A. The Role of Fiscal Buffers

Following a natural disaster, the government needs to secure funding for emergency relief and reconstruction efforts at the same time as revenues may be adversely affected. The government may reallocate budget spending towards disaster relief (often by cutting planned capital spending) and draw down its deposits in the central bank or commercial banks to address immediate relief needs. The international community (including other governments, international organizations, NGOs and the private sector) may also provide grants and aid to help the disaster-devastated country to recover quickly, though the amounts and timing of disbursements are quite uncertain. External insurance, including catastrophe bonds, is another option for financing, though it might be costly and has limited triggers for payouts.

Beyond the immediate response to a disaster, in order to finance recovery and reconstruction, the government may also seek to borrow from domestic or external sources, though the borrowing has an adverse effect on debt sustainability, especially for highly indebted countries. In the absence of sufficient funding for reconstruction, much of the financial burden of recovery will fall on the private sector, especially those who have been most adversely affected by the disaster. In addition, government resources will tend to be redirected away from previously planned investments and current expenditures towards relief efforts.

Fiscal buffers have a crucial role to play in managing the fiscal impacts of natural disasters and achieving sustainable and inclusive growth. Adequate fiscal buffers would reduce dependence on uncertain foreign donations, ensure quick disbursement for recovery and relief efforts, minimize disruption of pre-existing spending plans on health, education, and infrastructure, and provide flexibility in addressing post-disaster liquidity needs.

B. Definition of Fiscal Buffers

In general, fiscal buffers refer to fiscal space available to cover unexpected expenditure increases or revenue shortfalls, created by saving budgetary resources and reducing public debt in good times. Fiscal space is defined as the extent to which a government can raise spending or lower revenues without endangering market access and putting debt sustainability at risk (IMF, 2016).

In this paper, we define fiscal buffers for natural disasters as government financial assets which can be used to address relief and reconstruction needs, helping to insulate budgeted government spending from major disruption. They include government deposits and funds, which can be disbursed expeditiously in the event of natural disaster shocks without much restriction. They also include funds, which can be disbursed if withdrawal rules are changed, but do not include funds which are established for other purposes and which are not intended for use in dealing with substantial fiscal shocks. Consequently, although government cash buffers have great flexibility, they would normally be too small to be considered as a buffer for natural disasters. Conversely, a sovereign wealth fund (SWF) may be large, but may not be able to be used to cover natural disaster costs (Appendix).⁵

Currently, there are six PICs with contingency funds for natural disasters (Table 1). Fiji has established the National Disaster Relief and Rehabilitation Fund in 2004, which receives an annual contribution of F$ 1 million (US$ 0.5 million) from the government. In Marshall Islands and Micronesia, the Disaster Assistance Emergency Fund was established as part of the Compact of Free Association with the U.S. and is used by those countries to assist with disaster-related expenses (Box 1).

Table 1:

Natural Disaster Funds in PICs

\1. Pacific Disaster Contingent Savings Facility under the ADB-supported Pacific Disaster

Table 1:

Natural Disaster Funds in PICs

	Fund	Year Established
Fiji	National Disaster Relief and Rehabilitation Fund (known as the Prime Minister’s Fund)	2004
	Ongoing Contingency Funds for Disaster Risk	2015
Marshall Islands	Disaster Assistance Emergency Fund	2004
Micronesia	Disaster Assistance Emergency Fund	2004
Solomon Islands	Natural Disaster Council Fund	1989
Tonga	National Emergency Fund	2008
Tuvalu	Climate Change and Disaster Survival Fund	2015

\1. Pacific Disaster Contingent Savings Facility under the ADB-supported Pacific Disaster

View Table

Table 1:

Natural Disaster Funds in PICs

	Fund	Year Established
Fiji	National Disaster Relief and Rehabilitation Fund (known as the Prime Minister’s Fund)	2004
	Ongoing Contingency Funds for Disaster Risk	2015
Marshall Islands	Disaster Assistance Emergency Fund	2004
Micronesia	Disaster Assistance Emergency Fund	2004
Solomon Islands	Natural Disaster Council Fund	1989
Tonga	National Emergency Fund	2008
Tuvalu	Climate Change and Disaster Survival Fund	2015

\1. Pacific Disaster Contingent Savings Facility under the ADB-supported Pacific Disaster

In Tonga, a National Emergency Fund with around US$ 7.8 million was established in 2008. Tuvalu established a Climate Change and Disaster Survival Fund in 2015, with initial funding of AUD 5 million. The Consolidated Investment Fund (CIF) in Tuvalu can be considered as a fiscal buffer for natural disasters, as the government can freely draw down the CIF to finance budget spending. In Solomon Islands, the National Disaster Council Fund, which was established in 1989, was misused and often diverted away from disaster response activities. The fund has not received an appropriation since 2008.

The Compact Trust Funds (CTFs) in Marshall Islands and Micronesia, and the Climate Change Trust Fund (CCTF) in Tonga, however, do not fit the definition of a fiscal buffer for natural disasters. The CTFs will be disbursed as grants from 2024 onwards and no disbursement is allowed prior to 2024, while the purpose of the CCTF is to finance small, community-based climate adaptation and mitigation projects and fund the climate component of non-community-based projects.

The ADB has been working with several PICs to help strengthen their resilience to natural disasters under the Pacific Disaster Resilience Program by establishing Pacific Disaster Contingent Savings Facility as well as by strengthening the framework for disaster risk management. In December 2017, the ADB decided to provide a total of US$ 15 million in financing to Samoa, Tonga and Tuvalu.⁶ After Cyclone Gita hit Tonga in February 2018, the ADB provided US$ 6 million to the Government of Tonga to help fund priority early recovery activities.

