Policy Research Working Paper 4988
Disasters and Economic Welfare
Can National Savings Help Explain Post-disaster Changes
in Consumption?
Reinhard Mechler
The World Bank
Sustainable Development Network
Global Facility for Disaster Reduction and Recovery Unit
July 2009
WPS4988
Produced by the Research Support Team
Abstract
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the
names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Policy Research Working Paper 4988
The debate on whether natural disasters cause significant
macroeconomic impacts and indeed hinder development
is ongoing. Most analyses along these lines have focused
on impacts on gross domestic product. This paper looks
beyond this standard national accounting aggregate, and
examines whether traditional and alternative national
savings measures combined with adjustments for the
destruction of capital stocks may contribute to better
explaining post-disaster changes in welfare as measured
by changes in consumption expenditure. The author
concludes that including disaster asset losses may help to
better explain variations in post-disaster consumption,
albeit almost exclusively for the group of low-income
This paper—a product of the Global Facility for Disaster Reduction and Recovery Unit, Sustainable Development
Network—is part of a larger effort in Network to disseminate the emerging findings of the forthcoming joint World
Bank-United Nations’ Assessment of the Economics of Disaster Risk Reduction. Policy Research Working Papers are also
posted on the Web at http://econ.worldbank.org. The author may be contacted at mechler@iiasa.ac.at.
countries. The observed effect is rather small and in the
range of a few percent of the explained variation. For
low-income countries, capital stock and changes therein,
such as forced by disaster shocks, seem to play a more
important role than for higher-income economies, where
human capital and technological progress become crucial.
There are important data constraints and uncertainties,
particularly regarding the quality of disaster loss data
and the shares of capital stock losses therein. Another
important challenge potentially biasing the results is the
lack of data on alternative savings measures for many
disaster-exposed lower-income countries and small island
states.
Disasters and Economic Welfare:
Can National Savings Help Explain Post-disaster Changes
in Consumption?
Reinhard Mechler1
International Institute for Applied Systems Analysis
(IIASA)
JEL: E01, Q54, Q56
Keywords: Natural disasters, macroeconomic consequences, welfare effects, genuine
savings.
1 Support is gratefully acknowledged by the World Bank-UN project “Economics of Disaster Risk
Reduction.” We would like to thank the team leader of this project, Apurva Sanghi, and Sebnem Sahin of
the World Bank for ongoing support and stimulating discussions, as well as Norman V. Loayza, Kirk
Hamilton, Alejandro Fuente, Jesus Crespo Cuaresma, Stefan Hochrainer and two anonymous referees for
extremely helpful and stimulating comments.
2
1 POINT OF DEPARTURE
There is an ongoing debate on whether disasters cause significant macroeconomic
impacts and are truly a potential impediment to development. A position backed by
anecdotal evidence and a fair number of studies holds that natural disasters can set back
economic development (Otero and Marti, 1995; Benson, 1997a,b,c; Benson and Clay,
1998, 2000, 2001; ECLAC, 1999, 2002; Murlidharan and Shah, 2001; Crowards, 2000;
Charveriat, 2000; Mechler, 2004; Hochrainer, 2006; Cuaresma et al., 2008; Noy, 2009).
Then, there is a position suggesting that disasters have no effects on economic growth
(Albala-Bertrand, 1993, 2006; Skidmore and Toya, 2002; Caselli and Malhotra, 2004).
Most analyses along these lines have focused on aggregate impacts, and here on GDP as
the standard economic indicator for measuring changes in economic welfare. There is
almost no work on other indicators of welfare, such as consumption, which in economics
is usually taken as the basis for assessing changes in individual utility and social welfare.
Furthermore, it is well known that GDP or GNI2 are imperfect metrics for measuring
changes in welfare, as those aggregates generally do not account for the depletion of
natural resources, the value of household labor or investments in education. Several
alternative concepts have been proposed, an important one being genuine savings,3
In the context of natural disaster risk, an additional problem arises due to the fact that
the destruction of assets (capital stocks) is not considered in national accounting (which
essentially measures flows only), while the flow variables reconstruction and relief
spending add positively to GDP, yet in fact only contribute to a recovery to a prior
which
is an alternative welfare indicator based on concepts of green accounting (see, e.g.,
Hamilton and Atkinson, 2006). Genuine savings aims at better measuring the “true”
national savings by adding investments in human capital and subtracting the consumption
of capital stock, the depletion of natural resources and the adverse effects of air pollution.
The validity of savings measures is commonly tested by studying their ability to explain
variations in consumption changes. Although fraught with measurement problems,
genuine savings has gained acceptance and found applications in research and policy. It is
also standardly reported in the World Bank World Development Indicators.
2 In the following, we use GNI, as savings measures in the WDI dataset are indicated as a share of this
aggregate.
3 In the WDI genuine savings are referred to as adjusted savings.
3
economic status quo. Thus, relief and reconstruction spending in fact have to be
considered as a kind of “defensive spending,” and consequently disaster losses may need
to be adjusted for in national accounting statistics.
Given that disasters deplete capital stock and can be important in many countries, in
this paper we examine and test whether disaster losses should also be appropriately
considered in genuine savings and other savings measures.4
4 There may also be an anticipatory effect as individuals adjust their savings and consumption decisions
before or without an event, and thus baseline savings already incorporate part of the response to an
event. Yet, generally, the literature finds people to be myopic faced with rare events such as disasters,
and thus this effect may mostly be important in areas with frequent events, which are not considered in
this dataset examining the largest 200 plus events over the last three decades.
Almost no work has been
done on this issue. One paper, Barrito (2008), mentions this potential problem and
suggests a way for revising wealth accounting, yet does not empirically test it.
Using a sample of large disaster losses over the 30 year period from 1971 to 2000, we
examine whether factoring in such disaster shocks may help to better explain future
variations in welfare as measured by private and public consumption expenditure. Overall
we find some, albeit small and limited, evidence for adverse consequences of disasters on
consumption. Focusing on alternative measures of welfare and assessing the contribution
of disaster–related asset losses to changes in consumption, we conclude that accounting
for disaster asset losses for disaster-exposed, low-income countries may help to better
explain variations in post-disaster consumption and thus adjusting alternative welfare
indicators (negatively) can lead to improved predictability of future post–disaster
consumption changes.
The paper is organized as follows. We start in section 2 with introducing the concept
of national savings and a proposal for deriving alternative savings measures. We further
suggest considering disaster losses in savings measures and use a method to test the
suitability of doing so. In section 3, we present results based on cross-country regressions
for a sample of large-scale disasters. Section 4 ends with a discussion of these findings
and implications of the analysis.
4
2 USING AND APPLYING SAVINGS TO EXPLAIN CHANGES IN CONSUMPTION
In standard national accounting, gross national savings is calculated as the residual of
income and consumption. Gross savings is the amount of annual gross income that is not
consumed, and thus can be used for investment finally which adds to national wealth.
