Papageorgiou, and Klyuev (2018) explored how much grant financing is required to ensure debt sustainability, and how donor support for resilience can improve debt profile. Nevertheless, the following analysis departs from that in some key assumptions. It assumes resilient capital is more expensive than nonresilient capital, which creates an intertemporal trade-off for policymakers in choosing what kind of capital to invest in. It also introduces multiple shocks that are calibrated to a country’s own history in terms of frequency and size—which also play an important role in the intertemporal trade-off. Furthermore, it assumes that countries have borrowing constraints, which is common among disaster-vulnerable small states, many of which have high debt or are at high risk of debt distress (IMF 2019).

Stylized Facts

The number of natural disasters per year has been steadily increasing since the early 1990s (Figure 9.1; Chapter 14). In 2017, the number of disasters reported was more than double that of 1992. With climate change, this trend is likely to continue. These disasters cause significant damages, especially for small states. The average annual effect of natural disasters for small Caribbean and Pacific states ranges between 2.0 percent and 3.0 percent of GDP, which is about four to five times higher than for larger countries (see Figure 9.2, panel 1).

Frequency of Natural Disasters: 1980–2017

Source: Munich Re Group.

Impact of Natural Disasters and Disaster Aid Allocation

Note: EM-DAT = Emergency Events Database; WEO = World Economic Outlook.

Large natural disasters have significant macroeconomic impact—reducing growth and pushing up debt. Disasters that cause damage exceeding 20 percent of GDP are followed by a decline in growth of about 3 percentage points on average (Figure 9.2, panel 2) and increase debt by about 8 percent of GDP (Figure 9.2, panel 3). However, considerable heterogeneity is seen across countries, depending on initial conditions and size of the shock. In Dominica for example, growth is estimated to have collapsed to –9 percent in 2017–18 after Hurricane Maria in 2017.3

Fiji’s Efforts to Build Ex Ante Resilience to Natural Disaster and Climate Change

Fiji is particularly vulnerable to natural disasters. Tropical storms and floods cause about 5 percent of GDP damage annually, but damage can be as high as 20 percent of GDP, as it was for Tropical Strom Winston in 2016. The government has been steadily increasing investment in resilience, from about 4 percent of its annual budget in 2013 to about 10 percent in fiscal year 2016/17.

Recently Fiji conducted a climate vulnerability assessment to identify the investment needed to improve resilience to natural disasters and climate change. It has estimated these investment needs to be about 100 percent of GDP over 2018–2027, of which about half is on top of funds already earmarked in existing plans.

Increased spending to improve infrastructure is central to Fiji’s strategy to mitigate climate change, with transportation infrastructure getting the highest amount. It also envisages large investment in flood-risk management and coastal protection. These investments are expected to not only improve resilience to natural disasters but also improve livelihoods and productivity.

To finance resilience building, the government instituted the new Environmental and Climate Adaptation Levy in 2017, which is expected to yield 1 percent of GDP. Moreover, in November 2017 Fiji issued its first sovereign bond for financing climate and environmental resilience projects, becoming the first developing country to pursue such an initiative. Between 2011 and 2014, Fiji received about $10 million (about 0.25 percent of GDP) in concessional finance per year from multilateral and bilateral donors for climate resilience and disaster-risk management. It will need continued and additional support from donors to achieve its ambitious resilience plans.

Source: Government of Fiji, World Bank, and Global Facility for Disaster Reduction and Recovery 2017.

Building resilience is expensive. Estimates for total investment needed for climate adaptation vary significantly across countries and depend on risk tolerance. The Climate Change Policy Assessment for Belize (IMF 2018a) assesses the need for resilient investment at 28 percent of GDP.4 The climate vulnerability assessment for Fiji estimates costs to be about 100 percent of GDP over 10 years (see Box 9.1).5Table 9.1 shows the cost of adaptation for some Pacific island nations.

Table 9.1.Costs of Resilience in 2020s, in Selected Pacific Countries ($ millions per year at 2018 international prices)
CountryAdaptation Costs for Coastal ProtectionCosts of Protecting Infrastructure from High Temperature and RainfallCost of Adaptation to Higher Cyclone Winds for Public BuildingsTotal ($, Millions)Total (Percent of 2018 GDP)
Marshall Islands1446917.219.3417519.936.5
Solomon Islands8930820--1093287.923.8
Sources: World Bank 2016; and IMF staff estimates.
Sources: World Bank 2016; and IMF staff estimates.

