Environmental Impact Assessment Review 64 (2017) 5766

Contents lists available at ScienceDirect

Environmental Impact Assessment Review

journal homepage: www.elsevier.com/locate/eiar

Socio-economic exposure to natural disasters

Giovanni Marin a,b,c, Marco Modica b,c,
a
Department of Economics, Society, Politics, University of Urbino 'Carlo Bo', via Aurelio Saf, 2, 61029 Urbino, Italy
b
IRCrES - CNR, Research Institute on Sustainable Economic Growth, Via Corti 12, 20133 - Milano, Italy
c
SEEDS, Ferrara, Italy

a r t i c l e i n f o a b s t r a c t

Article history: Even though the correct assessment of risks is a key aspect of the risk management analysis, we argue that limited
Received 20 October 2016 effort has been devoted in the assessment of comprehensive measures of economic exposure at very low scale.
Received in revised form 13 February 2017 For this reason, we aim at providing a series of suitable methodologies to provide a complete and detailed list
Accepted 2 March 2017
of the exposure of economic activities to natural disasters. We use Input-Output models to provide information
Available online xxxx
about several socio-economic variables, such as population density, employment density, rms' turnover and
Keywords:
capital stock, that can be seen as direct and indirect socio-economic exposure to natural disasters. We then pro-
Economic exposure vide an application to the Italian context. These measures can be easily incorporated into risk assessment models
Disaster impact to provide a clear picture of the disaster risk for local areas.
Risk assessment 2017 Elsevier Inc. All rights reserved.
Risk management

1. Introduction appear to be highly correlated, it is important to consider at the same

The perception about the relevance of economic and social damages
generated by natural disasters has grown substantially in recent de-
cades (Blaikie et al., 2014). This greater awareness about natural disas-
ters triggered the demand, from both the public and the private
sectors, of actions aimed at preventing the occurrence of natural disas-
ters (when possible), at mitigating the damages and at adapting to in-
creasing risks (Zelekov and Zvijkov, 2017). Information about
risk and exposure to damages is fundamental both in an ex-ante per-
spective (e.g. risk reduction, risk assessment) and in an ex-post perspec-
tive (e.g. risk management, assessment of damage, reconstruction).
Even though the correct assessment of risks is a key aspect of the risk
management analysis (and for this reason, the academic literature con-
tains a large variety of sophisticated models for risk assessment), we
argue that less effort has been devoted in the denition of comprehen-
sive measures of socio-economic exposure at very low scale. For this
reason, we aim at discussing a selection of methodologies suitable to
provide a complete and detailed assessment of the exposure of econom-
ic activities to natural disasters. We are then able to provide detailed in-
formation on several socio-economic variables that can be easily
incorporated into risk assessment models to provide a clear picture of
the disaster risk. These variables are population density, employment
density, rms' turnover and capital stock (divided also in its main com-
ponents: buildings and machineries). Even though these measures
Corresponding author at: CNR-IRCrES, via Corti 12, 20133 Milan, Italy.
E-mail addresses: giovanni.marin@uniurb.it (G. Marin), marco.modica@ircres.cnr.it
(M. Modica).
time all the various aspects of socio-economic exposure.
In this respect, given the difculties to address all the exposed values
and the absence of clear methodologies able to dene proper socio-eco-
nomic exposure values, the existing literature employs different proxies
of socio-economic exposure (Chen et al., 1997) that depend on the fea-
tures of the disaster that is analyzed. In fact, the choice of the proxy used
in the evaluation of the economic losses due to natural disasters mostly
depends on the sequence of effects which are expected to occur when
different natural disasters affect a given area (Modica and Zoboli,
2016; Pelling, 2003). For instance, the density of the built environment
is a common proxy used in the case of ood risk assessment (e.g.
Jongman et al., 2012; Koks et al., 2014, 2015: Sterlacchini et al., 2016).
Gross Domestic Product, (GDP) or population density is commonly
used in earthquake risk assessment (Chen et al., 1997), as well as the
value of real estate assets (Field et al., 2005; Meroni et al., 2016). A sim-
ilar proxy can be used in the case of drought (land value, Simelton et al.,
2009) while for landslides an interesting measure is the mix between
social (population), physical (buildings and infrastructures), economic
(land value) and environmental (site of community importance) indi-
cators (Bloechl and Braun, 2005; Pellicani et al., 2014).
On these regards, there is extensive literature focusing on the deni-
tion of losses caused by extreme events (see ECLAC, 2003; FEMA, 1992;
and Pelling et al., 2002, for more details). Even though denitions are
not always coherent among each other, for simplicity we discriminate
between direct and indirect losses.
Direct losses refer to direct damages to people (injuries and fatali-
ties) and objects (e.g. goods, buildings, infrastructures; see ECLAC,
2003). For instance, earthquakes destroy buildings and infrastructures,
which in turn, generate damages to other goods and people. Floods