Positive savings indicate an increment to overall national wealth. Standardly, national
accounting only measures the increments to produced capital (or capital stock consisting
of machinery, equipment, physical structures including infrastructure, and urban land
area), yet social capital (human capital, quality of institutions, and the governance of
goods and people) and environmental capital (land, forests and sub-soil resources) are not
considered. Alternative savings measures have been proposed in order to also factor in
investments in those capital classes, a key concept being that of genuine savings. Genuine
savings can be derived from gross national savings, which is standardly reported in
national accounting statistics, and four types of adjustments can be distinguished as
suggested in World Bank (2006) as follows (see also figure 1):
(1) The depreciation of fixed capital representing the consumption of capital is deducted
from gross savings leading to net savings.
(2) Current education expenditures representing investments in human capital
(standardly, in national accounting these expenditures are considered a consumption
item)5
(3) The depletion of natural resources is factored in reflecting the decrease in the natural
asset base due to the extraction and harvesting of resources leading to genuine
savings excluding air and climate change damages.
are added in order to obtain net savings plus education expenditure.
(4) In a final step, social costs due to air pollution and climate change may be subtracted
leading to an estimate of genuine savings including air and climate change damages.
5 Although there is some discussion in the literature, which holds on the contrary that educational
expenditure is closer to consumption than investment.
5
.
Fig. 1: Calculating genuine savings
Source: World Bank, 2006
Testing the explanatory power of savings
In order to test whether different savings concepts may indeed lead to an improved
explanation of welfare changes, savings may be linked to consumption (see Dasgupta,
2001; Hamilton and Hartwick, 2005). In a competitive economy savings S (the increment
to capital or wealth) in a given year to should equal the present value of changes in
consumption ever after to, i.e., the future additional consumption produced thanks to the
wealth increment. In equation form this can be expressed as follows:
0
0
t
T
t t 1 t-1
t-1
t
t
t ) S
N
- C
N
(C
(1 r)
1 =
+
?
= +
with S, the savings measure, C consumption, N population, r discount rate, t time, and T
end of the time horizon considered. Accordingly, this relationship may be tested
empirically in a linear relationship as follows:
PV ?Ct0 +1,T = ?0 + ?1St0 +?? t0
6
with PV?Cto +1,T the present value of the change in per capita consumption from year
to+1 to T, and S the respective savings measure in t0, and ?0,1 the coefficients and error
term ?. In order to account for demographic change, tests should derive per capita
estimates. 6
6 Ferreira et al. (2008) conduct a more in depth assessment of the savings measure accounting for
population dynamics and omitted wealth. Yet, they find their results to only marginally improve, so we
do not consider these additional factors in the following.
Disaster adjusting savings
It seems intuitive to think about including disaster losses in genuine and others savings
measures. Overall, there may be two (interlinked) channels through which disasters and
associated losses may impact on future consumption: (1) by directly destroying capital
stock, output and income is decreased, depressing future consumption, which is generally
a function of income, (2) due to the need for rebuilding assets and livelihoods, planned
consumption is foregone in favor of reinvestments. Both channels seem important, yet
here we focus mostly on channel (1) and test whether capital accumulation and associated
future consumption opportunities are affected.
Disasters have the potential to cause substantial direct and indirect losses and destroy a
large portion of produced capital. Over the last 30 years, there have been about 50
instances where losses exceeded 10% of gross national product (GNP), another 50 events
where losses ranged from 5-10% of GNP, and about 110 events with losses exceeding
1% GNP. It is important to note that most of the very large events affected Small Island
States, or smaller, lower-income countries - countries small enough to have their entire
territory affected by one event, or countries too economically limited as to be able to
absorb the losses as very roughly proxied by GDP (such as observed for St. Lucia by
Hurricane Gilbert in 1988) (see figure 2).
7
0%
50%
100%
150%
200%
250%
300%
350%
400%
St. Lucia 1988
Samoa 1991
Samoa 1990
Mongolia 1996
Vanuatu 1985
St. Kitts and Nevis1998
Vanuatu 1987
Dominica 1979
Nicaragua 1972
St. Kitts and Nevis1995
Ant igua and Barbuda1995
Dominica 1995
Honduras 1998
St. Lucia1980
Honduras 1974
El Salvador 1986
Dominican Republic1979
Tonga 1982
Belize 2000
St. Kitts and Nevis1989
Samoa 1983
Samoa 1983
Zimbabwe 1981
Nicaragua 1998
Haiti 1980
Georgia1991
Jamaica 1988
Guatemala 1976
Nepal 1987
Lao PDR1993
Loss per GDP
Fig. 2: 30 largest monetary disaster losses since 1970
Source: own calculations based on data by EMDAT (CRED, 2009); Munich Re, 2008.
Although only a part of these losses are in fact capital stock losses, it seems evident that
losing a substantial portion of produced assets will impact the capital accumulation
process, affect produced wealth and as a consequence impair income creation. Barrito
(2008) suggests that, if factoring in reported disaster losses as the stock losses, capital
accumulation may fall significantly and permanently short of regularly reported
increments to capital (see Figure 3). Consequently, capital accumulation in vulnerable
countries such as El Salvador, Fiji and St. Lucia may be strongly affected by one or
multiple events whereas in large and diversified economies no significant effect may be
identified.
Given a lower capital accumulation path, it is straightforward to expect adverse
welfare effects, such as changes in consumption and consumption volatility over time.
The key question we pursue in this paper is to assess whether natural disaster losses can
be considered to affect consumption and whether including them in savings measures
may help to improve the predictive power of savings constructs. Our entry point is to
adjust for disasters by adding the disaster related depreciation in terms of losses of capital
stock to the other regular depreciation of capital.
8
Fig. 3: Capital accumulation and disaster-related capital depletion
Note: In millions of constant 2000 USD; NKF = Net Capital Formation; NKF’ = Net Capital Formation adjusted by
disaster losses. GKF=Gross Capital Formation; DpK = Depreciation of capital; EL (all) = Monetary disaster losses.
Source: Barrito, 2008.
A question is which savings measure to choose. A recent study by Ferreira, Hamilton,
and Vincent (2008) using panel data for 64 developing countries during the period 1970–
82 for different savings adjustments finds that the key step in such analysis is to account
for the depletion of natural resources, which leads to a significant improvement in the
relationship of savings and changes in consumption. The authors contend that genuine
savings are most meaningful if adjustments have been made for natural resource
9
depletion.7
Data and estimation procedure
We use this savings measure and further adjust it by subtracting disaster asset
losses in the given event, and in order to compare also examine gross and net savings,
overall leading to three disaster adjusted savings indicators: (i) gross disaster savings, (ii)
net disaster savings, and (iii) genuine disaster savings. We compare these constructs to
the savings measures unadjusted for disaster losses. The intuition behind the adjustment
is to include the adjustment of capital stock losses resulting from an exogenous disaster
shock. In line with World Bank (2006), we conduct standard bivariate regression and do
not simultaneously account for other explanatory variables beyond savings measures.