Donor support is at present heavily skewed toward postdisaster support instead of building ex ante resilience (Figure 9.2, panel 4). In 1990–2010, 85 percent of aid for disaster-related expenses was allocated to postdisaster recovery and humanitarian assistance. Only about 15 percent went toward building resilience. Climate financing is four times more focused toward mitigation than adaptation activities.

The Policy Experiment

The Setup

This section analyzes the returns of investment in resilient infrastructure, how they vary with economic and climate change parameters, and whether they are sufficiently high in net present value terms to cover the costs of investment. For this purpose, we use a multisector dynamic stochastic general equilibrium model to study the trade-off in building resilience. The model is a small open economy in which households purchase agricultural and manufacturing goods as well as services. There are five types of households: unskilled, skilled, government employees, entrepreneurs, and farmers. Households take prices and government policies as given. The government chooses tax policy and spending levels, including public investment. It can invest in infrastructure that is resilient and is not. The former is more expensive but is more durable.6 The latter is cheaper but depreciates at a higher rate when there is a natural disaster. Details of the model can be found in Annex 9.1 and in Guo and Quayyum (forthcoming).

Damage from Natural Disasters: 1980–2017 (Annual average, percent of GDP)

Source: IMF staff calculations, based on EM-DAT data.

Note: EM-DAT = Emergency Events Database.

The model is calibrated to six disaster-vulnerable economies.7 Country-specific data from this group inform assumptions made about the sizes of various sectors (for example, agriculture and manufacturing), factor endowments and allocation (for example, initial wealth, skill distribution of the labor force), the types of goods consumed by households, tax rates, disaster frequency, and damage (see Annex 9.2). Features of the production function, the labor participation rates, and depreciation of capital were estimated using national data.8

The disasters that hit countries can be divided into three types: small, medium, and large. Small disasters cause damage of less than 5 percent of GDP, medium disasters cause damage of 5 percent to 20 percent of GDP. Large disasters cause damage exceeding 20 percent of GDP. Table 9.2 shows the frequency of different-sized disasters for the six countries.

Table 9.2.Frequency of Disaster, Based on Experience over a 20-Year Period
Antigua and Barbuda111
St. Lucia401
St. Vincent and the Grenadines220
Source: IMF staff calculations, based on EM-DAT data.Note: Disaster size is based on the scale of damage (small refers to less than 5 percent of GDP damage; midsize refers to 5 percent to 20 percent of GDP damage; and large refers to GDP damage exceeding 20 percent).Note: EM-DAT = Emergency Events Database.
Source: IMF staff calculations, based on EM-DAT data.Note: Disaster size is based on the scale of damage (small refers to less than 5 percent of GDP damage; midsize refers to 5 percent to 20 percent of GDP damage; and large refers to GDP damage exceeding 20 percent).Note: EM-DAT = Emergency Events Database.

The depreciation rate for resilient and nonresilient capital varies with the size of shock. Small and medium shocks do not increase depreciation of resilient infrastructure over its nondisaster rate. Resilient capital depreciates when hit by a large shock, but at a rate lower than nonresilient capital would depreciate under a similar size of shock.9 In contrast, nonresilient capital depreciates in the face of all shocks.

Policy Scenarios

The model described in this chapter is used to study and compare the following two scenarios:

  • No action. The government does not invest in resilient capital but invests only in nonresilient capital to offset the depreciation of capital. However, when there is a natural disaster, the government rebuilds the damaged infrastructure with resilient capital. The cost of rebuilding is financed entirely by foreign grants.10
  • Build resilience. The government chooses to invest in resilient capital to offset the depreciation of capital.11 Given that resilient capital is more expensive, this requires additional financing of about 1 percent of GDP in nondisaster years. The additional financing is assumed to come from donor grants or a combination of grants and higher tax revenue. When there is a natural disaster, the destroyed infrastructure is rebuilt using resilient capital financed through grants, like in the previous scenario.