http://dx.doi.org/10.1016/j.eiar.2017.03.002
0195-9255/ 2017 Elsevier Inc. All rights reserved.
58 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
may generate minor damages to buildings if compared to damages to
other goods (e.g. vehicles) and people (FEMA, 1992; Luino et al.,
2009). Damages arising from the interruption of economic activities
due to the natural disaster are also considered to be direct losses (see
Rose and Lim, 2002; Rose et al., 2007). Interruptions may occur for sev-
eral reasons: damages to critical infrastructures such as energy and
water supply and transport network; damages to people involved in
production processes, destruction of production capital, etc. The inter-
ruption of economic activities in a region reduces the rms' turnover
for a certain amount of time, which in turn reduces region's GDP.
The category of indirect losses is, instead, broad and borderless. Lim-
iting the discussion to business interruption, foregone production and
turnover inuence the whole (local and global) supply chain of the pro-
duction activities that experience the interruption (e.g. Van Der Veen
and Logtmeijer, 2005). Suppliers of intermediate goods will experience
a reduction in the demand for their products and consequently a reduc-
tion in turnover. On the other hand, customers will experience potential
shortages of inputs needed for their production process and may be
forced to nd alternative suppliers, thus increasing production costs
and potentially reducing production. Foregone wages will also inuence
region's GDP as consumption will be reduced. Finally, if the interruption
lasts for a long period, producers may lose their customers permanently,
limiting the possibility of economic recovery even once the cause of in-
terruption is removed. For all these reasons indirect losses need to be
evaluated looking at general equilibrium effects by means of specic
economic models (see Okuyama, 2007).
Given these premises it turns out that policy makers and private ac-
tors have the vital need to have knowledge about both the direct and in-
direct components of economic exposure. Indeed, policy makers need to
know clearly what is the socio-economic value of the area under analy-
sis, as well as the possible interconnections with neighbouring areas. In
this way it is possible to dene optimal disaster mitigation policies, es-
timate the likely (or potential) maximum cost suffered by a region
and elaborate effective risk management strategies. Private actors,
such as insurance companies, can instead use this information to dene
more accurate risk analysis (i.e. to provide better insurance plans).
Direct components of exposure refer to those that might produce di-
rect losses because of a disaster (direct socio-economic exposure). In-
direct components refer to losses due to disruption of local and global
supply chains of the production activities (indirect socio-economic ex-
posure). However, measuring the socio-economic impacts of extreme
events is a difcult task due to the unpredictability of the different
types of natural events (Hallegatte and Przylusky, 2010), and to the
scarcity of information about economic activities for small geographical
units, both in an ex-ante and in an ex-post perspective. We propose sim-
ple methodologies to dene several socio-economic exposure measures
that can be used in different contexts and in relation to several natural
disasters at a very detailed scale, providing in this way a full map of
the potential exposure of socio-economic activities to natural disasters.
We then test these methodologies in the Italian context.
According to the literature (see Cardona et al., 2012), elements that
are exposed to natural disasters are the human beings and all the other
related basics constituting their livelihoods and assets. For this reason,
and according to previous works that have reviewed the literature of
the economic assessment of natural disasters (Modica and Reggiani,
2015; Modica et al., 2017), we mainly focus on a set of recurrent mea-
sures that consider the three aspects mentioned above (e.g. human be-
ings, livelihoods and assets): population density, employee density,
turnover) and capital stock. These measures provide interesting infor-
mation on the direct exposure since they are all proxy for the local
loss due to natural disasters. In order to consider the indirect socio-eco-
nomic exposure, we provide evidence about local linkages and possible
diffusion of economic damages by means of input-output inter-sectoral
linkages across neighbouring municipalities. Moreover, we also provide
descriptive evidence on the spatial clustering of the socio-economic ac-
tivities as measured by our set of indicators by means of the Local
Indicator of Spatial Autocorrelation (LISA, see Anselin, 1995 and Cutter
and Finch, 2008) in order to identify areas where high values are spatial-
ly concentrated.
The paper then aims at providing information on the direct and indi-
rect socio-economic exposure of territories to natural disasters by
means of selected variables that can be used by policy makers and pri-
vate actors for the elaboration and implementation of correct and effec-
tive risk management strategies. For these reasons, we dene a series of
suitable methodologies that are able to provide at a very low scale a
comprehensive assessment of the exposure of economic activities to
natural disasters. These methodologies offer robust and replicable
tools for the risk management at different administrative levels (munic-
ipalities, regions, national government) and provide important informa-
tion on local exposure per se. We then expect to identify areas with high
concentration of economic activities, indicating in this way a greater
economic exposure to risk and potentially higher indirect losses.
2. Material and method
This section describes different methodologies that are useful to de-
ne adequate socio-economic measures of exposure. While some of
these measures are readily available from statistical or administrative
sources, other measures need to be treated with appropriate methodol-
ogies. We deal in this section only with those measures that need to be
estimated and leave any further discussion to the Results section for
those that do not need any preliminary operations. We mainly differen-
tiate methodologies able to dene direct components of exposure
(Section 2.1) and indirect components (Section 2.2).
2.1. Direct components of exposure: estimating turnover and capital stock
at municipal level
According to the existing literature (Cardona et al., 2012; Modica
and Reggiani, 2015) suitable proxies able to provide useful information
for the potential direct local losses suffered by selected areas due to
natural disasters are: population density (proxy for potential life loss),
employee density (proxy for exposure during working hours), turn-
over (direct costs due to business interruption) and capital stock (direct
costs due to the destruction of capital goods). While population and em-
ployee density can be timely retrieved from ofcial statistics and cen-
suses, turnover and capital stock need to be estimated in order to
provide appropriate information of direct socio-economic exposure of
territories. The reason is the following: actual turnover and capital
stock data are not available for most of the rms as a public and ofcial
data (e.g. rms' turnover is available for turnover bands) and no direct
information is available for local units. Then, these facts do not allow
obtaining detailed information at low scale (e.g. municipal level) of
both the measures of capital and turnover and they need to be estimat-
ed. These estimates are based on the combination of statistical and ad-
ministrative data sources and rely on the validity of a set of clearly
dened assumptions.
The estimate of turnover is necessary because actual turnover is typ-
ically available only for a small number of large rms. For smaller rms,
it is possible to gain access to information on turnover bands rather than
to exact information on turnover. On the other hand, information on
employment for the universe of rms and local units is easily available
and it is usually collected in business registries, censuses or social secu-
rity registries. This claims for the possibility of estimating the value of
economic production (i.e. turnover) that is generated within a munici-
pality by exploiting information on employment of the local units. On
this regard, we adopt the following two-step procedure to estimate
rm-level turnover at the municipality level. First we estimate rm-
level turnover by means of an interval regression model (a generalisa-
tion of the Tobit model, in which both interval data and point data are
allowed). Second, we attribute the total turnover of the rm to its
local units by assuming that the ratio between turnover and number
 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
of employees is homogenous across local units within a rm. In this
way, we are able to estimate the rms' turnover at local level.
In the rst step we t an interval regression model on the population
of rms in which the log of turnover is function of the log of the number
of employees of the rm. The interval regression model (for details see
Stewart, 1983 and Wooldridge, 2010) is estimated separately for each
industry (NACE Subsection) to account for industry-specic relation-
ship between employment and turnover