Yet, we compose subsamples, such as differentiated by country income groups.
The analysis is based on observed and calculated savings and the present value of
consumption per capita from the disaster year into the future up to 2005, the last year
with rather complete information (in constant 2000 USD values). As savings are likely to
be affected in the event year, as discussed above, we use reported savings in the year
before the event and adjust this for the disaster asset loss, then start with calculating the
consumption change in the year of the event up to 2005, the last year of our time horizon.
Savings and consumption are reported as per cent of GNI of the year before the event. In
fact, this approach is equivalent to assuming that the event happened right at the start of
the disaster year.8 The global disaster sample initially consisted of 168 large natural
disaster events during the period 1971-2005, for which about 7,900 events are recorded
overall. The sample is based on information from two databases and was compiled by
Okuyama (2009) with the threshold for a large event defined arbitrarily by a loss
exceeding 1 percent of GDP.9
7 Pollution damage in terms of health effects of air pollution may be factored in a next step as well, yet such
inclusion does not add to our discussion here, and thus is omitted here, also as there are generally even
more data constraints.
8 This holds mostly for sudden onset disasters which are instantaneous events occurring over minutes and
hours, whereas droughts have a lead in of weeks to months.
9 In order to define the “event set” the threshold of stock losses is set as a share (1%) of flow effects (GDP).
It would have been more systematic to define an asset based threshold, yet we responded to the larger
intuitive appeal of using GDP as a denominator, and the fact that this threshold was also used by other
papers in the EDRR working paper series, which we wanted to be in line with.
One database is the open-source EMDAT database
(CRED, 2009) maintained by the Centre for Research on the Epidemiology of Disasters
at the Université Catholique de Louvain. EMDAT currently lists information on people
10
killed, made homeless, affected and financial losses for more than 16,000 sudden-onset
(such as floods, storms, earthquakes) and slow-onset (drought) events from 1900 to the
present. Data are generally collected from various sources, including UN agencies, nongovernmental
organizations, insurance companies, research institutes and press agencies.
The other database is the proprietary Munich Re NatCat Service database, which mainly
serves to inform insurance and reinsurance pricing. This database contains fewer entries
focusing on the about 300 largest events since 1950, yet data exhibit a higher reliability
as they are often crosschecked with other information. We focus on the monetary losses
listed in constant 2000 USD terms. In both datasets, loss data follow no uniform
definition and are collected for different purposes such as assessing donor needs for relief
and reconstruction, assessing potential impacts on economic aggregates and defining
insurance losses (see Provention, 2002). The sample comprises of sudden and slow onset
events. Key sudden-onset events are extreme geophysical events (earthquakes) and
hydrometeorological events such as tropical cyclones, floods and winterstorms. Slowonset
hydrometeorological disasters are either of a periodically recurrent or permanent
nature; these are mostly droughts, extreme temperature events and forest fires. Table 1
classifies the events in our sample according to cause and type of event.
Table 1: Classification of events in sample
Hazard Cause Type
Flood Hydrometeorological Sudden onset
Storm Hydrometeorological Sudden onset
Mass movement wet Hydrometeorological Sudden onset
Drought Hydrometeorological Slow onset
Extreme temperature Hydrometeorological Slow onset
Wildfire Hydrometeorological Slow onset
Earthquake Geophysical Sudden onset
Loss data may refer to direct and indirect loss, or stocks and flows, in unknown
proportions. As for our analysis only the direct stock losses are of importance, we resort
to assumptions and refer to evidence on the share of the direct capital stock losses in
productive sectors and infrastructure in different events in Latin America, for which good
information was at hand. Based on information listed on Table 2 and Table I-2 in the
Appendix for which a simple average would amount to 33%, given uncertainty around
11
this parameter, we finally take the simplifying assumption of using a share of 35% stock
losses in reported total losses.10
Table 2: Portion of total loss considered as direct, capital stock loss based on country cases
Event
Capital stock losses (productive sector and
infrastructure) as a share of total loss
Hurricane Stan in El Salvador, 2005 36%
Hurricane Stan in Guatemala, 2005 30%
Hurricane Mitch in Honduras 1998 48%
Hurricane Mitch in Nicaragua 1998 29%
Earthquake in El Salvador, 2001 21%
Arithmetic Average 33%
Parameter used in this study 35%
Sources: ECLAC 1999, 2002; Telford et al., 2004; CRED, 2009
For the savings measures and socio-economic information, we use World Bank
Development Indicators (World Bank, 2009) for calculating consumption, gross, net and
genuine savings and GNI per capita.11 The discount rate used for discounting future
consumption was 5% for all series (in line with World Bank, 2006). 12
Table 3: Overview over data used
Variable Data source Time horizon
Disaster losses (USD) EMDAT, Munich Re 1971-2000
Consumption (USD) WDI 2009 1970-2005
Gross National Income, GNI (USD) WDI 2009 1970-2005
Population WDI 2009 1970-2005
Gross savings (% of GNI) WDI 2009 1970-2005
Net savings (% of GNI) WDI 2009 1970-2005
Genuine savings (% of GNI) WDI 2009 1970-2005
Aid (% GNI) WDI 2009 1970-2005
GDP deflator13 WDI 2009 1970-2005
Country income groups WDI 2009 2005
10 Varying this parameter within reasonable bounds did not importantly affect the analysis.
11 Note, those data were indicated as a share of GNI, not GDP, so the analysis is based on GNI.
12 Conducting sensitivity analysis with discount rates related to country income groups, we find the
discount rate to be relatively insensitive to marginal changes, yet, an extension of this work might
explore using country-specific discount rates.
13 In the absence of a GNI deflator, we used the GDP deflator.
12
We restrict our analysis to events as late as the year 2000, so that a minimum of 5 years
of consumption data can still be used. As discussed in Ferreira, Hamilton, and Vincent
(2008), green accounting theory refers to an infinite time horizon, and it has been shown
that results become more valid with a longer time horizon; other studies have used a
minimum of 10 years, with a preference for 20 years. In contrast, we use a minimum of 5
years, yet in order to maintain a large number of observations, we keep the time horizon
flexible from a minimum of 5 to a maximum of 33 years of consumption changes
observed. We do find this flexibility in terms of time horizon to adversely affect the
analysis, which we discuss in the following. Given the data and adjustments done and a
lack of net and genuine savings data for a number of countries, only 99 observations
remained. Importantly, many of the low income countries with massive losses (such as
the biggest event of St. Lucia in 1988) dropped out due to a lack of savings or
consumption data, which is a constraint of the analysis to keep in mind. Table 4
quantitatively describes the key variables used for the analysis.