C. Institutional Framework for Fiscal Buffers for Natural Disasters

An effective institutional framework is needed to avoid the misuse of fiscal buffers, especially natural disaster funds, and enhance the transparency and accountability of the funds. The IMF (2016) recommends that a well-designed framework should have the following characteristics:

• The fund should be consolidated with budget information to allow the assessment of the overall fiscal situation; at a minimum, the fund balance should appear in financial statements, and drawdowns from the fund should appear in budget execution reports.

• There should be a standing appropriation that allows for spending immediately after a certain trigger event, such as a declaration of a disaster emergency by the executive.

• The fund should generally apply normal PFM rules. It should have clear rules governing the use of the resources; follow normal government accounting standards; prepare and publish audited financial statements; and define governance rules. Following these rules promotes transparency. However, procurement rules for immediate disaster response should be adjusted to allow for quicker procurement.

• It should be limited to respond to disasters with large fiscal impacts: hence, drawdowns should only start above a minimum fiscal cost. Smaller expenditure needs should be covered through budget contingencies.

• It may not be appropriate to let the fund get too large if: (i) its primary purpose is to cover immediate relief and recovery expenses while longer-term rebuilding is financed by borrowing, or (ii) it may generate pressures to tap it for other purposes or discourage international donors from providing grants or loans.

• The financial investment strategy of the fund should be to maintain a high degree of liquidity, given the potential urgency of relief expenditures. For PICs, these assets may be better kept offshore because domestic financial markets may be under stress after disasters and because relief and recovery efforts may involve substantial import content. Post-disaster repatriation of offshore investments would also strengthen the balance of payments at a time of economic weakness.

D. Opportunity Cost of Holding Fiscal Buffers

Given the high infrastructure needs in PICs, an argument can be made that there is a tradeoff between public spending on infrastructure and holding fiscal buffers. Accumulating public savings instead of investing in productive infrastructure projects forgoes the rate of the social return on the associated public investment. However, this argument does not reflect the fact that eventually the saved funds will be used for relief and recovery. There is a cost to waiting to use the buffer. But in the absence of a buffer, there would be a high cost to either delayed recovery owing to lack of resources, or the cost of borrowing to rebuild.

A slightly different argument is that investing in resilient infrastructure can directly reduce the cost of recovery from disasters, so that the two are to some extent substitutes. While this is likely to be true up to a point, there are limits to the amount of physical resilience that can be achieved (it is hard to protect crops from a cyclone, or houses from a tsunami), so financial and physical resilience become complements beyond that point.

Paying down debt might be better than accumulating fiscal buffers, especially for countries with high indebtedness. If the cost of public debt is higher than the benefits of holding fiscal buffers, the authorities may consider early debt repayment. A reduction of expensive debt may enhance future borrowing capacity with favorable terms. Nonetheless, reducing debt is not equivalent to building a buffer in an important way. Buffers can be used essentially immediately, which may be extremely important following a sudden disaster such as a cyclone, tsunami, or earthquake. Lower debt implies more room to borrow, but arranging a loan may take some time. In practice, authorities could consider a combination, with an asset buffer to cover immediate needs, and enough room to borrow to finance some of the longer-term recovery needs.

One way to reduce the delay associated with borrowing to finance post-disaster recovery is to have a pre-arranged line of credit that can be activated quickly. The World Bank has a program called the development policy loan with a Catastrophe Deferred Drawdown Option (Cat DDO), which is a contingent financing line that provides immediate liquidity following a natural disaster. Funds become available for disbursement after the drawdown trigger, which is typically the declaration of a state of emergency, is met. At approval a country must have an adequate macroeconomic policy framework and a satisfactory disaster risk management program in place (or under preparation). Some PICs, such as Vanuatu, are currently discussing the Cat DDO with the World Bank.

The attractiveness of a line of credit rather than building a buffer of own funds is not clear. For countries where natural disasters are quite infrequent, a line of credit may be preferred since it does not tie up valuable capital which could be used for development purposes. But for countries that are particularly vulnerable to frequent disasters, the line of credit is highly likely to be drawn. And such borrowing also needs to be repaid. In this case, the budgetary implications may be very similar to saving in a buffer fund. A buffer of own funds will have fewer strings attached, and can be as large as the country wishes. However, precisely because activating a line of credit is more restrictive, it may be less likely to be used for unintended purposes than a buffer of own funds. In other words, the differences between a buffer and a line of credit are not simply financial, but may be most significant in terms of governance.

A. Literature Review

Most studies estimating the economic costs of natural disasters use data from Emergency Events Database (EM-DAT) and focus on the impact on growth (e.g. Klomp And Valckx, 2014; Felbermayr and Gröschl, 2014; Fomby et al., 2013; Loayza et al., 2012; Cavallo et al., 2013)⁷. Recent studies on the fiscal effects of natural disasters generally use a broad range of countries as samples. Lis and Nickel (2010) examine the effect of extreme weather events on budget balances, using annual data for 138 countries in a fixed effects regression model. They conclude that, on average, budget balances would narrow by 0.23 percent to 1.4 percent of GDP, depending on the country income. Noy and Nualsri (2011) include quarterly data from 22 developed countries and 20 developing countries in a panel VAR model to quantify the fiscal consequences after natural disasters. They find that fiscal behavior is counter-cyclical in developed countries, and pro-cyclical in developing countries in the aftermath of disasters. This paper contributes to the literature by estimating the fiscal impacts in 12 PICs from severe natural disasters, and the lagged effects from disasters on fiscal indicators.