Key Insights

While policymakers spend about 1 percent of GDP in nondisaster years in the “build resilience” scenario, they save significantly in years when the countries are hit by disasters. Figure 9.4 shows the path of additional public investment for St. Vincent and the Grenadines under the two scenarios. In the “no action” scenario, no additional investment is made in nondisaster years (the orange line remains at zero). St. Vincent and the Grenadines faces two simulated medium-sized disasters and two small disasters over the 20-year period based on the simulation starting from a nondisaster year. After 20 years, the simulations would be back to the initial status for a new circle so that the situations in period 1 are identical to other situations after period 20. These can be seen in the four spikes in public investment. The blue line shows the path of public investment for a scenario in which policymakers invest in resilience. In nondisaster years, the blue line is higher than the orange line because of additional expenditure in resilience. However, the spikes in public investment are much smaller in disaster years.

Additional Public Spending on Infrastructure: St. Vincent and the Grenadines

Source: IMF staff calculations, based on simulation for a period of 20 years.

The difference in public spending between the two scenarios is calculated, along with the net present value of the difference as a percentage of initial-year GDP for each of the six countries.12 This is shown in Figure 9.5. For all six countries, the savings in lower recovery costs in disaster years from resilience building outweigh the additional expenses incurred in nondisaster years over the 20-year period. Net savings under the baseline scenario vary between 3 percent and 20 percent of the recipient’s GDP. The largest savings are in Fiji, which has the highest incidence of large disasters. The net savings increase significantly when one large disaster is added to the 20-year period, varying between 9 percent and 22 percent of GDP across the six countries.13

Savings from Building Ex Ante Resilience and Avoiding Large Recovery Costs

(Net present value, as a percent of first year’s GDP)

Source: IMF staff calculations, based on simulation for a period of 20 years.

Moreover, significant gains come from lower output losses in disaster years from building resilience beforehand. When public capital is more resilient, productive capacity shrinks less and fewer output disruptions occur. The gains range from 2 percent of initial-year GDP in St. Vincent and the Grenadines to about 8 percent of initial-year GDP in Dominica, where large disasters have been frequent in recent years (Figure 9.6). If a higher frequency of disasters is assumed, the gains range from about 4 percent of GDP in Fiji to over 8 percent in Dominica. These numbers are based on a scenario in which the additional investment needed for building resilience is financed through external grants. When VAT rates are raised to pay part of the higher cost of resilience, the output gains are only marginally smaller (less than 0.5 percent of GDP), but consumption drops significantly.14

Gains from Lower Output Loss from Building Ex Ante Resilience (Percent of first year’s GDP, net present value)

Source: IMF staff calculations, based on simulation for a period of 20 years.

Improving the efficiency of investment spending could significantly increase the gains from investment in building resilience. As noted in Chapter 3, improvements in infrastructure governance can impact the efficiency of spending significantly—countries can boost the quality and volume of infrastructure with limited increases in spending. In the model in this chapter, this can be captured through the production function. A test is carried out to determine how better public infrastructure governance affects the returns from investment in building resilience in vulnerable areas by increasing the productivity of public capital— better infrastructure governance would increase the response of output to an increase in the capital stock. In general, this leads to stronger output gains from investing in resilience. In particular, increasing the elasticity of output of public capital from 0.14 in the baseline to 0.2 leads to output gains in excess of the baseline estimates of between 0.3 percent of GDP (in St. Vincent and the Grenadines) and 1.3 percent of GDP (in Antigua and Barbuda). This is shown in Figure 9.7.15

Additional Saving with Higher Productivity of Public Investment (Percent of GDP)

Source: IMF staff calculations, based on simulation.

Policy Implications

The results support switching resources toward building resilience in countries that are vulnerable to natural disasters. By spending more on resilience, policymakers can expect to save on postdisaster recovery in disaster-vulnerable countries. As a significant part of the postdisaster recovery costs in small vulnerable states is usually financed by donors, they could expect to save on costs by providing more resources for resilience building.

Vulnerable countries and donors would need to take significant steps to raise financing necessary to build resilience. Donor assistance is particularly important. Concessional financing, preferably grants, would help countries build resilience while ensuring debt sustainability. Climate funds—which are also financed by the international community—are another source of funding but would need to improve access for projects by simplifying some of their administrative requirements, which many disaster-vulnerable states find cumbersome.16 However, country authorities will also need to generate fiscal resources through stronger domestic revenue mobilization and possibly switching low-priority expenditure toward resilience building. Countries will also need to spend better and improve efficiency of public spending (through better governance).