log yiik log Liik i 1 sector k while Lm = kLm
where yii k is either exact turnover (typically for large rms) or informa-
tion on the band of turnover of rm i belonging to industry k and Lik i is
rm employment for the same rm i. After estimating the model, we
predict the turnover at any level of employment. To improve the esti-
mates, whenever actual turnover is available (e.g. the commercial data-
base AIDA, provided by Bureau van Dijk, gives information on true
turnover data for medium and large European rms), we substitute
the predicted turnover with this true turnover. Instead, for the set of
rms with no information on actual turnover (typically small rms), it
could happen that the predicted value may exceed or be less than (re-
spectively) the upper and lower bound of the turnover interval that
has been used to estimate the predicted value. In this eventuality, we
substitute the predicted turnover with the upper bound of the interval,
whenever it exceeds the upper bound. Similarly, we substitute the pre-
dicted turnover with the lower bound whenever it is smaller than the
lower bound.
The second step of the procedure consists in attributing the total es-
timated turnover of the rm to its local units. This is particularly impor-
tant as many rms operate in a number of different locations.
Attributing all the turnover of these rms to the location of the head-
quarter is likely to underestimate the exposure to risk of small towns,
as headquarters tend to locate in large urban centres, and consequently
overestimate the exposure to risk of large urban centres. To attribute
turnover to local units, we assume that the ratio between turnover
and the number of employees is homogenous across local units j of
each rm i (y ^i =Li). Therefore, we attribute rm-level turnover
^i j =Li j y
^i ) to local units (j) according to the share of employment of the local
(y
unit j (Lj) over the total employment of the rm i (Li).
Lj
y ^i 
^j y
Li
Total turnover at the municipality level (m) is nally computed as
the sum of estimated turnover of local units operating in the municipal-
ity.
y ^j
^m jm y
While total estimated local turnover is already useful as an indicator
of exposure, it is important to scale this value by the total land area of
the municipality (both urban and non-urban areas), thus obtaining
the average turnover per square kilometre. This scaling factor is re-
quired as natural disasters (earthquakes, oods, landslides) affect spe-
cic areas of land that may not cover the entire municipal territory.
Another important component of the exposure at local level is the
net stock of capital that provides a measure for the quantication of
the assets at risk. However, the estimation of the net stock of capital of
economic activities at municipality level may result much more compli-
cated if compared with the procedure adopted to estimate the munici-
pal turnover. First, because book value of physical assets is only
available for those rms, typically large rms, for which the detailed bal-
ance sheet is available in commercial databases. Second, even that mea-
sure is the result of the application of accounting rules (e.g. about
depreciation) that are not consistent with statistical standards. In our
procedure, we assume that the ratio between capital stock and number
59
of employees is homogeneous in all the rms that belong to a specic
sector (k).
According to this assumption, the estimated capital stock for munic-
ipality m is given by:
Capitalk Lm
Capitalm k  km 4
Lk L
where Lm k is the total employment in local units in municipality m and
k is total employment in municipality m. This
means that the only source of variation across municipalities is the in-
dustry composition of economic activities of the municipalities. As
discussed for turnover, we scale the capital stock by the land area of
the municipality.
The main difference in the way turnover and capital stock are esti-
mated consists in the assumptions that need to be made to attribute a
value to each local unit and, consequently, to each municipality. For
what concerns turnover, two different mild assumptions are needed.
First, we assume that turnover is proportional to rm employment
within each turnover band, even though we allow the degree of propor-
tionality to differ across sectors. The second assumption is that within a
rm, production and consequently turnover, are generated by each local
unit proportionally to its employment level. For what concerns munici-
pality-level capital stock, only one but strong assumption is needed:
the capital/labour ratio is assumed to be the same for all local units
within a sector at the national level.
Summarizing, the indicators of turnover and capital are suitable
measures able to provide synthetic information on the value of ele-
ments at risk at municipal level. It also turns out that these indicators
could be useful to dene plausible damage scenarios. For instance, the
estimated turnover quanties the average potential expected loss of a
natural disaster affecting a local area in terms of drop of turnover. This
means that, supposing for simplicity homogeneity in the spatial distri-
bution of the productive activities in the area of the municipality, the
turnover indicator could be interpreted as an average economic loss
(e.g. zero production) if the overall economic activity is completely
interrupted for a year in a square kilometre of an area because of a nat-
ural disaster. This is due to the fact that a complete interruption of pro-
duction activities for one year entails a complete loss of turnover for that
year. The indicator can be rescaled to reect the expected loss of turn-
over due to temporary interruption (e.g. full interruption for just
2
7 days in a year, that is 1.9% of a year). At the same time, the indicator
of net capital stock reects the average expected loss in capital stock if
all productive physical capital is destroyed in a square kilometre of
land. Complete destruction of all production capital entails a permanent
direct loss of the monetary value of the capital that is destroyed. Also for
the case of capital, the indicator can be rescaled to evaluate partial loss
3 (e.g. just 50% of the xed capital is destroyed as a consequence of the
natural disaster).
The two indicators then describe two different dimensions of eco-
nomic exposure. Turnover loss is an interruption in the ow of econom-
ic production. Once the cause of the interruption is removed, the ow of
economic production can be recovered. On the other hand, capital loss
entails the irreversible destruction of the stock of economic capital.
This means that even when the cause of the destruction is removed,
the economic damage persists.
2.2. Indirect components of exposure: estimating turnover and capital stock
at municipal level
At the macro-economic level, the economic literature has focused on
inter-sectoral (input output) relationships as a mechanism through
which shocks diffuse throughout the economy (Acemoglu et al.,
2012). The general idea is that the collapse of a sector has an impact
on other sectors that are either upstream (suppliers) or downstream
(customers) in the supply chain. If the output of a sector experiences
60 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
an unexpected drop, rms in that sector will respond by reducing their
demand for intermediate inputs from upstream sectors. Suppliers, in
turn, will also reduce their production as one of their customer sectors
has reduced its demand. Moreover, a collapse in the production of a sec-
tor will also generate shortages (at least in the short run) of inputs for
those sectors that need goods or services produced by the collapsed sec-
tor (i.e. downstream sectors). These shortages may induce a reduction
in the production activities of downstream sectors.
Input-output modelling is a useful tool in this framework. Input-out-
put tables describe intersectoral monetary ows of intermediate inputs
across all sectors in the economy (Miller and Blair, 2009). The input-
output model, proposed by Wassily Leontief, allows estimating how
changes in the demand of certain commodities inuence the level and
structure of production of an economic system through changes in the
demand for intermediate inputs. Even though approaches based on
inputoutput modelling have been criticized for their reliance on strong
assumptions (i.e. rigid structure and lack of responses to price changes,
see Rose, 2004), they have been widely used in the assessment of disas-
ter damages, mostly because of their attitude to capture the economic
interdependencies within a regional economy (Okuyama and Santos,
2014; Perminova et al., 2016). For instance Tan et al. (2015) use
input-output modelling to estimate the maximum percentage of dam-
age that an economic system is able to absorb. Rose and Wei (2013)
simulate the impact of a seaport shutdown to the national economy,
while Van Der Veen and Logtmeijer (2005) simulate the impact of a
large-scale ooding in one province of Netherlands.
To translate this approach to the local level, a number of remarks are
needed. First, statistical agencies only provide estimates of national
input-output tables. This means that it is not possible to account for like-
ly geographic heterogeneity in technical input-output coefcients. Even
though these coefcients generally reect rather standard recipes for
producing nal goods, some heterogeneity exists.
Second, the share of inputs that a rm purchases from suppliers and
the share of output that a rm sells to customers are split into a domes-
tic and rest of the world transaction tables in national input-output ta-
bles, while no information is available about extra-municipality trade
with municipalities within the same country. This means that there is
no way of knowing the extent to which rms in the focal municipality
rely on inputs from suppliers (and sell products to nal or intermediate
consumers) within the same municipality or to any other municipality
within the country. In order to account these two issues, and because
of the absence of any information about the relevance of local suppliers
and customers, we use the national domestic input-output table as an
upper-bound of the relevance of local suppliers and customer. The im-
plicit assumption is that all national inputs are completely purchased
from rms in neighbouring municipalities. Even though this is a very
strong assumption, the absence of information about cross-regional
and cross-municipal trade ows limits the possibility of providing any
better estimate of the share of local sourcing of intermediates over
total national sourcing of intermediates.
A third precaution relates to the denition of local linkages, namely
how far should be the municipality affected by natural disasters to inu-
ence the economy of other municipalities (sth et al., 2016). To evaluate
the sensitivity of our mapping to this issue we report results for two dif-
ferent thresholds of distance: 20 km in the main text and 50 km in the
Supplementary material, see Fig. A1.
Finally, the diffusion of shocks rstly inuences direct suppliers and
customers. Then, the shock propagates to suppliers of suppliers and cus-
tomers of customers. To simplify, in our analysis we only consider rst
order effects and leave the evaluation of higher order effects (i.e. using
the Leontief model for suppliers and the Ghoshian model for customers)
for future research or specic case studies.
Keeping in mind these methodological premises, we build two dif-
ferent measures of socio-economic indirect exposure by using input-
output modelling. The rst (Destination of nal output in neighbouring
municipalities) measures the share of output (turnover) created in a
municipality that can be absorbed as intermediate inputs by rms that
operate in neighbouring municipalities (i.e. downstream impacts). The
second measure is Source of intermediate inputs in neighbouring munic-
ipalities, that computes the share of intermediate inputs that can be
possibly retrieved locally by rms that operate in the municipality of
reference by any sector (i.e. upstream impacts).
The method dening the rst measure (Destination of nal output in
neighbouring municipalities) is the following. For each sector: i) we
compute the total turnover in the municipalities within the radius of
the reference municipality (ykn) as the sum of estimated turnover of
local units that reside in these municipalities; ii) the diagonal matrix
of total turnover for all sectors (byn N) is then post-multiplied by the
technical coefcient matrix of the input-output matrix,1 resulting in
the hypothetical demand of intermediate inputs from each sector in
neighbouring municipalities; iii) we divide the hypothetical demand
by the output produced in the municipality of reference for each sector
(ykm) and take the weighted (by turnover) average across sectors.
More formally, the measure is given by:
0 1
1=y1m
ZbyN1 @ Ai
1=yKm
5
K
where i is the summation vector and K a scalar that indicates the num-
ber of sectors.
The second measure is Source of intermediate inputs in neighbouring
municipalities, the procedure is the following: i) we post-multiply the
vector of total turnover of the municipality of reference (ym) by the
technical coefcient matrix to obtain the potential demand for interme-
diate inputs from rms in the municipality (Z by N 1); ii) we divide
each element of this vector by the vector of total sectoral turnover in
neighbouring municipalities (yn) and take the average across sectors
(weighted by turnover). This measure proxies the potential share of in-
puts that can be retrieved by rms in the municipality of reference from
rms that operate in neighbouring municipalities. The measure shows
the potential drop in demand for intermediates experienced by
neighbouring municipalities in the case a production interruption in
the municipality of reference. More formally, the measure is given by:
0 1
1=y1n
ym Z byN1 @ A
0 0
1=yKn
6
K
We provide an application to the Italian municipalities of all the
methodologies proposed in this section.
2.3. Data sources
The main source of information is the ASIA database of Istat
(Archivio Statistico delle Imprese Attive). ASIA - Imprese includes in-
formation on the population of Italian companies since 1996. The total
number of rms in 2011 is 4,515,691 and they were distributed across
different classes of turnover as described in Table 1.
This information is characterized by two main limitations for the aim
of providing an appropriate estimate of local turnover. First, the true
value of turnover for each rm is unknown as only the turnover class
of the rm is reported in ASIA. Second, many rms produce a substan-
tial share of their turnover in local units other than the headquarter (see
Section 2.1). In fact, in single-unit rms all production occurs in the
headquarter, while for multi-unit rms (e.g. multi-plant) a potentially
1
The matrix A that describes the amount of euro of goods that are required from each
sector j to produce one euro of goods for sector i. More formally, being Z the matrix of
monetary intersectoral transaction of intermediate inputs and y the vector of total output,
A is dened as Z byN 1 where byN is the a diagonal matrix of output.
 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766 61