B. Data

We use annual natural disaster data during the period 1980–2016 from the EM-DAT database, which records disasters that meet the following criteria: (i) 10 or more people dead; (ii) 100 or more people affected; (iii) a declaration of a state of emergency; and (iv) a call for international assistance.⁸ All the economic indicators used in the estimation are from the World Economic Outlook and World Development Indicators databases. The panel is unbalanced mainly because of the limited availability of fiscal data.⁹

C. Natural Disaster Intensity Measure

Not all-natural disasters have significant economic effects on a country, either because the event is relatively small, or because it affects a sparsely populated or scarcely developed area (Table 2). For our purposes, we want to focus on disasters likely to have a substantial macroeconomic and fiscal impact. Previous literature constructed disaster intensity measures using information on the population affected and the numbers of deaths (Fomby et al, 2013; Becker and Mauro, 2006). However, different types of natural disasters influence the country’s economy and population differently. Historical data show that droughts affect more people than other types of disasters, but storms cause the most economic damage. This paper adopts the same intensity measure as proposed by Lee, et al. (2018), which takes account of both the population affected and economic damage.

Table 2:

Average Damage per Disaster, by type (1980 -2016)

Note: Others includes volcanic activity, epidemic, landslide, mass movement, and wildfireSources: EM-DAT

Table 2:

Average Damage per Disaster, by type (1980 -2016)

	Population Affected (number of person)	Total Damage (USD thousands)
Storm	36,629	62,463
Drought	290,931	45,000
Flood	27,177	26,843
Earthquake	3,942	20,952
Others	9,992	70,825

Note: Others includes volcanic activity, epidemic, landslide, mass movement, and wildfireSources: EM-DAT

View Table

Table 2:

Average Damage per Disaster, by type (1980 -2016)

	Population Affected (number of person)	Total Damage (USD thousands)
Storm	36,629	62,463
Drought	290,931	45,000
Flood	27,177	26,843
Earthquake	3,942	20,952
Others	9,992	70,825

Note: Others includes volcanic activity, epidemic, landslide, mass movement, and wildfireSources: EM-DAT

Population and damage are used as follows. Among the 204 observations in the sample period, not all have information on both economic damage and population affected. Therefore, the intensity of natural disasters is based primarily on the percentile rank of economic damage to GDP ratio and, secondarily, on the percentile rank of population affected in cases where damage to GDP is not available. In this paper, we define severe disasters as those that are above 75th percentile. This threshold is chosen based on trial and error approach. The results from regressions with different percentile natural disaster dummies show that disasters that are above 75th percentile have a statistically significant impact on GDP.

D. Definition of Fiscal Cost

The fiscal cost of natural disasters is defined as the sum of the increase in expenditure and decrease in revenue of the central government, excluding grants, in the aftermath of natural disasters. The data, therefore, do not include the impact on other levels of government or state enterprises.

The fiscal cost of a natural disaster is different from the economic cost, which usually covers damage and loss. Economic costs include not only costs borne by the government sector but also those borne by the private sector. For instance, the total economic value of the effects caused by Tropical Cyclone Pam, which struck Vanuatu in March 2015, is estimated to be approximately 64 percent of GDP (39 percent for damage and 25 percent for loss), while total fiscal costs after Pam are estimated to be around 16 percent of GDP. The fiscal cost of natural disasters is estimated below.

E. Empirical Specification

This paper uses a panel VAR to estimate the impact of natural disasters on fiscal expenditure and revenue in PICs, similar to Noy and Nualsri (2010). The panel VAR approach is adopted to capture the linear interdependencies between government revenue and government expenditure. Natural disasters are treated as exogenous shocks with contemporaneous and lagged macro-fiscal impacts. Given that individual PICs do not have enough high frequency data to run time-series analysis, this paper pools the experiences of the PICs to generate a regional average impact of disasters. While country-specific estimates might be preferred at first glance, it should also be borne in mind that the impact of natural disasters vary greatly even for a single country. Consequently, the within-country variance of disaster effects may be as great as the across-country variance. In estimated models, we estimate separately the impact of disasters on expenditure and revenues. In addition, we use a measure of revenues excluding grants. Most PICs depend on official transfers as a major source of government revenue. After natural disasters, these countries tend to get more grants to help cover the cost of rebuilding infrastructure and disaster relief. The basic model specification is:

This model includes two main endogenous variables, Y_i,t: (i) government expenditure (in percent of GDP) and (ii) government revenue excluding grants (in percent of GDP). X_i,t represents exogenous variables: natural disaster dummy (over 75^th percentile) and grants (in percent of GDP).

Y_i,t is a (1x2) vector of dependent variables, X_i,t is a (1x2) vector of exogenous variables; u_i is a vector of country fixed effects; λ_t is a vector of time fixed effects; and 𝑒_𝑖𝑡 is a vector of idiosyncratic errors. The country fixed effects capture the country-specific characteristics that are not quantified in this model, e.g. political system, infrastructure capacity, or land area. A and B are both (2x2) matrices of parameters to be estimated.

Since the economic impact of natural disasters last several years, we included lags of the natural disaster dummy in the following model:

D_i,t is the natural disaster dummy, and X_i,t is grants (in percent of GDP). To test the robustness of the model, we also include real GDP growth, and Australian and US real GDP growth as controls in alternative specifications.

