Content uploaded by Nick Brooks
Author content
All content in this area was uploaded by Nick Brooks on Oct 20, 2017
Content may be subject to copyright.
Content uploaded by Nick Brooks
Author content
All content in this area was uploaded by Nick Brooks
Content may be subject to copyright.
Global Environmental Change 15 (2005) 151–163
The determinants of vulnerability and adaptive capacity at the
national level and the implications for adaptation
Nick Brooks
a,
, W. Neil Adger
a,b
, P. Mick Kelly
c
a
Tyndall Centre for Climate Change Research, University of East Anglia, Norwich NR4 7TJ, UK
b
CSERGE, University of East Anglia, Norwich NR4 7TJ, UK
c
Climatic Research Unit, University of East Anglia, Norwich NR4 7TJ, UK
Accepted 8 December 2004
Abstract
We present a set of indicators of vulnerability and capacity to adapt to climate variability, and by extension climate change,
derived using a novel empirical analysis of data aggregated at the national level on a decadal timescale. The analysis is based on a
conceptual framework in which risk is viewed in terms of outcome, and is a function of physically defined climate hazards and
socially constructed vulnerability. Climate outcomes are represented by mortality from climate-related disasters, using the
emergency events database data set, statistical relationships between mortality and a shortlist of potential proxies for vulnerability
are used to identify key vulnerability indicators. We find that 11 key indicators exhibit a strong relationship with decadally
aggregated mortality associated with climate-related disasters. Validation of indicators, relationships between vulnerability and
adaptive capacity, and the sensitivity of subsequent vulnerability assessments to different sets of weightings are explored using expert
judgement data, collected through a focus group exercise. The data are used to provide a robust assessment of vulnerability to
climate-related mortality at the national level, and represent an entry point to more detailed explorations of vulnerability and
adaptive capacity. They indicate that the most vulnerable nations are those situated in sub-Saharan Africa and those that have
recently experienced conflict. Adaptive capacity—one element of vulnerability—is associated predominantly with governance, civil
and political rights, and literacy.
r2005 Elsevier Ltd. All rights reserved.
Keywords: Vulnerability; Adaptive capacity; Indicators; National-level; Risk; Mortality; Delphi survey; Governance; Literacy; Health
1. Introduction
The purpose of this study is to develop national-level
indicators of vulnerability and capacity to adapt to
climate hazards. The national level is an appropriate
scale for information utilised by central government
in determination of policy. In other policy areas,
the UNDP’s Human Development Index, measures of
Genuine Progress and other indicators provide compar-
able, transparent and meaningful information on
aspects of development, though often fail to capture
environmental sustainability (Neumayer, 2001;Lawn,
2003;Bell and Morse, 1999). Comparing vulnerability
across countries can identify leverage points in reducing
vulnerability to climate variability and, by inference, to
climate change, which is likely to be manifest through
changes in the frequency and severity of existing hazards
at least in the short- to medium-term (Easterling et al.,
2000;Frich et al., 2002). Identification of particularly
vulnerable nations or regions (i.e. those that are least
well equipped to cope with the impacts of climate
change) can act as an entry point for both under-
standing and addressing the processes that cause and
exacerbate vulnerability (Moss et al., 1999;Yohe
and Tol, 2002;Brooks and Adger, 2003;Leichenko
and O’Brien, 2002;O’Brien et al., 2004), although
sub-national spatial and social differentiation of
ARTICLE IN PRESS
www.elsevier.com/locate/gloenvcha
0959-3780/$ - see front matter r2005 Elsevier Ltd. All rights reserved.
doi:10.1016/j.gloenvcha.2004.12.006
Corresponding author. Tel.: +44 1603 593903.
E-mail address: nick.brooks@uea.ac.uk (N. Brooks).
vulnerability, and the way in which the impacts of
national-scale processes are mediated by local condi-
tions, should not be downplayed.