Table 4: Descriptive statistics for the whole sample (N=99)
Minimum Maximum Mean Std. Deviation
Time horizon (years) 5 33 15.1 8.18
Loss -0.57 -0.01 -0.04 0.08
Gross savings -0.04 0.56 0.17 0.12
Gross disaster savings -0.25 0.54 0.13 0.14
Net savings -0.14 0.45 0.08 0.11
Net disaster savings -0.33 0.44 0.04 0.13
Genuine savings -0.32 0.38 0.07 0.12
Genuine disaster
savings -0.42 0.37 0.02 0.14
PV ?Consumption -0.44 1.04 0.18 0.24
Losses exhibit a wide range from 1% to 57% of GNI per capita of the year of the event.
The means of the (non-disaster adjusted) savings variables seem broadly of similar
magnitude as the present value of consumption changes with means of 0.17, 0.08 and
0.07 of GNI as compared to the average consumption change of 0.18% of GNI per capita.
The “disaster” adjustment to the savings indicators reduces the means and increases
variability as measured by the standard deviation. To provide an idea for the distribution
behind the summary statistics reported in Table 4, Figure 4 shows gross savings and
gross disaster savings across the 99 cases. We observe that in the majority of cases
13
adjustments are small to moderate with a few very large events leading to severe disasterrelated
dissavings.
-50%
-40%
-30%
-20%
-10%
0%
10%
20%
30%
40%
50%
60% Gross savings
Gross disaster savings
Fig. 4: Gross and gross disaster savings for the whole sample (as a share of GNI in the year
before the event)
A key issue to consider is the effect of aid (and aid volatility) on consumption. The
question whether aid leads to higher investments (and thus to higher consumption in the
future), or is simply consumed, is of course at the heart of the development discourse and
there is no overall consensus (see, e.g. Arellano et al, 2009). This issue seems particularly
important for our analysis, as generally in large scale disasters additional aid in terms of
relief and reconstruction assistance is received. For example, in earlier work (Freeman et
al., 2002), based on a regression analysis of large-scale disaster events, we find that about
10% of losses in larger events will be compensated by relief and reconstruction
assistance. In order to revise for this “muddying” effect of international aid, particularly
for the case of lower-income countries, we also calculate scenarios, where we subtract aid
from consumption. In order to generally illustrate the calculation procedure, below we
outline the case of Honduras, which experienced four large events (1974, 1982, 1990 and
1998) over the time horizon of our study.
14
-800
-600
-400
-200
0
200
400
1971
1972
1973
1974
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Constant USD 2000
Total loss
Capital stock loss
Gross savings
Gross disasters savings
change in consumption less aid
change in consumption incl. aid
1974-2005
1982-2005
1990-2005
1998-2005
Fig. 5: Disaster losses and changes in consumption in Honduras with and without aid
Note: Values in 2000 constant USD per capita.
In Honduras, we observe multiple events leading to total and capital stock losses; beyond
the ones we look at in our sample with the 1% threshold, there are another 35 events of
smaller magnitude over the time horizon studied listed in the EMDAT database. Disasters
seem to have led to decreases in consumption spending in the year of and following
events (also depending on whether catastrophes happened early or late in a given year).
Aid seemingly has had a smoothing effect, and consumption generally exhibited some
volatility due to other reasons (for example, the hyperinflation in the 1990s). The chart
also shows the varying time horizons adopted in this study for examining effects on
consumption changes over time.
3 FINDINGS
For our sample of 99 events, we start out with assessing whether asset losses can be said
to affect the present value of post-disaster consumption changes. As shown on Table 5
for the whole sample and for two further samples for which we report results further
below, all hydrometeorological events and hydrometeorological events in the low-income
group of countries, the loss is highly insignificant (this is also the case for all further
regressions undertaken), and the (nonstandardized) coefficient is positive, which is
15
counterintuitive, yet may be explained by the fact that there are many other perturbations
affecting consumption positively as well as negatively, and thus the loss alone might have
little effect on future consumption.
Table 5: Regression losses on consumption changes for different samples
Model Whole sample
(N=99)
Sudden hydrometeorological
events (N=62)
Sudden hydrometeorological
events, low income group
(N=35)
PV dConsumption in per cent of per capita income of the year before the event
Constant 0.188*** 0.175*** 0.152**
(7.028) (4.582) (2.526)
Loss 0.206 0.083 0.345
(0.677) (0.150) (0.365)
R Square 0.005 0.001 0.004
Note: Significance at the * 10% level;** 5% level; *** 1% level
Focusing on the savings measures in a next step, we first assess how the savings
measures explain consumption changes irrespective of disaster adjustments (table 6).
Table 6: Regression results for the whole sample (N=99)
Model Gross savings Gross disaster
savings
Net savings Net disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant 0.061 0.104*** 0.124*** 0.155***
0.137*** 0.166***
(1.59) (3.42) (4.40) (6.54) (5.17) (7.17)
Savings
measures 0.692*** 0.585*** 0.654*** 0.560***
0.621*** 0.518***
(3.78) (3.67) (3.28) (3.25) (3.04) (2.99)
R Square 0.129 0.122 0.100 0.098 0.087 0.085
Note: Significance at the * 10% level;** 5% level; *** 1% level
The savings variables are all significant and most of the constants similarly so. Further,
for the whole sample, the size of these values is in line with findings from other studies
with about 13%, 10%, and 9% of the consumption change explained by gross, net and
genuine savings measures respectively in the baseline year. To provide some perspective,
Hamilton and colleagues in World Bank (2006) find that gross savings and genuine
savings explain about 15% resp. 24% of the variation in consumption using consecutive
20-year periods. Coefficients are of the right order of magnitude (ideally they should be
1) ranging from 0.7 to 0.6 compared to a range of 0.4 to 1.3 reported in World Bank
16
(2006). What is not in line with the literature is the fact that the explanatory power
decreases when going from gross to net to genuine savings. This can be explained by the
flexible time horizon chosen out of necessity to keep the number of observations as large
as possible. For example, as shown on table 7, when adopting a fixed 15 year time
horizon, thus “losing” 56 observations, the R square actually doubles for genuine savings.
Also, constants become unimportant and genuine savings very significant at the 1% level.
We suggest to keep this limitation in mind, yet in order to assess interesting subsamples
in the following, we propose to continue working with the variable time horizon given
our small dataset of 99 observations only. Accordingly, the analysis should not be
understood as shedding more light on the debate whether genuine savings better explain
consumption changes, but rather whether disaster “depreciation” helps improve the
explanatory power of savings measures generally.