Table 1 level as well as broken down in its main components: buildings and ma-
Information by turnover class. chinery) by sector (37 sectors, NACE rev 2) at the national level (Istat).
Class Min () Max () # Firms Average number Maximum
of local units number of local 3. Results
per rms units per rms

1 0 19,999 1,042,223 1.02 239 In this section, we discuss the estimates of the measures illustrated
2 20,000 49,999 1,161,047 1.01 28 in Section 2 to evaluate the exposure to natural disasters of socio-eco-
3 50,000 99,999 788,529 1.02 21
nomic activities in 8056 Italian municipalities.
4 100,000 199,999 603,462 1.05 37
5 200,000 499,999 464,785 1.11 65
6 500,000 999,999 198,311 1.19 146 3.1. Maps of direct socio-economic exposure
7 1,000,000 1,999,999 120,353 1.29 173
8 2,000,000 3,999,999 66,543 1.43 190 As mentioned above the elements that are exposed to natural disas-
9 4,000,000 4,999,999 13,791 1.59 133
10 5,000,000 9,999,999 28,355 1.84 380
ters are the human beings and their livelihoods and assets. Population
11 10,000,000 19,999,999 14,312 2.42 838 and employment can be useful indicators to evaluate the human beings
12 20,000,000 49,999,999 8487 3.41 546 exposition to natural disasters. In fact, population density based on
13 50,000,000 199,999,999 4230 7.71 1599 place of residence is a proxy for potential life loss across different
14 200,000,000 1263 42.42 12,392
types of hazards (Dilley, 2005). However, because of commuting for
work reasons, we also use employment density as an indirect measure
of how exposed is a municipality during its working hours.
large share of production occurs in units other than the headquarter. The data used for these measures are very detailed and are provided
However, it is important to know where the production actually occurs by the Italian National Statistics ofce (ISTAT). ISTAT annually estimates
for measuring the exposure of a rm's production to natural disasters. these variables at municipality level while in Census years (every ten
In order to attribute the total turnover of the rm to its various years) ISTAT provides very detailed data also at sub-municipal level
branches (local units) we match ASIA Imprese with ASIA Local (i.e. census area). The data are then very detailed and can be easily
Units, which contains information on the population of local units of disaggregated at sub-municipal level with low effort and mild
Italian rms from 2004 to 2012. Table 1 shows some descriptive statis- assumptions.
tics on the number of local units per rm. As mentioned previously, at- Turnover is a good proxy for evaluating the livelihoods as it provides
tributing total rm production to the location of the headquarter may a measure of the gross economic value produced by an area in one year
result in a substantial over-estimation of the turnover generated in and it might be considered as a synthetic value of the economic capacity
the municipality of the headquarter and in an underestimation of the of a territory. Even the data used to assess the turnover at municipal
turnover generated in the municipalities of the local units. Indeed, head- level, as well those for population and employment, are very detailed
quarters are more likely to locate in big urban areas (e.g. Milan, Rome). and might be disaggregated at sub-municipal level, allowing for a great-
Therefore, the use of rm-level data only would result in a systematic er usefulness of the applied methodologies.
over-estimation of turnover (but also employment and capital stock) Finally, information on estimated net capital stock is used to evalu-
in big urban areas. ate the physical capital exposed to natural disasters. Using national sec-
Finally, to retrieve more detailed information about turnover (be- tor-specic information about capital/labour ratios from national
yond simple turnover classes), we use data on true turnover at rm- accounts (ISTAT), we exploit the detailed information on employment
level from the AIDA database (Bureau van Dijk). This information is distribution by municipality-sector to estimate the monetary value of
available only for 733,458 rms (16.24% of total). However, these the capital stock at the municipality level.
rms are responsible for 41.77% of total employees as AIDA over-repre- Fig. 1 shows the exposure maps for population density (Fig. 1a), em-
sents large rms. This information is particularly important because de- ployee density (Fig. 1b), turnover (Fig. 1c) and net capital stock (Fig. 1d
tailed (and true) knowledge on these large players may limit the risk of 1f). As a rst result, it is important to notice the differences between
systematically over-estimating or under-estimating turnover. Table 2 Northern and Southern Italy. Northern Italy shows higher exposure in
reports the estimated turnover as well as the total turnover of rms terms of all the elements considered (e.g. human beings, livelihoods
for which actual turnover was available in the AIDA database. and assets). Southern Italy instead, is characterized by high exposure
To estimate the value of the capital stock owned by economic activ- in population density but low economic exposure. This evidence is in
ities, we retrieve information on net stock of capital (at the aggregate line with the very well known gap between the two Italian macro-
regions.
As a consequence, this evidence might shed light on the different
damage impact of natural disasters between Northern and Southern
Table 2
Italy. Indeed, it is possible to argue that, ceteris paribus, the economic
Estimated turnover by turnover class (billion euro).
damage suffered by the southern regions should be lower than that suf-
Class Total turnover Share of total Total turnover Share of turnover fered by northern ones. At the same time, Northern regions will be able
(rms in AIDA) in rms in AIDA
to recover rstly because of their economic capacity. This evidence is
1 18.53 0.58% 0.32 1.74% also supported by Barone and Mocetti (2014) that show evidence of a
2 40.40 1.26% 1.48 3.66% different recovery pattern of two Italian regions (Friuli Venezia Giulia,
3 49.80 1.55% 3.87 7.77%
4 73.96 2.30% 10.85 14.67%
in the north, and Campania in the south) affected by two earthquakes
5 124.68 3.88% 38.85 31.16% that are comparable in terms of strength of impact.
6 124.24 3.87% 57.77 46.50% If we consider the direct exposure variables singularly we are able to
7 151.75 4.73% 88.19 58.12% underline other important differences in the exposure of the Italian mu-
8 170.69 5.31% 116.21 68.08%
nicipalities. Population density is reported in Fig. 1 (panel a). As expect-
9 59.46 1.85% 36.47 61.34%
10 186.71 5.81% 145.18 77.76% ed, population density is greater in urban areas (e.g. Napoli, Milano and
11 190.28 5.92% 152.43 80.11% the corresponding adjacent municipalities) and lower in mountainous
12 243.32 7.58% 184.87 75.98% areas. The rst six municipalities in terms of population density all be-
13 355.65 11.07% 237.03 66.65% long to the Napoli province (see Supplementary material, see Tables
14 1422.13 44.28% 602.79 42.39%
A1 and A2). Employees per square kilometre (Fig. 1, panel b) seem to
62 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
Fig. 1. Direct socio-economic exposure for (a) population; (b) employees; (c) turnover; (d) net capital stock; (e) buildings and (f) machineries (per km2 in 2011).
be more concentrated in big cities. At the same time, municipalities that
surround big cities appear now less dense than when considering pop-
ulation. This reects the fact that many people commute from sub-
urban municipalities to the urban centre to work. Looking at the rst
ten municipalities, we now observe a predominance of municipalities
that belong to the province of Milano (5).2
The underlined spatial concentration of these variables may produce
intriguing policy conclusion in terms for instance of mitigation and pre-
vention policies or for land use planning. First of all, when dening mit-
igation policies, policy makers should keep in mind the distribution of
people over space (e.g. concentration of people in risky prone areas)
but also the commuting between municipalities as to dene targeted
mitigation policies weighted for the concentration of people that live
or work in a certain area. Second, Fig. 1a and b might be incorporated
in emergency risk plan as to properly consider the number of exposed
human beings as a sum of resident and commuters.
To capture the exposure of economic activities, we instead focus on
turnover and capital stock of rms. Turnover intensity is higher in
Northern Italy than in the Southern part of the country (Fig. 1, panel
c). Lombardy and, more specically, the metropolitan area of Milano,
is the area with the highest intensity in turnover. Municipalities on
the Alps and Appennini, on the other hand, tend to show rather small
turnover intensity. Moving to total physical capital stock (Fig. 1, panels
2
The small municipality of Fiera di Primiero, in the province of Trento, in the Alps, ranks
second in terms of employees per square kilometre. This is an outlier as it is the second
smallest municipality (in terms of area) in Italy, with a size of just 0.15 km2. Also the third
municipality in terms of employees per square kilometre (Casavatore, near to Naples),
might be considered an outlier because it is the most densely populated Italian
municipality.
d, e, f), we observe that its value is rather high also for medium-sized
urban centres in the northern part of Italy (Padova, Bologna, Bergamo,
Brescia, Genova) as well as municipalities in the northern part of the
Adriatic coast. Among the top 10, municipalities in the province of Mila-
no still lead (5).3 As a large share of the total capital stock is composed
by buildings, the distribution of our measure of capital stock in buildings
is very similar to the one of total capital. The stock of capital in machin-
ery is mostly concentrated in municipalities which are specialised in in-
dustrial production (panel f). Again, 6 municipalities of the top 10 are
located in the province of Milano. The wide-spread distribution of the
physical capital stock even in medium-sized urban centres might lead
to an increase of the difculties in implementing mitigation and preven-
tion plans due to a greater portion of territory that is exposed to poten-
tial risks, especially so for small municipalities that cannot have the
nancial resources to face the risk management in vast areas. Coordina-
tion between neighbouring municipalities can be useful to adopt more
appropriate measures against natural disaster risks.
To sum up, descriptive evidence highlights a rather heterogeneous
exposure of municipalities to natural disasters. However, the correlation
between the various measures is very high and exposure according to
different metrics tends to overlap to a great extent. Table 3 reports the
correlation matrix between our measures of exposure to natural disas-
ters. Correlation between capital stock (and its components) and em-
ployment however is far from perfect, despite the former is estimated
3
For capital stock per square kilometre, whose estimate depends on the number of em-
ployees working in a municipality, the small municipality of Fiera di Primiero ranks rst.
This is driven by the very high intensity of workers employed in local units that reside
in the municipality and by the relatively capital intensive sectoral composition of the
economy in the municipality.
 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766 63

Table 3
Correlation matrix.