Published studies of national-level vulnerability to
date generally have been characterised by indicators
chosen subjectively by the authors, based on assump-
tions about the factors and processes leading to
vulnerability, informed by literature review and intuitive
understandings of human–environment interaction. The
approach presented here uses an empirical approach to
develop indicators of vulnerability to a range of climate
hazards at the national level. We specifically address
vulnerability to mortality resulting from exposure to
climate hazards, assessed at the national level for
decadal periods. This analysis is carried out using a
conceptual framework in which risk is viewed as a
function of physically defined climate hazard and
socially constructed vulnerability, elaborated in the
following section.
National-level outcome risk is represented in terms of
mortality outcomes from climate-related disasters using
data from the emergency events database (EM-DAT)
data set. Vulnerability is represented by a suite of socio-
economic, political and environmental variables that
represent the sensitivity and exposure of national
populations to climate hazards. Key indicators of
vulnerability are identified by examining the statistical
relationships between a large number of potential
proxies for vulnerability, and measures of mortality
outcome.
The paper also addresses validation of the indicators
through expert elicitation, and an assessment of the
sensitivity of country rankings to different sets of
weightings, based on subjective weightings generated
by a focus group. Finally, a set of ‘‘most vulnerable’’
countries is identified via an assessment of country
rankings across a number of composite indicators, each
constructed using a different set of weightings, and the
implications of the study for adaptation discussed.
2. Conceptual framework: risk, vulnerability and
adaptive capacity
2.1. Defining the relationship between vulnerability and
risk
Definitions of risk are commonly probabilistic in
nature, relating either to (i) the probability of occurrence
of a hazard that acts to trigger a disaster or series of
events with an undesirable outcome, or (ii) the prob-
ability of a disaster or outcome, combining the
probability of the hazard event with a consideration of
the likely consequences of the hazard (Smith, 1996;
Stenchion, 1997;Downing et al., 2001;Brooks, 2003;
Jones and Boer, 2003). In this study, risk is conceptua-
lised as relating to compound ‘‘climate-related disas-
ters’’, triggered by climatic or meteorological hazards
(storms, droughts, extreme precipitation events, circula-
tion changes and so on) but mediated by the sensitivity
or vulnerability of the exposed systems. We therefore
view risk as a function of hazard and vulnerability
(UNDHA, 1992), a definition which is compatible
with that of risk as the product of probability and
consequence (Smith, 1996;Brooks, 2003;Jones and
Boer, 2003). In this study, conducted at a global scale
and focusing on the national level, the exposed systems
are individual countries, and we are concerned with
mortality risk.
As an alternative to the probabilistic approach, we
may use quantitative measures of outcome as proxies for
risk, particularly where we are concerned with historical
data. Probabilistic and outcome-based measures repre-
sent alternative but complementary ways of approach-
ing risk assessment. In particular, we may use data
relating to adverse socio-economic impacts as a retro-
spective measure of historical risk, representing out-
comes arising from the interaction of hazard and
vulnerability (UNDHA, 1992;Brooks, 2003). Here we
use numbers of people killed by climate-related disasters
per decade, expressed as a percentage of national
population, as a proxy for climate mortality risk at the
national level. We then use the relationship
‘‘risk ¼hazard vulnerability’’ to assess vulnerability
using the risk proxies in conjunction with a variety of
socio-economic and other data.
2.2. Factors influencing vulnerability
The IPCC Third Assessment Report (TAR) contains
two conflicting definitions of vulnerability. The glossary
of the TAR (IPCC, 2001, p. 995) defines vulnerability as
‘‘The degree to which a system is susceptible to, or
unable to cope with, adverse effects of climate change,
including climate variability and extremes. Vulnerability
is a function of the character, magnitude, and rate of
climate variation to which a system is exposed, its
sensitivity, and its adaptive capacity.’’ However, Smit et
al. (2001) in the IPCC TAR, citing Smit et al. (1999),
describe vulnerability as the ‘‘degree to which a system is
susceptible to injury, damage, or harm (one part—the
problematic or detrimental part—of sensitivity)’’. Sensi-
tivity in turn is described as the ‘‘degree to which a
system is affected by or responsive to climate stimuli’’
(IPCC, 2001, p. 894). Here we use the latter definition, in
which vulnerability is essentially a state variable,
determined by the internal properties of a system; for
social systems, we are considering what may be referred
to as social vulnerability (Brooks, 2003;Adger, 1999;
Adger and Kelly, 1999;Kelly and Adger, 2000).