Table 7: Regression results for the fixed time horizon of 15 years (N=43)
Model Gross
savings
Gross
disaster
savings
Net savings Net disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant 0.001 0.027 0.035 0.055** 0.044* 0.070***
(0.029) (0.957) (1.328) (2.315) (1.931) (2.998)
Savings
measure 0.437* 0.344** 0.470* 0.374* 0.670*** 0.556***
(1.948) (1.807) (1.836) (1.732) (2.897) (2.717)
R Square 0.085 0.074 0.076 0.068 0.170 0.153
Note: Significance at the * 10% level;** 5% level; *** 1% level
Yet, overall, deducting disasters does not improve the explained variation, and R squares
actually decrease slightly for both specifications and all savings measures, so, for the
sample looking at all hazard types and income classes we do not find disaster
depreciation to better explain regressions.
We now further test different subsamples, such as for sudden, slow onset and sudden
hydrometeorological events separately. As can be seen on Table II-1 in the Appendix, the
sudden onset group had similar explanatory power in terms of R squares as the whole
dataset, while for genuine savings the R square measure as indicator of the explained
variation now actually slightly improved when introducing disaster losses. Then, as
shown on Table II-2 for the slow-onset events (while probably too small for robust results
17
with only 21 observations), all variables become insignificant indicating that indeed slow
onset events may largely lead to indirect, flow losses rather than to direct, stock impacts
to be explained by savings measures. As a next sample, sudden–onset,
hydrometeorological events (storms and floods) are examined leading to the strongest
results in terms of R squares (from 0.33 to 0.19), while also the disaster adjustment
decreases the quality of the regression. A factor explaining this difference in results is
clearly that most of the earthquakes (13 out of 15) in the sample occurred in high and
medium income countries, while many low income countries in the sample are prone to
massive flooding and storms (hurricanes). Thus it seems to be income, further discussed
below, which picks up most of the explained variation. We do not feel confident going
beyond this in trying to explain the variation by the types of sudden onset events.
Overall, we find for the sample undifferentiated by per capita income that revising
savings for disaster shocks does not reliably improve regression results in terms of better
explaining post-disaster consumption variation.
Low and middle-income sample
As a next step, we further divide the sample into country groups differentiated by per
capita income with the expectation that the explanatory power of savings may increase as
we zoom into the group of medium to low-income countries, where capital stock should
become more important.14,15
14 We did not separately examine the middle and high income samples due to the limited sample sizes.
15 Classification is also dynamic, as countries may change their income status, yet none of the countries
analyzed in this group did leave the low to middle income group.
Generally, work on genuine savings has contended that in
developing countries, produced capital due to its sheer scarcity is a more important
component of wealth than in higher-income countries, where human capital seems more
critical (see, e.g., Ferreira et al., 2008). Also, as discussed above, for this group of
developing countries, aid inflows play an important role, and we further test the effect of
subtracting normal and disaster-related aid inflows from consumption. Table 8 shows that
all variables remain significant at the 1% level, while R squares substantially increase to,
e.g., 0.33 for gross savings. Also, the size of the coefficients in four instances increases
above 1. Yet, adding in capital stock losses does not help with explaining consumption
18
changes and actually diminish the quality of the regression (e.g., from 33% to 28%
explained variation for gross savings).
Table 8: Regression results for the low and middle income sample, sudden hydrometeorological
events (N=62)
Model Gross savings Gross
disaster
savings
Net
savings
Net
disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant -0.031 0.041 0.061* 0.120 *** 0.075 ** 0.124 ***
0.67 (1.07) (1.79) (4.03) (2.01) (3.95)
Savings measures 1.214 *** 1.014 *** 1.229 *** 0.995 *** 1.109 *** 0.974 ***
(5.33) (4.74) (5.05) (4.42) (3.96) (3.71)
R Square 0.325 0.276 0.302 0.249 0.210 0.189
Note: Significance at the * 10% level;** 5% level; *** 1% level
We also analyze the effect of revising for aid by subtracting aid from consumption (see
Table 9), and results do not change substantially, neither for the unadjusted nor genuine
savings indicators, indicating that regular and post event aid in this sample does not lead
to substantial changes in welfare post-disaster.
Table 9: Regression results for the low and middle income sample, sudden hydrometeorological
events revised for aid (N=62)
Model Gross savings Gross
disaster
savings
Net
savings
Net
disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant -0.047 0.026 0.052 0.112*** 0.067* 0.116***
1.02 (0.68) (1.49) (3.69) (1.76) (3.63)
Savings measures 1.283*** 1.086*** 1.269*** 1.044*** 1.133*** 1.016***
(5.53) (5.00) (5.07) (4.53) (3.92) (3.76)
R Square 0.338 0.295 0.300 0.255 0.204 0.191
Note: Significance at the * 10% level;** 5% level; *** 1% level
Low-income group
Finally, we turn to assessing events in low-income countries only. As expected, this
produces the best results in terms of explaining consumption changes by savings
measures. Table 10 reports sample information for the group of sudden
hydrometeorological events.
19
Table 10: Model results for the low income sample, sudden hydrometeorological events (N=35)
Model Gross savings Gross
disaster
savings
Net savings Net disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant -0.094 -0.043 -0.001 0.046 0.042 0.073
(1.522) 0.817 0.027 (1.043) (0.741) (1.500)
Savings
measures
1.414*** 1.403*** 1.434*** 1.471*** 1.148*** 1.323***
4.894 (5.028) (4.474) (4.736) (2.827) (3.228)
R Square 0.421 0.434 0.378 0.405 0.195 0.240
Note: Significance at the * 10% level;** 5% level; *** 1% level
To start with, while the savings measures are all highly significant, coefficients increase
to above 1. Overall, the statistically explained variation also increases substantially to,
e.g., a R square of 0.42 for gross savings. Then, most interestingly, the disaster
adjustment finally makes a difference and the explained variation increases by about 1%,
2.5% and 4.5% for gross savings, net savings and genuine savings, respectively. Best
results are still obtained when using the conventional gross savings measure. Further
revising for aid inflows improves regression results slightly and increments in
explanatory power are 3%, 4% and 5.5% respectively (see table II-4). It is important to
remember that due to a lack of genuine savings and consumption data, a number of
highly vulnerable countries, such as disaster-prone Caribbean countries (e.g., St. Lucia,
which in 1988 experienced the largest ever loss as compared to national income) are not
considered in this data set, which may improve the results in terms of explained variation
with and without accounting for disaster shocks.
We tentatively conclude that for this group of low-income countries and events,
produced capital, and thus losses therein, play a stronger role in explaining consumption
changes; furthermore, disasters losses seem to have a small adverse impact on
consumption streams, although the disaster loss variable is again highly non-significant.
To provide a graphical impression of these relationships, on Figure 6, we chart out gross
disaster savings vs. the present value of changes in consumption for this sample.
20
-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6
Gross disaster savings
-0.5
0.0
0.5
1.0
PV dc
Fig. 6: Gross disaster savings vs. the change in consumption for the low income sample,
sudden hydrometeorological events (as a share of GNI)
Further focusing in on this sample, such as separately studying flood and storm events, is
not reliably possible, as, e.g., only 20 flood, 13 storm and two wet mass movement events
remain for this income group.