Annex 9.1. Model Setup and Parameters

The model consists of a small open economy with three consumption goods: agriculture, manufacturing, and services. There are five types of households: unskilled, skilled, government employees, entrepreneurs, and farmers. Households solve dynamic optimization problems taking prices and government policies as given. In addition to consumption decisions, the different participants of the economy make decisions that affect output and incomes, taking into account uncertainty from natural disasters:

  • Unskilled households. Choose to work for farmers or in the informal sector.
  • Skilled households. Choose to work for entrepreneurs or to migrate.
  • Government employees. Choose to work for the government or to migrate.
  • Farmers. Hire labor to produce agriculture goods and invest.
  • Entrepreneurs. Hire labor for the formal manufacturing and services sector and invest.

Three goods are produced: (1) agriculture (produced by farmers), (2) manufacturing (produced by entrepreneurs), and (3) services (produced by entrepreneurs in the formal sector and by unskilled workers in the informal sector). Services are produced by both the formal and the informal sectors.

As input for the production function, the agriculture sector uses unskilled labor and capital. The manufacturing sector employs skilled labor and invests in capital to produce manufacturing goods. Production in the services sector can be formal using skilled labor or informal using unskilled labor. Agriculture and manufacturing goods are tradable and sold domestically or internationally, while service goods are nontradable and sold only domestically.

Households decide how much to consume of each good. In addition, skilled households have an occupational choice; they decide the share of time devoted to work in the domestic formal sector and the share time devoted to work abroad.17 Unskilled households also have an occupational choice between working in the agriculture sector or working in the informal sector. Government workers have a similar occupational choice, since they can receive a public sector wage or migrate and work abroad.

The possibility of labor migration is a key feature of the model, with important macroeconomic implications. The model allows households to choose between working in the local economy, or migrating to work in other countries and sending back remittances. Notice that this is a macro-critical feature for small states and small open economies, including in the Caribbean, Central America, and Asia Pacific. Migration typically has a significant impact on both the quantity and the quality of labor, given that migration can be skill biased. For example, Dominica’s population has shrunk in recent years despite an average natural growth rate, with emigration being more prominent for high-skilled workers, at 80 percent of the population. These skilled workers contribute to the economy by sending remittances, of around 5 percent of GDP per year.

Farmers own their capital and decide how much unskilled labor to hire and how much to invest in capital to produce the agriculture goods. Entrepreneurs also own capital; they decide how much skilled labor to hire to produce formal services and manufacturing and invest in capital that is used as an input on the production of manufacturing goods.

The government sector includes a granular menu of fiscal policy instruments. The government collects tax revenue (VAT, corporate taxes, and personal income tax) and nontax revenue (mainly through citizenship by investment programs and donor grants). Government revenue is used to fund expenditure (including public sector wages, public investment in resilience and nonresilience capital, and transfers) and to service public debt. Public investment plays a significant role in this economy and affects the productivity of the manufacturing and agriculture sectors. Government decisions ultimately affect budget constraints and the accumulation of public debt. The model structure is summarized in Annex Table 9.1.1.

Annex Table 9.1.1.Model Structure
AgricultureFarmersUnskilled labor and capitalConsumptionTradable
ManufacturingEntrepreneursSkilled labor and capitalConsumption and investmentTradable
ServicesUnskilled labor and entrepreneursSkilled and unskilled laborConsumptionNontradable
Source: IMF staff, based on country staff report.
Source: IMF staff, based on country staff report.

Resilient and Nonresilient Infrastructure

Public infrastructure contributes to the manufacturing and agriculture sector as public capital with sector-specific elasticity of production. In the modeling, we would like to consider the impact from natural disasters and climate change on

infrastructure investment and disaster financing. Therefore, we have two types of infrastructure: resilient Ktr and nonresilient Ktn. They function the same and depreciate the same in regular time and Ktn+Ktr=Ktg. However, when the economy is hit by natural disasters, the resilient infrastructure does not sustain extra damage unless the disaster is large, inflicting damage of more than 20 percent of GDP. Nonresilient infrastructure depreciates more when there is a natural disaster and the additional depreciation depends on the intensity of the disaster. When a unit price of nonresilient infrastructure is set as 1, the unit price for resilient infrastructure is 1.1 in the baseline specification. The price is varied from 1.05 to 1.2 for robustness checks.