The parameters can be estimated by using fixed effects or equation-by-equation ordinary least squares (OLS) (Abrigo and Love, 2015). For fixed effects estimators, within transformation eliminates country fixed effects, however, the estimator is still biased with the lagged dependent variable on the right-hand side (Nickell, 1981; Baltagi, 2013). Alternatively, the estimation can be done using the generalized method of moments (GMM), which gives the consistent estimates for a panel VAR and may yield efficiency gains from estimation a system of equations (Holtz-Eakin, Newey and Rosen, 1988). Therefore, we use GMM to estimate the model and specifically forward orthogonal deviation or Herlmert transformation, which was introduced by Arellano and Bover (1995). In contrast to the first difference transformation, the Herlmert transformation removes the mean of all available future observations, and therefore avoids data loss and enables lagged regressors to be used as instruments to estimate the coefficient in GMM. Country fixed effects are eliminated by this transformation.

F. Results

The estimation results show that the impact of natural disasters on government expenditure is likely to extend to the third year after the initial disaster year, while the impact on revenue excluding grants is not statistically significant. These findings were robust across numerous variations on the model specification. On average, a severe natural disaster is likely to increase government expenditure by 13.8-20.6 percent of GDP over a three-year period.¹⁰ In the disaster year, government expenditure is likely to increase by 6.4 percent, and two years after the disaster, government expenditure tends to further increase by 6.8 percent. This paper’s results complement the economic impacts estimated by Lee et al (2018): one severe natural disaster would cause the country about a 2 percent decrease in real GDP growth in the disaster year. It may be noted that the finding that government (net) spending typically increases following a disaster implies that, with respect to this source of shocks, the fiscal response if generally counter-cyclical.

The persistent increase in public expenditure is not surprising. A rise in government spending on relief and emergency repairs to infrastructure in the immediate aftermath of a disaster is expected. However, beyond the immediate responses to a disaster, there is the longer process of rebuilding infrastructure, especially housing, schools, clinics and so on, and this may take several years. This is especially likely to be the case if rebuilding is dependent on external aid, as this may take time to deliver. Yang (2008) shows that the official development assistance flow increases significantly after natural disasters, but with a two-year lag.

The non-significant results on government revenues could be potentially explained by the special characteristics in PICs. Without highly diversified economies, most PICs depend on few sectors (e.g. tourism, fishing license fees, or natural resources) for government revenues. Natural disasters could have significant negative effects on tourism-oriented countries, including Fiji, Vanuatu and Palau, if the tourism infrastructure is significantly damaged. However, the impact of natural disasters countries dependent on fishing license fees is limited. Government revenues of countries that are big oil or commodity exporters are highly sensitive to the global oil/commodity prices and may not be significantly vulnerable to natural disasters.

Table 3:

Estimation Results

Standard errors in parentheses

^***p<0.1

^**p<0.05,

^*p<0.01,

Table 3:

Estimation Results

VARIABLES	(1) Expenditure	(2) Revenue excl grants	(3) Expenditure	(4) Revenue excl grants	(5) Expenditure	(6) Revenue excl grants	(7) Expenditure	(8) Revenue excl grants
Expenditure (1st lag)	0.590***	-0.235**	0.591***	-0.228**	0.611***	-0.235*	0.580***	-0.236**
	(0.125)	(0.109)	(0.126)	(0.110)	(0.136)	(0.121)	(0.123)	(0.107)
Revenue excl grants (1st lag)	0.274**	1.186***	0.283**	1.163***	0.290**	1.193***	0.260**	1 187***
	(0.133)	(0.0994)	(0.137)	(0.100)	(0.146)	(0.103)	(0.127)	(0.100)
Natural Disaster Dummy	6.420**	-0.0918	6.175**	0.325	5.373*	-0.379	6.112**	-1.145
	(2.922)	(2.562)	(2.986)	(2.645)	(2.892)	(2.537)	(2.773)	(2.783)
Natural Disaster Dummy (1st lag)	6.780*	-2.396	6.353*	-1.691	5.521	-2.189	8.290*	-5.265**
	(3.731)	(1.661)	(3.731)	(1.774)	(4.245)	(1.662)	(4.232)	(2.521)
Natural Disaster Dummy (2nd lag)	7.374**	-0.673	7.028**	-0.294	6.585**	-0.265	6.558**	-0.675
	(3.199)	(2.176)	(3.156)	(2.583)	(3.094)	(1.623)	(2.931)	(1.711)
Grants	0.466***	0.280***	0.464***	0.291***	0.482***	0.306***	0.472***	0.282***
	(0.152)	(0.106)	(0.149)	(0.106)	(0.164)	(0.116)	(0.149)	(0.106)
Real GDP growth			-0.0604	0.223*
			(0.105)	(0.123)
Real GDP growth (lag)					0.0673	0.0205
					(0.121)	(0.108)
Australia RGDP growth							-0.0336	-0.0804
							(0.0388)	(0.0497)
US RGDP growth							0.0163	0.373*
							(0.265)	(0.218)
Observations	230	230	230	230	227	227	230	230
Hansen	0.728	0.728	0.735	0.735	0.899	0.899	0.646	0.646

Standard errors in parentheses

^***p<0.1

^**p<0.05,

^*p<0.01,

View Table

Table 3:

Estimation Results

VARIABLES	(1) Expenditure	(2) Revenue excl grants	(3) Expenditure	(4) Revenue excl grants	(5) Expenditure	(6) Revenue excl grants	(7) Expenditure	(8) Revenue excl grants
Expenditure (1st lag)	0.590***	-0.235**	0.591***	-0.228**	0.611***	-0.235*	0.580***	-0.236**
	(0.125)	(0.109)	(0.126)	(0.110)	(0.136)	(0.121)	(0.123)	(0.107)
Revenue excl grants (1st lag)	0.274**	1.186***	0.283**	1.163***	0.290**	1.193***	0.260**	1 187***
	(0.133)	(0.0994)	(0.137)	(0.100)	(0.146)	(0.103)	(0.127)	(0.100)
Natural Disaster Dummy	6.420**	-0.0918	6.175**	0.325	5.373*	-0.379	6.112**	-1.145
	(2.922)	(2.562)	(2.986)	(2.645)	(2.892)	(2.537)	(2.773)	(2.783)
Natural Disaster Dummy (1st lag)	6.780*	-2.396	6.353*	-1.691	5.521	-2.189	8.290*	-5.265**
	(3.731)	(1.661)	(3.731)	(1.774)	(4.245)	(1.662)	(4.232)	(2.521)
Natural Disaster Dummy (2nd lag)	7.374**	-0.673	7.028**	-0.294	6.585**	-0.265	6.558**	-0.675
	(3.199)	(2.176)	(3.156)	(2.583)	(3.094)	(1.623)	(2.931)	(1.711)
Grants	0.466***	0.280***	0.464***	0.291***	0.482***	0.306***	0.472***	0.282***
	(0.152)	(0.106)	(0.149)	(0.106)	(0.164)	(0.116)	(0.149)	(0.106)
Real GDP growth			-0.0604	0.223*
			(0.105)	(0.123)
Real GDP growth (lag)					0.0673	0.0205
					(0.121)	(0.108)
Australia RGDP growth							-0.0336	-0.0804
							(0.0388)	(0.0497)
US RGDP growth							0.0163	0.373*
							(0.265)	(0.218)
Observations	230	230	230	230	227	227	230	230
Hansen	0.728	0.728	0.735	0.735	0.899	0.899	0.646	0.646

Standard errors in parentheses

^***p<0.1

^**p<0.05,

^*p<0.01,

G. Fiscal Cost of Natural Disaster for Individual PICs

As Table 4 below shows, the cumulative fiscal cost of a severe natural disaster is estimated, on average, at 13.8 to 20.6 percent of GDP.¹¹ Because each country has a different frequency of disaster, each would face a different timeframe to build a fiscal buffer able to cover the fiscal cost of one disaster. Assuming the estimated fiscal cost for one disaster in each country is the regional average, the annual fiscal cost can be calculated as below:

Table 4.

Fiscal Cost of Severe Natural Disasters

(in percent of GDP)

Note: Average probability is calculated using the total severe disaster observations divided by the total number of years (37) in sample period 1980–2016. Based on this probability, implied average disaster interval can be calculated, and could be compared with the median disaster interval. NA indicates that median interval can’t be calculated because they only have one severe disaster observation, and to calculate an interval needs at least two severe disasters.

Table 4.

Fiscal Cost of Severe Natural Disasters

(in percent of GDP)

Total Fiscal Cost (In percent of GDP)
		Disaster year	1st yr post-disaster	2nd yr post-disaster	Cumulative fiscal cost for one disaster
Regional Level		6.4	6.8	7.4		13.8 ~ 20.6
Probability and Intervals for a Severe Disaster (by country)
	Actual Average Probability	Implied Average Interval	Actual Median Interval	Implied probability	Annual Fiscal Cost (based on median interval)	Annual Fiscal Cost (based on average interval)
Samoa	16.2	6.2	3.0	33.3	4.6 ~ 6.9	2.2 ~ 3.3
Vanuatu	18.9	5.3	4.0	25.0	3.5 ~ 5.2	2.6 ~ 3.9
Solomon Islands	8.1	12.3	5.5	18.2	2.5 ~ 3.7	1.1 ~ 1.7
Marshall Islands	8.1	12.3	12.5	8.0	1.1 ~ 1.6	1.1 ~ 1.7
Fiji	10.8	9.3	15.0	6.7	0.9 ~ 1.4	1.5 ~ 2.2
Micronesia	5.4	18.5	18.0	5.6	0.8 ~ 1.1	0.7 ~ 1.1
Tonga	5.4	18.5	19.0	5.3	0.7 ~ 1.1	0.7 ~ 1.1
Tuvalu	5.4	18.5	25.0	4.0	0.6 ~ 0.8	0.7 ~ 1.1
Kiribati	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Timor-Leste	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Papua New Guinea	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Palau	0.0	NA	NA	NA	NA ~ NA	0.0 ~ 0.0

Note: Average probability is calculated using the total severe disaster observations divided by the total number of years (37) in sample period 1980–2016. Based on this probability, implied average disaster interval can be calculated, and could be compared with the median disaster interval. NA indicates that median interval can’t be calculated because they only have one severe disaster observation, and to calculate an interval needs at least two severe disasters.

View Table

Table 4.