4 DISCUSSION AND IMPLICATIONS OF THE RESEARCH FINDINGS
There is an ongoing debate on whether disasters cause significant macroeconomic
impacts and are truly a potential impediment to economic development. The discussion is
almost exclusively focused on impacts on GDP. We suggested that, as disasters, inter
alia, destroy capital stocks, there may be important medium-longer-term welfare effects
in terms of consumption opportunities foregone as a consequence of reduced produced
capital accumulation. Taking a longer term perspective (5 up to 33 years after an event)
we examined welfare changes in consumption potentially caused by the loss of capital
stock; we hypothesized that, if indeed those existed, national savings measures adjusted
for disaster asset losses should better explain changes in post-disaster consumption
streams.
21
Overall, we tentatively conclude that adjusting savings for disaster effects helps in better
explaining post-disaster changes in welfare, yet mostly for the low-income group of
countries. Furthermore, the estimated effect is rather small. For the whole sample, and the
combined medium and low-income groups, disaster capital stock loss adjustments to
savings does not reliably lead to improvements in explaining post-disaster consumption
changes. Also, losses by themselves do not significantly explain changes in consumption,
probably due to the small size of the effect and the many other pressures on consumption.
Furthermore, sudden onset events, and here floods and storms, perform best which can be
attributed to the fact that sudden onset events predominantly destroy assets, whereas slow
onset events such as droughts or extreme temperature incidences rather lead to longer
term indirect effects, which are not well picked up by the savings measure focusing on
accounting for investments into capital stock. Furthermore, switching from gross savings
to genuine savings mostly does not improve results in this regard. This result, somewhat
at odds with theory and empirical work, can be explained by the flexible time horizon
adopted. When using the fixed 15-year time horizon, indeed genuine savings measures
better explain the consumption changes than gross and net savings do. Accordingly, our
analysis is not to be understood as aiming to shed more light on the debate whether
genuine savings better explain consumption changes, but rather whether adjusting
disaster “depreciation” helps improve the explanatory power of savings measures
generally.
An implication of our work may be that accounting for disaster asset losses in savings
measures for disaster-exposed, low-income countries may help better explain variations
in post-disaster consumption changes and thus adjusting alternative welfare indicators
(negatively) leads to improved predictability of future post–disaster consumption
changes. For some highly disaster exposed and vulnerable countries it may be worthwhile
to explore using such further refined measures when planning policy, also given the
increasing availability of country-wide risk estimates and savings indicators.
Overall, however, we have to acknowledge the small size of the sample and the fact
that data exhibit important constraints hindering us to reasonably go beyond tentative
conclusions. A key bottleneck has been the limited number of observations, mainly due
to a lack of genuine savings data for a number of highly vulnerable countries, such as for
22
many disaster-prone Caribbean countries (e.g., St. Lucia, which in 1988 experienced the
largest ever loss as compared to national income). This lack of data for countries
expected to be particularly vulnerable to natural hazards may have lead to an important
bias in the analysis. Consequently, we might expect our findings to improve if more
observations are added. Constructing a more comprehensive and detailed database of
disaster losses and savings measures may help to some extent with better addressing this
problem, yet, by definition, the study of extreme events will always be constrained by
scarce and imprecise data.
23
5 REFERENCES
Albala-Bertrand, J. M. (1993). Political Economy of Large Natural Disasters With Special
Reference to Developing Countries. Oxford, Clarendon Press.
Albala-Bertrand, J. M. (2006). The Unlikeliness of an Economic Catastrophe: Localization &
Globalization. Working Papers 576, Queen Mary, University of London, Department of
Economics.
Arellano, C., Bulí?, A., Lane, T., Lipschitz, L. (2009). The dynamic implications of foreign aid
and its variability. Journal of Development Economics 88: 87–102
Baritto, F. (2008). Disasters, Vulnerability and Resilience. Working paper. UN-ISDR.
Benson, C. (1997a). The economic impact of natural disasters in Fiji. Working Paper No.97.
London, UK, Overseas Development Institute (ODI).
Benson, C. (1997b). The economic impact of natural disasters in the Philippines. Working Paper
No.99. London, UK, Overseas Development Institute (ODI).
Benson, C. (1997c). The economic impact of natural disasters in Viet Nam. Working Paper
No.98. London, UK, Overseas Development Institute (ODI).
Benson, C. (1998). The cost of disasters. Development at risk? Natural disasters and the third
World. J. Twigg. Oxford, Oxford Centre for Disaster Studies, UK National Coordinated
Committee for the International Decade for Natural Disaster Reduction (IDNDR): 8-13.
Benson, C. and Clay, E. (1998). The impact of drought on Sub-Saharan African economies.
World Bank Technical Paper Series No. 401. Washington, D.C., The World Bank.
Benson, C. and Clay, E. (2000). Developing countries and the economic impacts of catastrophes.
Managing disaster risk in emerging economies. A. Kreimer, Arnold, M. (eds.). Washington,
D.C., World Bank Publication.
Benson, C. and Clay, E. (2001). Dominica: Natural disasters and economic development in a
small island state. Disaster Risk Management Working Paper Series No.2. Washington, D.C.,
The World Bank.
Caselli, F. and Malhotra, P. (2004). Natural Disasters and Growth: from Thought Experiment to
Natural Experiment. Washington DC, IMF.
Charveriat, C. (2000). Natural disasters in Latin America and the Caribbean: An overview of risk.
Working Paper No. 434. Washington, D.C., Inter-American Development Bank (IDB).
CRED (2009). EM-DAT: International Disaster Database. Brussels, Belgium, Centre for
Research on the Epidemiology of Disasters, Université Catholique de Louvain.
Crowards, T. (2000). Comparative vulnerability to natural disasters in the Caribbean. Charleston,
South Carolina, Caribbean Development Bank.
Cuaresma, J., Hlouskova, J. ,Obersteiner, M (2008). Natural disasters as creative destruction.
Evidence from developing countries. Economic Inquiry 46: 214-226
Dasgupta, P., and Maeler, K.-G. (2001). Wealth as a Criterion for Sustainable Development.
Discussion Paper 139. Stockholm, Beijer International Institute of Ecological Economics.
ECLAC (1999). Evaluacion de los danos ocasionados por el Huracan Mitch. Mexico City.
ECLAC (2002). Handbook for estimating socio-economic and environment effects of disasters.
Mexico City.
Ferreira, S., and Vincent, J. (2005). Genuine Savings: Leading Indicator of Sustainable
Development?” Economic Development and Cultural Change 53:737–54
24
Ferreira,S., Hamilton, K, and Vincent, J. (2008). Comprehensive Wealth and Future
Consumption: Accounting for Population Growth. The World Bank Economic Review, 22:
233–248
Freeman, P. K., L. A. Martin, R. Mechler, K. Warner with P. Hausmann (2002). Catastrophes and
Development. Integrating Natural Catastrophes into Development Planning. Washington DC,
World Bank.