Population Employees Turnover Net capital stock Net capital stock - machinery Net capital stock - buildings

Population 1
Employees 0.8367 1
Turnover 0.6032 0.8269 1
Net capital stock 0.7123 0.9222 0.7581 1
Net capital stock - machinery 0.7571 0.9612 0.8385 0.8906 1
Net capital stock - buildings 0.6895 0.8946 0.7273 0.9968 0.8522 1

by using information of the latter. This means that heterogeneity in cap-
ital intensity across sectors and differences in sectoral composition
across municipalities is substantial and results in rather heterogeneous
estimates of exposure of productive capital stock.
It should be noticed that all the measures listed in this section are per
se able to provide information on the different components of economic
value that are directly exposed to natural disasters. However, in order to
account for the indirect effects induced by natural disasters in a given
area, we have to account also for the dependency of each industry in a
given municipality from all other industries in neighbouring municipal-
ities: i) to retrieve intermediate inputs, and ii) as a market for nal puted as:
goods and service.
3.2. Maps of indirect socio-economic exposure
(population, employees and net capital stock), while turnover appears
to be more randomly distributed across municipalities.
This evidence claims for further considerations on the way in which
exposure is spatially distributed. Clustering is particularly useful as it al-
lows to easily identify possible hot-spots or critical areas in terms of ex-
posure to natural disasters.
To this purpose, we compute the Local Indicators of Spatial Associa-
tions (LISA, see Anselin, 1995). We rst compute the local Moran's I
(based on the 5-neighbours weighting matrix) and identify clusters of
municipalities and outliers. The local Moran's I for municipality i is com-
 
j wij xi x x j x N
Ii 8
i xi x
2 j wij

In order to have a rst assessment of the indirect component of ex- Whenever the local Moran's I is not statistically different from zero
posure we evaluate some geographical features of the measures of ex- the municipality does not belong to a cluster and it is not an outlier.
posure. We make use of spatial autocorrelation and clustering as an On the other hand, four outcomes are possible when the local Moran's
implicit way of assessing indirect exposure to risk. Indeed, the adjacency I is statistically signicant:
of municipalities with similar direct exposure may implicitly indicate
some sort of economic linkage between them. As a preliminary analysis, the municipality belongs to a cluster with high values (HH);
we estimate the spatial autocorrelation for all variables of interest the municipality belongs to a cluster with low values (LL);
(Moran's I, see Moran, 1950) using as geographical weights either the the municipality is an outlier with high value surrounded by munici-
contiguity matrix which considers only direct neighbours or the conti- palities with low value (HL);
guity matrix that considers up to the fth group of neighbours. The the municipality is an outlier with low value surrounded by munici-
Moran's I is computed as: palities with high value (LH).
 