Fiscal Cost of Severe Natural Disasters

(in percent of GDP)

Total Fiscal Cost (In percent of GDP)
		Disaster year	1st yr post-disaster	2nd yr post-disaster	Cumulative fiscal cost for one disaster
Regional Level		6.4	6.8	7.4		13.8 ~ 20.6
Probability and Intervals for a Severe Disaster (by country)
	Actual Average Probability	Implied Average Interval	Actual Median Interval	Implied probability	Annual Fiscal Cost (based on median interval)	Annual Fiscal Cost (based on average interval)
Samoa	16.2	6.2	3.0	33.3	4.6 ~ 6.9	2.2 ~ 3.3
Vanuatu	18.9	5.3	4.0	25.0	3.5 ~ 5.2	2.6 ~ 3.9
Solomon Islands	8.1	12.3	5.5	18.2	2.5 ~ 3.7	1.1 ~ 1.7
Marshall Islands	8.1	12.3	12.5	8.0	1.1 ~ 1.6	1.1 ~ 1.7
Fiji	10.8	9.3	15.0	6.7	0.9 ~ 1.4	1.5 ~ 2.2
Micronesia	5.4	18.5	18.0	5.6	0.8 ~ 1.1	0.7 ~ 1.1
Tonga	5.4	18.5	19.0	5.3	0.7 ~ 1.1	0.7 ~ 1.1
Tuvalu	5.4	18.5	25.0	4.0	0.6 ~ 0.8	0.7 ~ 1.1
Kiribati	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Timor-Leste	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Papua New Guinea	2.7	37.0	NA	NA	NA ~ NA	0.4 ~ 0.6
Palau	0.0	NA	NA	NA	NA ~ NA	0.0 ~ 0.0

Note: Average probability is calculated using the total severe disaster observations divided by the total number of years (37) in sample period 1980–2016. Based on this probability, implied average disaster interval can be calculated, and could be compared with the median disaster interval. NA indicates that median interval can’t be calculated because they only have one severe disaster observation, and to calculate an interval needs at least two severe disasters.

The annual fiscal cost based on the regional level median disaster interval (13.8 years) is 1– 1.5 percent of GDP. We use the median interval as a more robust estimator than the sample mean, which can be strongly influenced by a few outlier observations. Applying the same approach to each PIC, we can see the results in Table 4. Samoa, Vanuatu, and Solomon Islands have the highest annual fiscal cost, reflecting the high frequency of disasters, which leaves less time to rebuild buffers than for countries hit less frequently.

It should be emphasized that these figures are indicative rather than precise. Part of this is because we use pooled data to estimate a “typical” disaster fiscal cost in the region. In addition, the intervals between disasters that we use are based on data over the whole period since 1980. We have made no allowance for the impact of climate change, which may be significantly shortening the interval between disasters. Nor has allowance been made for the fiscal costs of smaller disasters. Consequently, the fiscal costs shown in Table 4 should be probably be regarded as lower bound estimates. Moreover, because both the timing of disasters is stochastic, and the magnitude of the impact can vary greatly from one disaster to the next, this means that even if a country saved at the rates indicated in Table 4, there is no guarantee that this would be sufficient to cover all major disasters. Sometimes these savings rates would be more than sufficient, and other times not.

H. Data Limitations/Caveats

The estimation results are based on an average across countries and disasters. But each country is different. So too is each disaster, even in the same country. The damage and fiscal cost of a cyclone, for example, will tend to be less in sparsely populated countries than in densely-settled countries, and less in countries with very limited infrastructure than in more developed countries with more, and more vulnerable, infrastructure.

The estimated costs focus only on relatively severe natural disasters, so the fiscal costs of small natural disasters are excluded from our analysis.¹²

To keep the consistency of our analysis, we use information from EM-DAT as our only source of natural disaster data. For some natural disasters, EM-DAT has different damage or population affected numbers compared with numbers from other sources (e.g. World Bank). Therefore, in certain cases, the damage of a natural disaster might be under or over estimated.

No allowance is made for the impact of climate change on the frequency or severity of disasters. The estimates can therefore be seen as optimistic, although they do take into account the experience of PICs in the 1980s and 1990s, which saw more frequent large disasters than the first decade of the 2000s. Certainly the experiences of Fiji, Tonga, and Tuvalu in recent years suggest that the frequency with which they are affected by severe cyclones has increased sharply.

A. External Financing

The size of external financial support is generally determined by the degree of the damage by natural disasters. The larger the damage and loss is, the larger are likely to be the grants provided from international community. Figure 4 shows the humanitarian aid flows for the PICs that were struck by severe natural disasters in the past ten years. The international community expressed their large commitment for humanitarian aid right after natural disasters took place. It ranged from US$ 2.8 million to US$ 43.7 million, depending on the severity of disasters. Regarding insurance payment, Vanuatu and Tonga received a payout of US$ 1.9 million (about 0.25 percent of GDP) and US$ 3.5 million (about 0.8 percent of GDP), following the Tropical Cyclones Pam and Gita, respectively, under the Pacific Disaster Risk Financing and Insurance (DRFI) Program.

Pacific Island Countries : Humanitarian Aid after Severe Disasters ^1/

(In US$ millions)

Citation: IMF Working Papers 2019, 152; 10.5089/9781498320801.001.A001

^1/: Severe disasters as defined in section III.Sources: UNOCHA Financial Tracking Service.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Pacific Island Countries : Humanitarian Aid after Severe Disasters ^1/

(In US$ millions)

Citation: IMF Working Papers 2019, 152; 10.5089/9781498320801.001.A001

^1/: Severe disasters as defined in section III.Sources: UNOCHA Financial Tracking Service.

• Download Figure
• Download figure as PowerPoint slide

Pacific Island Countries : Humanitarian Aid after Severe Disasters ^1/

(In US$ millions)

Citation: IMF Working Papers 2019, 152; 10.5089/9781498320801.001.A001

^1/: Severe disasters as defined in section III.Sources: UNOCHA Financial Tracking Service.