Hamilton, K., and Hartwick, J. (2005). Investing Exhaustible Resource Rents and the Path of
Consumption. Canadian Journal of Economics 38 :615–21
Hamilton, K., and Atkinson, G. (2006). Wealth, Welfare and Sustainability: Advances in
Measuring Sustainable Development. Cheltenham, UK: Edward Elgar.
Hochrainer, S. (2006). Macroeconomic risk management against natural disasters. Wiesbaden,
German University Press (DUV).
Mechler, R. (2004). Natural Disaster Risk Management and Financing Disaster Losses in
Developing Countries. Karlsruhe, Verlag für Versicherungswissenschaft.
Munich Re (2008). Disaster loss data. Munich Re NatCat Service. Munich, Munich Re.
Murlidharan, T. L. and H. C. Shah (2001). Catastrophes and Macro-economic Risk Factors: An
Empirical Study. Proceedings of First Annual IIASA-DPRI Meeting on Integrated Disaster
Risk Management: Reducing Socio-Economic Vulnerability, 1-4 August 2001, Laxenburg,
IIASA.
Noy, I. (2009). The macroeconomic consequences of disasters. Journal of Development
Economics 88: 221–231
Okuyama, Y. (2008). Critical review of Methodologies on disaster impacts estimation.
Background paper for EDRR report.
Okuyama, Y. (2009). Impact estimation of higher order effects. Background paper for EDRR
report, GFDRR, Washington DC.
Otero, R. C. and R.Z. Marti (1995). The impacts of natural disasters on developing economies:
implications for the international development and disaster community. In M. Munasinghe
and C. Clarke (eds.). Disaster Prevention for Sustainable Development: Economic and Policy
Issues. Washington DC, World Bank: 11-40
Provention Consortium (2002). The quality and accuracy of disaster data. A comparative analysis
of three global data sets. Geneva, ProVention Consortium.
Rasmussen, T.N. (2004). Macroeconomic implications of natural disasters in the Caribbean. IMF
Working Paper WP/04/224.
Sanderson, W.C. and Striessnig, E. (2009). Demography, Education, and the Future of Total
Factor Productivity Growth. IIASA Interim Report IR-09-002, Laxenburg, Austria.
Skidmore, M. and Toya, H. (2002). Do Natural Disasters Promote Long-Run Growth? Economic
Inquiry 40: 664-687
Telford, J., Arnold, M., Harth, A. with ASONOG (2004). Learning Lessons from Disaster
Recovery. The Case of Honduras. Disaster Risk Management Working papers Series No. 8.
World Bank, Washington DC.
Toya, H. and Skidmore, M (2007). Economic development and the impacts of natural disasters.
Economics Letters 94: 20-25
World Bank (2006). Where is the wealth of nations?, Washington DC, World Bank.
World Bank (2009). World Development Indicators 2008, Washington DC, World Bank.
25
APPENDIX I: ADDITIONAL DATA AND INFORMATION
Table I-1: List of countries, events, losses and country characteristics (N=99)
Country year Income
group
Type Loss %
GNIt-1
Algeria 1980 M Earthquake -0.03
Armenia 1997 M Earthquake -0.007
Armenia 2000 M Drought -0.018
Australia 1974 H Flood -0.008
Australia 1981 H Drought -0.011
Bangladesh 1974 L Flood -0.022
Bangladesh 1987 L Flood -0.011
Bangladesh 1988 L Flood -0.027
Bangladesh 1991 L Storm -0.032
Bangladesh 1995 L Storm -0.007
Bangladesh 1998 L Flood -0.033
Belize 2000 M Storm -0.135
Bolivia 1982 L Flood -0.037
Bolivia 1983 L Drought -0.05
Bolivia 1983 L Drought -0.06
Bolivia 1992 L Mass movement wet -0.026
Cambodia 2000 L Flood -0.015
Canada 1977 H Drought -0.005
Chile 1985 M Earthquake -0.021
China 1991 L Flood -0.012
China 1993 L Flood -0.008
China 1994 L Drought -0.01
China 1996 L Flood -0.013
China 1998 L Flood -0.011
Colombia 1999 M Earthquake -0.007
Costa Rica 1996 M Flood -0.007
Czech Republic 1997 M Flood -0.012
Dominica 1989 M Storm -0.045
Dominica 1995 M Storm -0.295
Dominica 1995 M Storm -0.034
Dominican
Republic 1979 M Storm -0.127
Dominican
Republic 1998 M Storm -0.047
Ecuador 1987 M Earthquake -0.034
Ecuador 1993 M Mass movement wet -0.015
Egypt, Arab Rep. 1992 M Earthquake -0.01
El Salvador 1982 M Flood -0.027
El Salvador 1986 M Earthquake -0.142
El Salvador 1998 M Storm -0.012
Fiji 1993 M Storm -0.026
26
Georgia 2000 M Drought -0.02
Guatemala 1976 M Earthquake -0.095
Guatemala 1998 M Storm -0.015
Haiti 1994 L Storm -0.009
Haiti 1998 L Storm -0.019
Honduras 1974 L Storm -0.162
Honduras 1982 L Storm -0.012
Honduras 1990 L Flood -0.008
Honduras 1998 L Storm -0.303
Hungary 1986 M Drought -0.009
Hungary 1992 M Drought -0.005
India 1993 L Flood -0.008
Indonesia 1997 M Wildfire -0.012
Iran, Islamic Rep. 1999 M Drought -0.01
Japan 1995 H Earthquake -0.008
Jordan 1992 M Extreme temperature -0.033
Macedonia, FYR 1995 M Flood -0.032
Madagascar 1977 L Storm -0.049
Madagascar 1981 L Storm -0.019
Madagascar 1982 L Storm -0.021
Madagascar 1984 L Storm -0.024
Madagascar 1986 L Storm -0.016
Mauritania 1979 L Drought -0.016
Mauritius 1989 M Storm -0.01
Mauritius 1994 M Storm -0.014
Mauritius 1999 M Drought -0.014
Mexico 1985 M Earthquake -0.009
Moldova 1997 M Flood -0.008
Moldova 2000 M Storm -0.006
Mongolia 1990 L Wildfire -0.011
Mongolia 1996 L Wildfire -0.567
Mongolia 2000 L Storm -0.027
Morocco 1999 M Drought -0.009
Mozambique 1990 L Drought -0.007
Mozambique 2000 L Flood -0.036
Nepal 1980 L Earthquake -0.04
Nepal 1987 L Flood -0.082
Nepal 1988 L Earthquake -0.026
Nepal 1993 L Flood -0.017
Nicaragua 1972 M Earthquake -0.332
Nicaragua 1982 M Storm -0.052
Nicaragua 1988 M Storm -0.039
Nicaragua 1991 M Wildfire -0.02
Nicaragua 1998 M Storm -0.104
27
Pakistan 1973 L Flood -0.019
Pakistan 1976 L Flood -0.015
Pakistan 1992 L Flood -0.007
Philippines 1972 M Flood -0.012
Philippines 1990 M Earthquake -0.008
Poland 1997 M Flood -0.009
Senegal 1976 L Drought -0.053
Sri Lanka 1978 M Storm -0.007
Sri Lanka 1992 M Flood -0.01
Swaziland 1984 M Storm -0.03