Natural Disaster

A natural disaster generates the separation of both total productivity loss θ instantly and loss in public capital with a one-time depreciation drop δd. The productivity shock is the same for all sectors. The θ is calibrated for small, middle, and large shocks separately to match the real GDP drop, based on the historical disaster series from the Emergency Events Database (EM-DAT). We define the size of the disaster based on its impact on the damage-to-GDP ratio. If the ratio is below 5 percent, it is regarded as a small natural disaster shock. If it is between 5 percent and 20 percent, it is middle-sized. Otherwise it is a large natural disaster shock. For small states in this study, Table 9.2 shows the frequency of different types of shocks for countries in 20 years.

We assume that GDP would drop 5 percent for large disaster shocks, 1 percent for middle ones, and 0.3 percent for small shocks, based on empirical calculation of the impact of various-sized disasters on growth.

The following depreciation is assumed for when public infrastructure is hit by natural disaster, which varies with the intensity of disaster (Annex Table 9.1.2).

Annex Table 9.1.2.Depreciation of Nonresilient and Resilient Infrastructure (By size of natural disaster)
Source: IMF staff.
Source: IMF staff.
Annex 9.2. Application to Small States

The model is calibrated to match the quantitative parameters of the six small states’ economies separately. The calibration accounts for sector sizes, labor participation, capital investment, and intersector links, and for the consumption basket for different goods.

Preferences. Households’ preferences over manufacturing goods (i//) and services (γ) are calibrated so that consumption shares in the model match those in the consumer price index (CPI) basket, mapping the different types of goods and services to the sectors in the model.

Labor force. Labor market parameters, including the distribution of labor across different types of households in the model, are based on data from each country’s national administration.18 Specifically, the sectoral data on employment are allocated as follows (Annex Table 9.2.1): (1) government workers (μg): public sector, (2) skilled workers (μs): manufacturing, utilities, trade, tourism, and financial sector, and (3) unskilled workers (μu): agriculture, fishing, mining, and construction. The remaining sectors are distributed between entrepreneurs (μe) and farmers (μf). Wages in the United States (wus) are calibrated so that remittance flows in the model match the actual data (Annex Table 9.2.2).

Annex Table 9.2.1.Parameters
μgShare of government workers
μsShare of skilled workers
μuShare of unskilled workers
μeShare of entrepreneurs
μfShare of big farmers
r*Interest rate in government debt
βDiscount factor
c¯Lower bound of agricultural consumption
aEmigration elasticity
amTradable elasticity to private capital
a*Agriculture elasticity to private capital
agElasticity to public capital
δPrivate capital depreciation
δgPublic capital depreciation
δhHuman capital depreciation
HHuman capital stock
Source: IMF staff, based on country staff reports.
Source: IMF staff, based on country staff reports.
Annex Table 9.2.2.Moment Calibration Summary
Preferences (percent)
ψManufacturing share in total consumptionYe s
γServices share in total consumptionYe s
Economic Indicators (percent of GDP)
zaAgricultural outputYe s
ze,sServices outputYe s
zsInformal sector outputYe s
wusRemittancesYe s
Fiscal Policy (percent of GDP)
τa, τm, τsRevenue from consumption taxYes
τwRevenue from personal income taxYe s
τkRevenue from corporate taxesYe s
GrGrantsYe s
NRNontax revenuesYe s
wgPublic sector wage billYe s
Tu, Ts, Tf, Te, TgTransfers to householdsYe s
θGDP deviation from nondisaster periodYe s
Source: IMF staff, based on country staff reports.
Source: IMF staff, based on country staff reports.

Economic sectors. The production function is assumed to be Cobb-Douglas. Productivity in the agricultural (za), formal (ze,s), and informal services (zs) sectors is calibrated so that the sizes of the sectors in the model match the national accounts data. Elasticity to private capital (αm, α*) is calculated as the difference between total capital income share and capital income attributed to structures,

while elasticity to public capital (αg) is assumed to equal capital income because of structures in which capital income shares are estimated by Valentinyi and Herrendorf (2008) (Annex Table 9.2.1). Capital stock is calculated as the stock in the previous period (net of depreciation) increased with investments in which private (δ) and public physical capital depreciation (δg) are calculated as the weighted average of depreciation rates by type of capital from Feenstra, Inklarr, and Timmer (2015) with capital income shares estimated by Valentinyi and Herrendorf (2008) serving as weights. Human capital depreciation (δh) is calculated based on mortality rates in peer countries (United Nations 2015). Human capital stock (H) is normalized to 1.