Vulnerability depends critically on context, and the
factors that make a system vulnerable to a hazard will
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163152
depend on the nature of the system and the type of
hazard in question. The factors that make a rural
community in semi-arid Africa vulnerable to drought
will not be identical to those that make areas of a
wealthy industrialised nation such as Norway vulnerable
to flooding, wind storms and other extreme weather
events. Isolation and income diversity might be im-
portant determinants of vulnerability to drought for
rural communities in Africa, whereas the dominant
factors mediating vulnerability to storms and floods in
Norway might be the quality of physical infrastructure
and the efficacy of land use planning. Nonetheless, there
are certain factors that are likely to influence vulner-
ability to a wide variety of hazards in different
geographical and socio-political contexts. These are
developmental factors including poverty, health status,
economic inequality and elements of governance, to
name but a few. These may be referred to as generic
determinants of vulnerability, as opposed to specific
determinants relevant to a particular context and hazard
type, such as the price of a particular food crop, the
number of storm shelters available for the use of a
coastal community, or the existence of regulations
concerning the robustness of buildings. Although the
relative importance of different generic factors will
exhibit some variation, such factors may be viewed as
the foundation on which specific measures for reducing
vulnerability and facilitation adaptation are built. For
example, a rural community is more likely to be serviced
by transport infrastructure if it is effectively represented
at the political level. Building codes are more likely to be
enforced if corruption in the building industry and
regulatory agencies is minimised.
The concept of generic, as opposed to hazard- and
context-specific, determinants of vulnerability, is a
useful one if we wish to undertake comparative
assessments of vulnerability at the national level. While
developmental and governance indicators do not repre-
sent a complete description of vulnerability, and while
vulnerability will exhibit substantial sub-national geo-
graphical and social differentiation, assessments of
‘‘generic’’ vulnerability can tell us how well equipped a
country is to cope with and adapt to climate hazards.
The aim of the study presented here is to identify key
indicators of generic vulnerability from general devel-
opmental data representing the national level, using the
risk–hazard–vulnerability framework outlined above, in
a global analysis.
2.3. Relationships between vulnerability and adaptive
capacity
Adaptive capacity is defined in the glossary of the
IPCC (2001, p. 982) TAR as ‘‘The ability of a system to
adjust to climate change (including climate variability
and extremes), to moderate potential damages, to take
advantage of opportunities, or to cope with the
consequences.’’ Because adaptation does not occur
instantaneously, the relationship between adaptive capa-
city and vulnerability depends crucially on the timescales
and hazards with which we are concerned. The vulner-
ability, or potential vulnerability, of a system to climate
change that is associated with anticipated hazards in the
medium- to long-term will depend on that system’s ability
to adapt appropriately in anticipation of those hazards.
However, vulnerability to hazards associated with climate
variability that may occur in the immediate future will be
related to a system’s existing short-term coping capacity
rather than its ability to pursue long-term adaptation
strategies. For a more detailed discussion of the relation-
ship between vulnerability and adaptive capacity within
the context of different types of climate-related hazard
the reader is referred to Brooks (2003).
3. Data and methods
3.1. Proxies for risk
The study presented here specifically addresses
mortality risk associated with climate-related disasters
and, by extension, deals with vulnerability to death
associated with climate hazards. It is therefore not a
comprehensive assessment of all the dimensions of
vulnerability, focusing instead on the most important
dimension, and the one for which relatively reliable data
are most readily available.