• Download Figure
• Download figure as PowerPoint slide

International Financial Institutions (IFIs) also provide loans to alleviate balance-of-payment pressures after natural disasters and to support reconstruction activities. The World Bank (WB), Asian Development Bank (ADB), and IMF, can provide loans on concessional terms. In the case of Vanuatu, the IMF provided emergency loans (US$ 23.8 million, about 3.0 percent of GDP) in June 2015 with no interest payment, while the ADB and WB are expected to provide Vanuatu with a total of around US$40 million (about 3.8 percent of GDP) for the purpose of the reconstruction of roads, schools, and public buildings during 2017–2022, with grant elements more than 50 percent (Box 2). In the case of Fiji, the ADB and WB provided US$ 50 million emergency assistance loan, respectively, to the country for disaster recovery program (Box 3). High concessionality of loans reduces the debt service pressure in the medium term. Excessive borrowing would raise the issue of debt sustainability, especially for some PICs with high debts. The PICs with high debts would face not only debt service pressure in the medium term but also the limited capacity to borrow in the event of future natural disasters.

Box 2.

Cyclone Pam in Vanuatu

Cyclone Pam, one of the most damaging natural disasters in Vanuatu’s history, hit the country in March 2015.It caused overall damages amounting to more than 60 percent of GDP and affected almost 72 percent of the population (more than 188,000 inhabitants). Key sectors such as tourism, transportation, and agriculture were highly damaged. In its aftermath, development partners contributed significant amounts of aid-in-kind, grants, and loans. Vanuatu received around US$ 40.2 (4.6 percent of GDP) of grants from bilateral and multilateral donors and around 85 percent of the total funds was disbursed as of end -2017. Regarding loans, the IMF provided financial assistance (about US$ 23.8 million, equivalent to about 3 percent of GDP) in June 2015 under the RCF and RFI to address an urgentbalance of payments need. While the IMF’s quick disbursement catalyzed the internationalcommunity’s efforts to support Vanuatu’s recovery from Cyclone Pam, the fund has not been used due to limited BOP pressure. The IDA is expected to provide US$ 50 million (about 4.7 percent of GDP) in the form of loans (50 percent) and grants (50 percent) during 2017– 2022 to support reconstruction and improvement of roads, schools, and public buildings, while the ADB is expected to provide about US$29 million (about 3 percent of GDP) in the form of loans (50 percent) and grants (50 percent) during 2017–2019. The government itself used its own resources, amounting to VUV 401.6 million (about 0.5 percent of GDP), for the relief and recovery efforts as well as refunds for the disaster-related expenses.

Table 4.

Financial Supports from International Community for Cyclone Pam

Source: Vanuatu authorities; and IMF staff estimates.

Table 4.

Financial Supports from International Community for Cyclone Pam

	Source	Amount (US$ million)	Amount (In Percent of GDP)
Grants	International community	40.2	4.6%
	IDA for reconstruction projects	25.0	2.3%
	ADB for reconstruction projects	14.5	1.5%
	Total	79.7	8.5%
Loans	IMF Facility	23.8	3.0%
	IDA for reconstruction projects	25.0	2.3%
	ADB for reconstruction projects	14.5	1.5%
	Total	63.3	6.8%

Source: Vanuatu authorities; and IMF staff estimates.

View Table

Table 4.

Financial Supports from International Community for Cyclone Pam

	Source	Amount (US$ million)	Amount (In Percent of GDP)
Grants	International community	40.2	4.6%
	IDA for reconstruction projects	25.0	2.3%
	ADB for reconstruction projects	14.5	1.5%
	Total	79.7	8.5%
Loans	IMF Facility	23.8	3.0%
	IDA for reconstruction projects	25.0	2.3%
	ADB for reconstruction projects	14.5	1.5%
	Total	63.3	6.8%

Source: Vanuatu authorities; and IMF staff estimates.

We can roughly calculate the total of humanitarian aid and loans as a percent of GDP by adding humanitarian aid in the disaster year and loan disbursement from disaster year onwards. The total external support the PICs received after the disasters is expected to range from 1.7 to 18.5 percent of GDP, depending on the type and severity of natural disasters. This is a very wide range and makes it difficult for policy makers to gauge how much assistance they should count on in the wake of a severe disaster. Moreover, the bulk of such assistance may go directly to the private sector, so that only a small part may be available for budget support. In such circumstances, PIC governments should probably make quite conservative assumptions regarding post-disaster external budget support.

B. Domestic Financing

Given relatively small financial markets in PICs, the scope for governments to borrow domestically is very limited and the cost of borrowing is generally high. Moreover, domestic borrowing by the government risks crowding out private sector borrowing to finance recovery.

In Vanuatu the government did not borrow from domestic sources after Cyclone Pam for the purpose of recovery and reconstruction efforts. In Fiji, domestic borrowing increased partly to fund recovery efforts. In FY16-17, infrastructure bonds amounting to almost 2 percent of GDP were sold, largely to the Fiji National Provident Fund (FNPF), commercial banks, and insurance companies. Another potential source of domestic financing is the central bank, but in general central bank financing should be strictly limited on account of risks of fiscal dominance that would destabilize the economy through possible higher inflation.

National Provident Funds (NPFs) in PICs would be a possible source for financing rebuilding, though more likely for the private sector than for the government (Box 3). Some PICs have mobilized pension funds, but as a disaster relief instrument instead of as a government financing source. For example, Fiji allowed the members of its FNPF to withdraw up to 30 percent of their pension account balance after Cyclone Winston. Vanuatu’s NPF allowed its members to withdraw up to 20 percent of their retirement savings, with total drawdowns amounting to VUV 1.7 million (2.1 percent of GDP) for three months after Cyclone Pam. Si et al. (2018) suggested that early pension withdrawal could improve welfare by helping households finance necessary expenditure when they recover from the damage caused by a natural disaster. Without any subsequent measures to strengthen the financial position of NPFs, such as adjusting the mandatory pension contribution rate, a contingent government liability for under-saving for retirement would likely materialize.