i j wij xi x x j x N If a cluster HH experiences a natural disaster, that will generate
I 7
i xi x
2 i j wij large damages no matter which specic municipality is directly affected.
This happens because many municipalities in the cluster have high
where wij represents spatial weights and xi and xj the value of the mea- values of exposure. On the contrary, LL clusters will experience rather
sure of interest for municipalities i and j. This measure reects how cor- low expected damages as economic activities are not concentrated in
related are some specic features (risk exposure, in our cases) across any of the municipalities of the clusters. Finally, HL (as well as LH) mu-
adjacent locations at national level. A positive value would indicate nicipalities will be characterized by a very high expected damage that is
that similar locations tend to be contiguous, while a negative value concentrated only in one part of the region and the expected damages of
would mean that highly exposed locations tend to be systematically the events should be carefully analyzed according to the affected area. It
closer to locations with low exposure. should be notice that these measures might provide a rst evidence of
Results are reported in Table 4. Spatial autocorrelation is rather high the propagation of the shock in neighbouring municipalities, regardless
for population density, employee density and capital stock (total and its of the probability for a specic region of experiencing a natural disaster.
various components) per square kilometre, while it is rather low for Municipalities in risky areas should be relatively more interested in the
turnover. Spatial correlation is much greater when using one-neigh- evaluation of direct effects of disasters. At the same time, municipalities
bours spatial weights than when using ve-neighbours. This evidence that lie in non-risky areas need to consider whether they belong primar-
is particularly interesting because it indicates a rather similar exposure ily to HH or LH cluster in which at least one of the near municipalities
of contiguous municipalities for most of the selected measures lies in a risky area. In this case, the propagation of the economic damage
can even reach non-risky areas.
Table 4
Results for population, employment, turnover, total capital stock,
Spatial autocorrelation (Moran's I). capital stock in buildings and capital stock in machinery are shown in
Fig. 2. For what concerns population (panel a), we observe HH clusters
Variable Moran's I Moran's I
in medium-large urban areas (i.e. Milan, Rome) while LL clusters
1 neighbour 5 neighbours
emerge in the mountains of Alps and Appennini. Few HL outliers also
Population 0.728 0.266
exist in correspondence of medium-size cities. LH outliers are very un-
Employees 0.695 0.236
Turnover 0.087 0.040 common. Evidence for employment density (panel b) is very similar
Net capital stock total 0.540 0.169 to what is found for population while for turnover (panel c) we observe
Net capital stock buildings 0.505 0.152 a much smaller number of HH clusters. Finally, the various measures of
Net capital stock machinery 0.638 0.245 capital stock (panels df) tend to reect to a great extent what we found
Net capital stock intangible 0.571 0.170
for employment.
64 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
Fig. 2. Clustering using Local Moran's I for (a) population; (b) employees; (c) turnover; (d) net capital stock; (e) buildings and (f) machineries (per km2 in 2011).
As we mention above, it is important to account for differences in the
clustering between turnover and the other exposure measures. While
population density, employee density and net capital stock follow a sim-
ilar pattern, turnover is spatially clustered around important hot-spots
like Milan and Turin in the north of Italy. Then it results that the polar-
ization of economic activities in precise areas needs to be carefully
planned in order to avoid and prevent economic damages over large ter-
ritories. For this reason, more explicit measures of indirect exposure are
needed. Indeed, agglomeration forces that enable to take advantage of
geographically concentrated economic linkages might drive proximity
and concentration.
We show results for Destination of nal output in neighbouring mu-
nicipalities in Fig. 3 (panels ac) and for Source of intermediate inputs
in neighbouring municipalities in Fig. 3 (panels df). To ease the graphi-
cal representation, we rescaled the indicators to be in the range 01
(darker shade indicates greater value of the indicator). It is interesting
to note that at the top of the list of municipalities that are destination
of nal output we have municipalities that host large industrial facilities
(i.e. the steel districts of Piombino and Taranto) or large cities (e.g.
Rome, Milan).
On the same line, the list of municipalities that are source of interme-
diate inputs is quite the same as before at the very top (i.e. industry-
based municipalities and large cities). However, the correlation with
the previously discussed measure is positive but not very large (0.45),
thus conrming that these measures depict different dimensions of in-
direct exposure.
Results should be interpreted as follows: the overall density of activ-
ities (of any sector) in large cities generates a large demand for
intermediate inputs, which usually is larger than the output produced
in neighbouring municipalities. On the other hand, manufacturing sec-
tors are characterized, on average, by much larger upstream multi-
pliers (or, in our case, technical coefcients) than service sectors as
they require a large amount of intermediate inputs to carry out produc-
tion activities. This generates a large demand for intermediates that
often cannot be completely met by neighbouring municipalities, leading
to high values for our indicators for industry-based municipalities.
Not all sectors are likely to be inuenced by natural disasters in the
same way. On the one hand, manufacturing sectors are generally more
capital intensive than service sectors. In case the natural disaster has rel-
evant consequences in terms of destruction of physical capital goods,
capital intensive sectors are more likely to be forced to interrupt their
production activities. On the other hand, given that manufacturing sec-
tors are tradable sectors (i.e. they are exposed to foreign competition),
there is the risk that temporary interruptions in production due to the di-
saster induce a permanent change in the structure of the supply chain in
favour of manufacturing rms operating in other areas. For these rea-
sons, we also provide evidence for manufacturing-to-manufacturing
linkages as well as manufacturing - to (from) - all sectors linkages.
These measures are reported in panels b, c, e and f of Fig. 3.
What is interesting to observe is that these measures positively cor-