Mortality risk is represented by indicators constructed
from the EM-DAT, developed by the US Office of
Foreign Disaster Assistance (OFDA) and the Centre for
Research into the Epidemiology of Disasters (CRED) at
the Universite
´Catholique de Louvain in Brussels,
Belgium (http://www.cred.be/emdat). The version of
the data set used here records ‘‘natural’’ disasters,
defined as events associated with 10 or more people
reported killed, 100 or more people affected, a call for
international assistance, or the declaration of a state of
emergency. The data set was processed in order to
remove events without a climatic component, leaving a
version representing climate-related disasters only. The
identification of climate-related events is discussed in
Brooks and Adger (2003). The total numbers of people
killed by climate-related disasters for each of the three
final decades of the twentieth century were then
calculated for each country represented in the database,
and the results scaled by national population in order to
create the risk indicators.
3.2. Vulnerability proxies: initial shortlist
On the basis of previous work and our own expert
judgement, we identified a shortlist of 46 variables
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 153
representing generic vulnerability, representing economic
well-being and inequality, health and nutritional status,
education, physical infrastructure, governance, geo-
graphic and demographic factors, agriculture, ecosystems
and technological capacity. Proxy data representing each
variable were acquired from a variety of sources,
including the World Bank, UNDP (Human Development
Index), UNEP and CIESIN. Governance data from the
data set of Kaufmann, Kray and Zoido-Lobaton were
used, hereafter referred to as the KKZ data set
(Kaufmann et al., 1999a, b); these were augmented by
data on civil and political rights from Freedom House
(www.freedomhouse.org). Inevitably, the nature of the
shortlist was influenced by data availability, with the
variables used being acquired from data sets in the public
domain. Nonetheless, the shortlist included a wide range
of variables representative of the categories listed above.
The shortlisted variables are detailed in Table 1.
3.3. Identifying key indicators of vulnerability
Proxy data representing the 46 shortlisted vulner-
ability variables were averaged over the decadal periods
represented by the mortality data or, where data were
unavailable for multiple years, data from a single year
were used to represent an entire decade. For each proxy,
a data series was constructed, in which each datum
represented a specific country in a specific decade. The
mortality risk proxy series was then correlated with each
of the vulnerability proxy series in turn, in order to
identify statistically significant relationships between
mortality risk and the various socio-economic and other
variables represented by the candidate vulnerability
data. Significance was assessed using a Monte-Carlo
randomisation procedure based on the method of
Ebisuzaki (1997). This method calculates the probability
that the correlation between a pair of series is exceeded
when the data are randomised, while also accounting for
autocorrelation in the data. Proxy data that were
significantly correlated with mortality risk at the 10%
level were adopted as key indicators of vulnerability.
This approach does not address the influence of
hazard type, frequency and severity on mortality
outcomes. If different human systems were subject to
identical hazards, differences in outcome would be a
result purely of variations in vulnerability across the
different systems. In other words, if hazard is constant,
risk is a function of vulnerability alone. However, we
cannot, at the national level, examine outcomes under
conditions of constant hazard, as no two countries have
identical hazard profiles. Furthermore, given the high
degree of geographical variation in the nature of climate
hazards and the resulting difficulty in comparing ‘‘levels
of hazard’’ across different countries, it was concluded
that the development of a hazard index would be
inappropriate. Physical hazard may therefore be viewed
simply as an additional variable influencing mortality
risk, albeit one that is not explicitly addressed in this
analysis. Given that outcomes are determined by the
interaction of a variety of social, economic, political and
environmental factors, including the nature of the
hazards associated with those outcomes, we are essen-
tially assessing the role of the majority, rather than all,
of the factors determining mortality risk. This in no way
invalidates the statistical importance of the vulnerability
proxies identified as significant.
3.4. Validation and interpretation of indicators through
expert elicitation
Delphi surveys and iterative Delhi surveys have
been used occasionally to derive quantitative estimates
of elements of climate change risks and costs (e.g.
Nordhaus, 1994;Vaughan and Spouge, 2002). The
approach here is to utilise expert judgement as a tool for
broad validation of the empirical determination of
indicators outlined above.