Fiscal policy. Government revenue and expenditure parameters are attuned to central government data (Annex Table 9.2.2). Specifically, consumption (τa, τm, τs), personal income (τw), and corporate tax rates (τk) are calibrated to yield revenue close to effective revenue collections. Nontax revenues (NR) are calibrated to match flows mainly from the government budget plan. Similarly, grants (Gr) are aligned with actual current and capital grants. Transfers to households are calibrated to match actual spending on transfers, while public sector wages (wg) are calibrated so that the public sector wage bill is in line with actual compensation of employees. Interest rate on government debt (r*) is set as the implied interest rate on outstanding debt (Annex Table 9.2.1).


    ArslanalpSerkanFabianBornhorstSanjeevGuptaandElsa Sze. 2010. “Public Capital and Growth.” IMF Working Paper 10/175International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    BarroRobert. 1990. “Government Spending in a Simple Model of Endogenous Growth.” Journal of Political Economy98 (5): 10325.

    BarroRobert andXavierSala-i-Martin. 1992. “Public Finance in Models of Economic Growth.” Review of Economic Studies59 (4): 64561.

    • Search Google Scholar
    • Export Citation

    BayoumiTamimSaadQuayyum andSibabrataDas. Forthcoming. “Growth at Risk from Natural Disaster.” IMF Working PaperInternational Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    BevanDavidL. andSamanthaJane Cook. 2015. “Public Expenditure Following Disasters.” Policy Research Working Paper 7355World BankWashington, DC.

    • Search Google Scholar
    • Export Citation

    CavalloEduardo andIlanNoy. 2011. “Natural Disasters and the Economy: A Survey.” International Review of Environmental and Resource Economics5 (1): 63102.

    • Search Google Scholar
    • Export Citation

    CavalloEduardoSebastianGalianoIlanNoyandJuan Pantano. 2013. “Catastrophic Natural Disasters and Economic Growth.” Review of Economics and Statistics95 (5): 154961.

    • Search Google Scholar
    • Export Citation

    FeenstraRobertC.RobertInklaar andMarcel P.Timmer2015. “The Next Generation of the Penn World Table.” American Economic Review105 (10): 315082.

    • Search Google Scholar
    • Export Citation

    FutagamiKoichiYuichiMorita andAkihisaShibata. 1993. “Dynamic Analysis of an Endogenous Growth Model with Public Capital.” Scandinavian Journal of Economics95 (4): 60725.

    • Search Google Scholar
    • Export Citation

    Government of FijiWorldBankandGlobal Facility for Disaster Reduction and Recovery. 2017. “Fiji 2017: Climate Vulnerability Assessment—Making Fiji Climate Resilient.” World BankWashington, DC.

    • Search Google Scholar
    • Export Citation

    GuersonAlejandroWeiGuoBalazsCsonto andMarinaMendes Tavares. Forthcoming. “A Model for Fiscal Policy Analysis in Small States: An Application to Dominica.” IMF Working PaperInternational Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    GuoWei andSaadQuayyum. Forthcoming. “Building Resilience to Natural Disaster in Vulnerable States: Savings from Ex Ante Interventions.” IMF Working PaperWashington, DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund (IMF). 2016. “Small States’ Resilience to Natural Disasters and Climate Change—Role for the IMF.” IMF Policy PaperWashington, DC.

    • Search Google Scholar
    • Export Citation

    IMF. 2018a. “Belize: Climate Change Policy Assessment.” IMF Country Report 18/239Washington, DC.

    IMF. 2018b. “Dominica: Staff Report for the 2018 Article IV Consultation.” IMF Country Report 18/265Washington, DC.

    IMF. 2019. “Building Resilience in Countries Vulnerable to Large Natural Disasters.” IMF Board PaperWashington, DC.

    KnutsonThomasJoseph J.SirutisGabriel A.VecchiStephenGarnerMingZhaoHyeong-SeogKimMorrisBenderandothers. 2013. “Dynamical Downscaling Projections of Twenty-First-Century Atlantic Hurricane Activity: CMIP3 and CMIP5 Model-Based Scenarios.” Journal of Climate 26.

    • Search Google Scholar
    • Export Citation

    LeeDongyeolHuanZhang andChauNguyen. 2018. “The Economic Impact of Natural Disasters in Pacific Island Countries: Adaptation and Preparedness.” IMF Working Paper 18/108International Monetary FundWashington, DC.