Tajikistan 1992 L Flood -0.053
Tajikistan 1993 L Mass movement wet -0.026
Tajikistan 1998 L Flood -0.024
Tajikistan 2000 L Drought -0.019
Turkey 1999 M Earthquake -0.02
Venezuela, RB 1999 M Flood -0.014
28
Table I-2 Information on direct, indirect and capital stock losses for selected disaster events
Hurricane Stan in El Salvador, 2005
(million USD)
Losses
Sector Direct Indirect Total
Social (housing, education, health) 48 102 150
Productive (agriculture, industry, commerce, tourism) 22 34 56
Infrastructure (water and sanitation, electricity,
transport)
106 8 114
Environment 21 1 22
Emergency and relief expenditure 11 11
Total (ECLAC) 196 145 352
Total loss according to EMDAT 356
Capital stock losses (productive sector and
infrastructure) as a share of total loss
36%
Capital stock losses (productive sector and
infrastructure) as a share of total loss (EMDAT)
36%
Hurricane Stan in Guatemala, 2005
(million Quetzales)
Losses
Sector Direct Indirect Total
Social (housing, education, health) 630 543 1,173
Productive (agriculture, industry, commerce, tourism) 306 1,736 2,042
Infrastructure (water and sanitation, electricity,
transport) 1,960 1,437 3,396
Environment 308 … 308
Emergency and relief expenditure 595 595
Total (ECLAC) 3,203 3,716 7,514
Total loss according to EMDAT 7,542
Capital stock losses (productive sector and
infrastructure) as a share of total loss 30%
Capital stock losses (productive sector and
infrastructure) as a share of total loss (EMDAT) 30%
Hurricane Mitch in Honduras 1998
(million USD)
Losses
Sector Direct Indirect Total
Social (housing, education, health) 305 719 1,024
Productive (agriculture, industry, commerce, tourism) 1,478 577 2,055
Infrastructure (water and sanitation, electricity,
transport) 348 164 512
Environment 47 0 47
Emergency and relief expenditure 156 156
Total (ECLAC) 2,178 1,460 3,794
Total loss according to EMDAT 3,794
Capital stock losses (productive sector and
infrastructure) as a share of total loss 48%
Capital stock losses (productive sector and
infrastructure) as a share of total loss (EMDAT) 48%
29
Table I-2 Direct, indirect and capital stock losses for selected disaster events (continued)
Hurricane Mitch in Nicaragua 1998
(million USD)
Losses
Sector Direct Indirect Total
Social (housing, education, health) 225 45 270
Productive (agriculture, industry, commerce, tourism) 128 57 185
Infrastructure (water and sanitation, electricity,
transport) 159 147 306
Environment na Na -
Emergency and relief expenditure 227 227
Total (ECLAC) 512 249 988
Total loss according to EMDAT 988
Capital stock losses (productive sector and
infrastructure) as a share of total loss 29%
Capital stock losses (productive sector and
infrastructure) as a share of total loss (EMDAT)
29%
Earthquake in El Salvador, 2001
(million USD)
Losses
Sector Direct Indirect Total
Social (housing, education, health) 496 120 616
Productive (agriculture, industry, commerce, tourism) 244 96 340
Infrastructure (water and sanitation, electricity,
transport) 97 375 472
Environment 102 1 103
Emergency and relief expenditure 73 73
Total (ECLAC) 939 591 1,604
Total loss according to EMDAT 1,500
Capital stock losses (productive sector and
infrastructure) as a share of total loss 21%
Capital stock losses (productive sector and
infrastructure) as a share of total loss (EMDAT) 23%
Sources: ECLAC, 1999, 2002; Telford et al., 2004; CRED, 2009
30
APPENDIX II: FURTHER RESULTS OF THE REGRESSION ANALYSIS
Table II-1 Regression results for the sudden onset events (N=77)
Model Gross
savings
Gross
disaster
savings
Net
savings
Net
disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dconsumption in per cent of per capita income of the year before the event
Constant 0.050 0.096*** 0.114*** 0.147*** 0.127*** 0.154***
(1.13) (2.73) (3.48) (5.36) (3.90) (5.62)
Savings measures 0.730*** 0.606*** 0.689*** 0.571*** 0.614** 0.561**
(3.40) (3.27) (3.00) (2.92) (2.45) (2.57)
R Square 0.133 0.125 0.107 0.102 0.074 0.081
Note: Significance at the * 10% level;** 5% level; *** 1% level
Table II-2 Regression results for the slow onset event sample (N=21)
Model Gross
savings
Gross
disaster
savings
Net
savings
Net disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant 0.097 0.131* 0.156** 0.182*** 0.172*** 0.204***
(1.17) (1.99) (2.60) (3.58) (3.50) (4.24)
Savings measures 0.577 0.515 0.552 0.537 0.729* 0.494
(1.56) (1.56) (1.29) (1.38) (2.00) (1.68)
R Square 0.113 0.113 0.081 0.091 0.174 0.129
Note: Significance at the * 10% level;** 5% level; *** 1% level
31
Table II-3 Regression results for the sudden-onset, hydrometeorological events sample
(N=62)
Model Gross savings Gross
disaster
savings
Net
savings
Net
disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant -0.031 0.041 0.061* 0.120*** 0.075** 0.124***
0.67 (1.07) (1.79) (4.03) (2.01) (3.95)
Savings measures 1.214*** 1.014*** 1.229*** 0.995*** 1.109*** 0.974***
(5.33) (4.74) (5.05) (4.42) (3.96) (3.71)
R Square 0.325 0.276 0.302 0.249 0.210 0.189
Note: Significance at the * 10% level;** 5% level; *** 1% level
Table II-4 Regression results for the low income sample, sudden hydrometeorological events
revised for aid (N=35)
Model Gross savings Gross
disaster
savings
Net
savings
Net
disaster
savings
Genuine
savings
Genuine
disaster
savings
PV dConsumption in per cent of per capita income of the year before the event
Constant -0.112* -0.063 -0.013 0.033 0.033 0.062
1.77 1.17 0.25 (0.73) (0.55) (1.22)
Savings measures 1.472*** 1.483*** 1.461*** 1.526*** 1.155*** 1.369***
(4.94) (5.21) (4.36) (4.75) (2.73) (3.23)
R Square 0.425 0.452 0.366 0.406 0.184 0.240
Note: Significance at the * 10% level;** 5% level; *** 1% level