relate with each other even though this correlation is usually far from
perfect (see Table A3 in the Supplementary material).
As a further implication, it should be important to underline that,
while providing a full discussion of the results for these alternative mea-
sures goes beyond the aim of this paper, the maps provided in Fig. 3 are
able to dene municipalities that are hot-spots in terms of both source
 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
Fig. 3. Measures of indirect economic exposure.
and destination of goods. Natural disasters affecting these municipali-
ties are then able to produce economic damages to areas that might
be unaffected by the extreme event.
4. Discussion and conclusions
In recent years, public and private sector have put always-greater ef-
fort in implementing actions aimed at preventing the occurrence of nat-
ural disasters, at mitigating the damages and at adapting to increasing
risks. In this study we provide evidence that the effect of and the dam-
ages produced by a natural disaster may propagate between local areas
because of the economic inter-connections of the productive systems.
However, very often local administrators are unconscious about the po-
tential i) direct damage that local areas are going to suffer when hit by a
natural disaster; ii) indirect damage that results from a natural disaster
that affected another (even distant) area.
Very little effort has been devoted until now to addressing the issue
of mapping the exposure of socio-economic values to natural disasters.
This study aims to ll this gap. We have provided a rst attempt to de-
ne suitable indicators able to capture the different aspects of the socio-
economic exposure of territories. We also have used administrative and
statistical data to estimate a set of economic information that can be
employed to map the potential economic exposure of narrowly dened
geographical units (e.g. municipalities) in terms of direct and indirect
socio-economic consequences of natural disasters. These types of
maps can be integrated into existing and newly developed models for
risk assessment or as benchmark for urban planning, risk mitigation
policies and risk prevention.
Besides providing estimates of direct socio-economic exposure to
natural disasters, we also developed different measures of indirect
65
socio-economic exposure. These indicators are particularly useful to
quantify the overall potential economic impact of natural disasters on
specic areas, also beyond the boundaries of the location that is directly
affected by natural disasters. Our results show that even though all indi-
cators (direct and indirect exposure) are positively correlated, these
correlations are far from perfect and suggest that differences in the eco-
nomic structure of municipalities may result in different transmission of
shocks to neighbouring municipalities as a consequence of natural
disasters.
Then, the spatial concentration of the variables that are underlined
in this study (both in direct and indirect terms) may produce relevant
policy conclusions for land use planning but also in order to dene
more precise and targeted mitigation and prevention strategies. Indeed,
local governments should have a correct knowledge about: i) the distri-
bution of people over space (e.g. concentration of people in risky prone
areas), ii) the commuting patterns across municipalities (e.g. employees
that commute between municipalities), iii) the spatial distribution of
assets (e.g. the net capital stock) and iv) the inter-connection between
productive systems (destination and source of goods). This is important
for being able to dene targeted mitigation policies that account for the
concentration of the underlined socio-economic variables in certain
areas.
Finally, the approach developed in this study as well as the type of
data that are employed may be particularly useful for academic and pol-
icy-relevant evaluations of risk assessment and damage cost evaluation
of natural disasters, from both an ex-ante and ex-post perspective. The
indicators we propose could be further rened to account for the spec-
icities of the case study for which they need to be employed focusing
on: i) narrowly dened areas (i.e. smaller than the administrative
boundaries of a municipality); ii) distances between the municipality
66 G. Marin, M. ModicaEnvironmental Impact Assessment Review 64 (2017) 5766
of reference and the economic activities that could potentially suffer
from indirect damages; iii) reduced number of sectors due to their
higher or lower vulnerability to direct or indirect damages.
Acknowledgment
The authors thank to the anonymous referees and the editor for the
constructive comments and recommendations. Authors acknowledge
nancial support from the research project La valutazione economica
dei disastri economici in Italia funded by the Fondazione Assicurazioni
Generali. Usual disclaimer applies.
Appendix A. Supplementary data.
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.eiar.2017.03.002.
References
Acemoglu, D., Carvalho, V.M., Ozdaglar, A., Tahbaz-Salehi, A., 2012. The network origins of
aggregate uctuations. Econometrica 80 (5), 19772016.
Anselin, L., 1995. Local indicators of spatial association LISA. Geogr. Anal. 27 (2), 93115.
Barone, G., Mocetti, S., 2014. Natural disasters, growth and institutions: a tale of two
earthquakes. J. Urban Econ. 84, 5266.
Blaikie, P., Cannon, T., Davis, I., Wisner, B., 2014. At Risk: Natural Hazards, People's Vulner-
ability and Disasters. Routledge.
Bloechl, A., Braun, B., 2005. Economic assessment of landslide risks in the Swabian Alb,
Germanyresearch framework and rst results of homeowners' and experts' surveys.
Nat. Hazards Earth Syst. Sci. 5 (3), 389396.
Cardona, O.D., van Aalst, M.K., Birkmann, J., Fordham, M., McGregor, G., Perez, R.,
Pulwarty, R.S., Schipper, E.L.F., Sinh, B.T., 2012. Determinants of risk: exposure and
vulnerability. In: Field, C.B., Barros, V., Stocker, T.F., Qin, D., Dokken, D.J., Ebi, K.L.,
Mastrandrea, M.D., Mach, K.J., Plattner, G.-K., Allen, S.K., Tignor, M., Midgley, P.M.
(Eds.), Managing the Risks of Extreme Events and Disasters to Advance Climate
Change Adaptation. A Special Report of Working Groups I and II of the Intergovern-
mental Panel on Climate Change (IPCC). Cambridge University Press, Cambridge,
UK, and New York, NY, USA, pp. 65108.
Chen, Q.F., Chen, Y., Liu, Jie, Chen, L., 1997. Quick and approximate estimation of earth-
quake loss based on macroscopic index of exposure and population distribution.
Nat. Hazards 15 (23), 215229.
Cutter, S.L., Finch, C., 2008. Temporal and spatial changes in social vulnerability to natural
hazards. Proc. Natl. Acad. Sci. 105, 23012306.
Dilley, M., 2005. Natural Disaster Hotspots: A Global Risk Analysis. Vol. 5. World Bank
Publications.
ECLAC, 2003. Handbook for Estimating the Socio-Economic and Environmental Effects of