Box 3.

Post-Cyclone Winston Financing in Fiji

In early 2016, Fiji was struck by Cyclone Winston (TC Winston), one the most severe tropical cyclones in the Southern Hemisphere. The cyclone caused US$ 1.42 million (about 31 percent of GDP) damage and losses and affected more than 60 percent of population ¹. In the Disaster Recovery Framework (DRF)² published by the Fijian government responding to the TC Winston, the government dedicated the recovery efforts on restoring public assets and services, providing assistance and support to affected populations, and increasing resilience to future disasters. The DRF identified recovery priorities and estimated the costs of recovery at F$ 730 million for the period covering fiscal year 2016 to 2018. The government was able to allocate F$ 136 million from its own sources and received F$ 22 million from donors, leaving a financing gap of F$ 575 million. Recognizing the long-term nature of recovery and reconstruction, the Fijian government has integrated recovery efforts beyond the first two years into its National Development Plan.

Post-disaster financing after TC Winston includes humanitarian aid from UN agencies, bilateral donors, and private non-profit organizations and external loans from multilateral organizations like World Bank and ADB. In total, Fiji received about US$ 42.1 million humanitarian aid³, of which US$ 21.5 million was from UN Flash Appeal, making TC Winston the best funded emergency response in year 2016. The major bilateral donors were Australia (US$ 27 million), New Zealand (US$ 10.3 million) and the EU (US$ 4.8 million), and about 75 percent recovery support is in-kind⁴. At the same time, Fiji signed US$ 50 million loan agreements with both World Bank and the Asian Development Bank, bringing the total loan amount to US$ 100 million. These loans are mainly used for post-disaster schoolreconstruction and Help for Homes program. As part of the disaster relief efforts, Fiji National Provident Fund (FNPF) allowed members to withdraw up to F$ 1000 (US$ 465) plus additional F$ 5,000 (US$ 2,325) if proof of damaged housing was provided. Within two months of the disaster, FNPF approved 170,000 withdraw application and disbursed about F$ 250 million cash to affected population.

Projected Costs of Recovery from TC Winston, in F$ millions

Sources: Disaster Recovery Framework by MOE.

Projected Costs of Recovery from TC Winston, in F$ millions

Recover Priority	Total Budget	Government	Donors	Financing Gap
Priority 1: Rebuilding homes	183.94	72.14	0	111.8
Priority 2: Restoring Livelihoods	169.65	36.17	9.89	123.7
Priority 3: Repairing and Strengthening Critical Infrastructure	353.39	25.79	12.04	315.57
Priority 4: Building Resilience	23.88	2	0	23.88
Total	730.87	135.99	21.93	574.95
In percent of GDP	7.41	1.38	0.22	5.83

Sources: Disaster Recovery Framework by MOE.

View Table

Projected Costs of Recovery from TC Winston, in F$ millions

Recover Priority	Total Budget	Government	Donors	Financing Gap
Priority 1: Rebuilding homes	183.94	72.14	0	111.8
Priority 2: Restoring Livelihoods	169.65	36.17	9.89	123.7
Priority 3: Repairing and Strengthening Critical Infrastructure	353.39	25.79	12.04	315.57
Priority 4: Building Resilience	23.88	2	0	23.88
Total	730.87	135.99	21.93	574.95
In percent of GDP	7.41	1.38	0.22	5.83

Sources: Disaster Recovery Framework by MOE.

¹Fiji: Post Disaster Needs Assessment on Tropical Cyclone Winston, February 20, 2016, Government of Fiji. May 2016²Disaster Recovery Framework: Tropical Cyclone Winston, Ministry of Economy, Republic of Fiji. September 2016³UNOCHA Financial Tracking Service Website, retrieved in October 2018.⁴Post-Cyclone Winston Emergency Development Policy Operation, World Bank, June 2016.

The need to ensure that the domestic financial system is able to support private sector post-disaster recovery efforts suggests that, in general, governments should not plan to draw heavily on domestic borrowing as a significant part of public sector post-disaster financing.

C. Contingency Credit / Provision and Budget Reallocation

A contingent provision in the annual budget allows the government to fund unforeseen expenditure, including from natural disaster shocks. Some PICs, such as Fiji, Solomon Islands and Vanuatu, have a contingency provision in place in their annual budget (Box 4). Solomon Islands puts aside SI$ 26 million (about 0.2 percent of GDP and 0.7 percent of total expenditure), while Vanuatu reserves VUV 50 million (about 0.05 percent of GDP and 0.2 percent of total expenditure, excluding grants-funded expenditure) in the 2018 budget. Clearly these contingency funds are very limited, especially relative to the costs associated with natural disasters, and cannot realistically be considered as a significant source of disaster financing.

The government may use its own resources to cope with natural disasters through budget reallocation in a supplementary budget. Vanuatu has spent VUV 402 million (about 0.5 percent of GDP) for Cyclone Pam. Budget reallocation allows for spending for immediate relief needs, while awaiting legislative process of appropriating funds for relief and reconstruction. However, substantial budget reallocation from budgeted spending earmarked for other purposes to recovery-related expenditure may complicate budget execution and fiscal policy. In Fiji, following Cyclone Winston, significant budgetary reallocation also took place. Budget reallocation is, in effect, a means of limiting additional spending rather than a source of additional financing. Moreover, although it is natural for governments to reallocate funds following a major disaster, some reallocations may be very disruptive or costly. For example, if some investment projects are halted to divert funds to disaster relief, it may be very costly to later re-start the projects.