    • Search Google Scholar
    • Export Citation

    LoayzaNormanEduardoOlaberriaJameleRigoliniandLuc Christiaensen. 2012. “Natural Disasters and Growth: Going Beyond the Averages.” World Development40 (7): 131736.

    • Search Google Scholar
    • Export Citation

    MartoRicardoChrisPapageorgiou andVladimirKlyuev. 2018. “Building Resilience to Natural Disasters: An Application to Small Developing States.” Journal of Development Economics135 (C): 57486.

    • Search Google Scholar
    • Export Citation

    NoyIlan. 2009. “The Macroeconomic Consequences of Disasters.” Journal of Development Economics88 (2): 22131.

    NoyIlan andAekkanushNualsri. 2007. “What Do Exogenous Shocks Tell Us about Growth Theories?University of Hawaii Working Paper 07–28Honolulu.

    • Search Google Scholar
    • Export Citation

    RaddatzClaudio. 2009. “The Wrath of God: Macroeconomic Costs of Natural Disasters.” Policy Research Working Paper 5039World BankWashington, DC.

    • Search Google Scholar
    • Export Citation

    StroblEric. 2012. “The Economic Growth Impact of Natural Disasters in Developing Countries: Evidence from Hurricane Strikes in the Central American and Caribbean Regions.” Journal of Development Economics97 (1): 13041.

    • Search Google Scholar
    • Export Citation

    United Nations. 2015. “World Population Prospects: The 2015 Revision.” United NationsNew York.

    United Nations Environment Programme. 2016. The Adaptation Finance Gap Report.

    United Nations Environmental ProgrammeNairobi.ValentinyiAkos andBertholdHerrendorf. 2008. “Measuring Factor Income Shares at the Sectoral Level.” Review of Economic Dynamics11 (4): 82035.

    • Search Google Scholar
    • Export Citation

    World Bank. 2016. “Pacific Possible: Climate and Disaster Resilience.” World BankWashington, DC.

    World Bank. 2017. “Pacific Possible: Long-Term Economic Opportunities and Challenges for Pacific Island Countries.” Washington, DC.


See IMF (2016) for a detailed discussion on international financing of natural disasters and climate change.


In this scenario, the country is hit by one additional large natural disaster exceeding 20 percent of GDP damage in the 20-year period.


The estimated growth is the average of the two years based on IMF (2018a).


Climate Change Policy Assessments are joint World Bank and IMF exercises carried out as pilots for assessing policy gaps in mitigation and adaptation strategies to climate change.


The climate vulnerability assessment was prepared by the Government of Fiji with assistance from the World Bank to assess interventions and investments needed to make the country climate resilient.


For illustrative purposes, we assume resilient infrastructure is 10 percent more expensive than non-resilient infrastructure, while Marto, Papageorgiou, and Klyuev (2018) assumed costs are the same, and Guerson and others (forthcoming) assumed resilient infrastructure costs 20 percent more. The robustness of results is tested by varying this assumption.


These countries suffered average annual damage of 0.4 to 16 percent of GDP (see Figure 9.3).


The discount factor was borrowed from the literature and set to 0.9.


For simplicity, it is assumed that resilient and nonresilient capital depreciate at the same rate in nondisaster years.


Many of the disaster-vulnerable countries, especially in the Caribbean, are fiscally constrained (see IMF 2019) and rely heavily on donor support after a disaster. For the sake of simplicity, it is assumed that the rebuilding costs are all donor financed.


It is assumed that the investment rates are the same in the two scenarios.


A discount factor of 5 is used for the net present value calculations.


These savings are independent of whether the spending on resilience is financed through grants or a combination of grants and tax revenue.


In this scenario about one-third of the cost of building resilience is borne by the vulnerable countries by increasing the VAT rate. Increasing the burden-sharing ratio through higher taxes lowers GDP, but only marginally, in the model.


In the baseline, it is assumed that a 1 percent increase in public capital leads to a 0.14 percent increase in output, based on data from Dominica. In the new scenario, the assumption is that a 1 percent increase in public capital leads to a 0.2 percent increase in output, based on the average estimate in Arslanalp and others (2010) for countries that are not members of the Organisation for Economic Co-operation and Development.


Climate funds provide financing for climate mitigation and adaptation activities.


This occupational choice captures the brain drain problem faced by many countries in the Caribbean.


The latest available data are for 2015.

    Other Resources Citing This Publication