Disasters. United Nations Economic Commission for Latin America and the Caribbean.
FEMA (1992). Indirect Economic Consequences of a Catastrophic Earthquake, edited:
Milliman, J.W. and Sanguinetty, J.A., (National Earthquake Hazards Reduction
Program).
Field, E.H., Seligson, H.A., Gupta, N., Gupta, V., Jordan, T.H., Campbell, K.W., 2005. Loss es-
timates for a Puente Hills blind-thrust earthquake in Los Angeles, California. Earth-
quake Spectra 21 (2), 329338.
Hallegatte, S., Przylusky, V., 2010. The economics of natural disasters. Concepts and
Methods. Policy Research Working Paper 5507. World Bank. World Bank.
Jongman, B., Kreibich, H., Apel, H., Barredo, J.I., Bates, P.D., Feyen, L., Ward, P.J., 2012. Com-
parative ood damage model assessment: towards a European approach. Nat. Haz-
ards Earth Syst. Sci. 12 (12), 37333752.
Koks, E.E., de Moel, H., Aerts, J.C., Bouwer, L.M., 2014. Effect of spatial adaptation measures
on ood risk: study of coastal oods in Belgium. Reg. Environ. Chang. 14 (1),
413425.
Koks, E.E., Jongman, B., Husby, T.G., Botzen, W.J.W., 2015. Combining hazard, exposure
and social vulnerability to provide lessons for ood risk management. Environ. Sci.
Pol. 47, 4252.
Luino, F., Cirio, C.G., Biddoccu, M., Agangi, A., Giulietto, W., Godone, F., Nigrelli, G., 2009.
Application of a model to the evaluation of ood damage. GeoInformatica 13 (3),
339353.
Meroni, F., Pessina, V., Squarcina, T., Locati, M., Modica, M., Zoboli, R., 2016. The economic
assessment of seismic damage: an example for the 2012 event in Northern Italy. In:
CERU European Centre on Urban Risks (Ed.), International Conference on Urban
Risk. ISBN: 978-989-95094-1-2, pp. 323330 (Lisbon (Portugal)).
Miller, R.E., Blair, P.D., 2009. Input-Output Analysis: Foundations and Extensions. Cam-
bridge University Press.
Modica, M., Reggiani, A., 2015. Spatial economic resilience: overview and perspectives.
Networks and Spatial Economics 15 (2), 211233.
Modica, M., Zoboli, R., 2016. Vulnerability, resilience, hazard, risk, damage, and loss: a
socio-ecological framework for natural disaster analysis. Web Ecology 16, 5962.
Modica, M., Reggiani, A. and Nijkamp (2017). Vulnerability, resilience and exposure:
methodological aspects and empirical applications to the german labour market
(forthcoming).
Moran, P.A.P., 1950. Notes on continuous stochastic phenomena. Biometrika 37 (1),
1723.
Okuyama, Y., 2007. Economic modeling for disaster impact analysis: past, present, and fu-
ture. Econ. Syst. Res. 19 (2), 115124.
Okuyama, Y., Santos, J.R., 2014. Disaster impact and inputoutput analysis. Econ. Syst. Res.
26 (1), 112.
sth, J., Lyhagen, J., Reggiani, A., 2016. A new way of determining distance decay param-
eters in spatial interaction models with application to job accessibility analysis in
Sweden. Eur. J. Transp. Infrastruct. Res. 16 (2), 344363.
Pellicani, R., Van Westen, C.J., Spilotro, G., 2014. Assessing landslide exposure in areas
with limited landslide information. Landslides 11 (3), 463480.
Pelling, M., 2003. The Vulnerability of Cities: Natural Disasters and Social Resilience.
Earthscan, London.
Pelling, M., Ozerdem, A., Barakat, S., 2002. The macro-economic impact of disasters. Prog.
Dev. Stud. 2, 283305.
Perminova, T., Laratte, B., Sirina, N., Baranovskaya, N., Rikhvanov, L., 2016. Methods for
land use impact assessment: a review. Environ. Impact Assess. Rev. 60, 6474.
Rose, A., 2004. Economic principles, issues, and research priorities in hazard loss estima-
tion. In: Okuyama, Y., Chang, S.E. (Eds.), Modeling Spatial and Economic Impacts of
Disasters. Springer, NewYork, pp. 1336.
Rose, A., Lim, D., 2002. Business interruption losses from natural hazards: conceptual and
methodological issues in the case of the Northridge earthquake. Environmental Haz-
ards 4 (1), 114.
Rose, A., Wei, D., 2013. Estimating the economic consequences of a port shutdown: the
special role of resilience. Econ. Syst. Res. 25 (2), 212232.
Rose, A., Oladosu, G., Lia, S.Y., 2007. Business interruption impacts of a terrorist attack on
the electric power system of Los Angeles: customer resilience to a total blackout. Risk
Anal. 27 (3), 513531.
Simelton, E., Fraser, E.D., Termansen, M., Forster, P.M., Dougill, A.J., 2009. Typologies of
crop-drought vulnerability: an empirical analysis of the socio-economic factors that
inuence the sensitivity and resilience to drought of three major food crops in
China (19612001). Environ. Sci. Pol. 12 (4), 438452.
Sterlacchini, S., Zazzeri, M., Genovese, E., Modica, M., Zoboli, R., 2016. Flood damage in
Italy: towards an assessment model of reconstruction costs. EGU General Assembly
2016, held 1722 April, 2016 in Vienna Austria, p. 13135.
Stewart, M.B., 1983. On least squares estimation when the dependent variable is grouped.
Rev. Econ. Stud. 50 (4), 737753.
Tan, R.R., Aviso, K.B., Promentilla, M.A.B., Solis, F.D.B., Yu, K.D.S., Santos, J.R., 2015. A shock
absorption index for inoperability inputoutput models. Econ. Syst. Res. 27 (1),
4359.
Van Der Veen, A., Logtmeijer, C., 2005. Economic hotspots: visualizing vulnerability to
ooding. Nat. Hazards 36, 6580.
Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel Data. MIT Press.
Zelekov, M., Zvijkov, L., 2017. Risk analysis within environmental impact assess-
ment of proposed construction activity. Environ. Impact Assess. Rev. 62, 7689.
Marco Modica, Ph.D., Post-doctoral Research Fellow at IRCrES-CNR. Main elds of exper-
tise: regional and urban economics, new economic geography, economics of disasters, sus-
tainability and applied econometrics. He has got a PhD in 'Economics, Markets and
Institutions' at IMT (Institute for Advanced Studies Lucca, Italy) in 2013, with a thesis on
regional complex systems. He has been visiting research student at the Maxwell School
of Citizenship and Public Affairs of Syracuse University, Syracuse, NY (US). He is involved
in several joint projects with particular focus on economic evaluation of natural disasters,
coastal management and urban sustainability, with a special attention on the resilience of
socio-economic systems.
Giovanni Marin, Ph.D., Assistant professor of Economics at the Department of Economics,
Society and Politics, University of Urbino 'Carlo Bo' (Italy). Main elds of expertise: statis-
tical, econometric, and input-output models applied to the economic analysis of the ener-
gy and eco-innovation. He has got a PhD in 'Economics, Markets and Institutions' at IMT
(Institute for Advanced Studies Lucca, Italy) in 2012, with a thesis on eco-innovation. He
has been Visiting research scholar at the University of California Berkeley (US) and at
the OFCE-SciencesPo (France).He has been post-doctoral researcher at the IRCrES-CNR
from 2012 to 2016. He is author of papers published in international economics journals
(Journal of Evolutionary Economics, Ecological Economics, Industrial and Corporate
Change, Research Policy, Environmental and Resource Economics).