A focus group of experts in the field of climate
impacts and vulnerability, convened in August 2003 in
Southampton (UK), provided weightings and interpre-
tation of the indicators identified in the statistical
analysis described above. This type of expert focus
group is frequently used for the elicitation of both
specific refining information and for the generation of
new data and insights through direct interaction
between participants (Morgan, 1996). Each expert was
asked to define their area of expertise in terms of subject
matter and geographical area and the results are shown
in Table 2. Each participant was asked to comment from
their own perspective, however broad or limited in terms
of subject matter or geographical range.
Each participant was asked to consider which key
indicator they felt was the ‘‘most important’’ in terms of
defining or predicting vulnerability, and then to rank the
different indicators according to importance, based on
their experience in different areas of vulnerability
assessment. They were given the opportunity to perform
multiple ranking exercises to represent different hazards
and contexts. Some participants chose to undertake
separate ranking exercises for vulnerability and adaptive
capacity. Participants were also asked to consider
whether they felt a distinction could be made between
those key indicators that represented vulnerability and
those that represented adaptive capacity, and whether a
distinction between vulnerability and adaptive capacity
was meaningful within the context of this exercise.
3.5. Methodology for the construction of composite
vulnerability indices
Those interested in the structural factors behind
vulnerability for particular countries or groups of
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163154
ARTICLE IN PRESS
Table 1
Potential proxies for national-level vulnerability to climate change
Category Variable Proxy Source
Economy National wealth GDP per capita (US$ PPP) WB
Inequality GINI coefficient WIID
Economic autonomy Debt repayments (% GNI, averaged over decadal
periods)
WB
National wealth GNI (total, PPP) WB
Health and nutrition State support for health Health expenditure per capita (US$ PPP) HDI
State support for health Public health expenditure (% of GDP) HDI
Burden of ill health Disability adjusted life expectancy WHO
General health Life expectancy at birth HDI
Healthcare availability Maternal mortality per 100,000 HDI
Removal of economically active
population
AIDS/HIV infection (% of adults) HDI
Nutritional status Calorie intake per capita GRID
General food availability Food production index (annual change averaged over
1981–90 and 1991–99)
WB
Access to nutrition Food price index (annual change averaged over
1981–90 and 1991–99)
WB
Education Educational commitment Education expenditure as % of GNP HDI
Educational commitment Education expenditure as % of government expenditure HDI
Entitlement to information Literacy rate (% of population over 15) HDI
Entitlement to information Literacy rate (% of 15–24 year olds) HDI
Entitlement to information Literacy ratio (female to male) HDI
Infrastructure Isolation of rural communities Roads (km, scaled by land area with 99% of
population)
WB/CISEIN
Commitment to rural communities Rural population without access to safe water (%) HDI
Quality of basic infrastructure Population with access to sanitation (%) HDI
Governance Conflict Internal refugees (1000s) scale by population WB
Effectiveness of policies Control of corruption KKZ
Ability to deliver services Government effectiveness KKZ
Willingness to invest in adaptation Political stability KKZ
Barriers to adaptation Regulatory quality KKZ
Willingness to invest in adaptation Rule of law KKZ
Participatory decision making Voice and accountability KKZ
Influence on political process Civil liberties FH
Influence on political process Political rights FH
Geography and
demography
Coastal risk km of coastline (scale by land area) GRID
Coastal risk Population within 100 km of coastline (%) GRID
Resource pressure Population density CIESIN
Agriculture Dependence on agriculture Agricultural employees (% of total population) WB
Dependence on agriculture Rural population (% of total) WB
Dependence on agriculture Agricultural employees (% of male population) WB
Dependence on agriculture Agricultural employees (% of female population) WB
Agricultural self-sufficiency Agricultural production index (1985, 1995) WB
Ecology Environmental stress Protected land area (%) GRID
Environmental stress Forest change rate (% per year) GRID
Environmental stress % Forest cover GRID
Environmental stress Unpopulated land area CIESIN
Sustainability of water resources Groundwater recharge per capita GRID
Sustainability of water resources Water resources per capita GRID
Technology Commitment to and resources for
research
R&D investment (% GNP) WB
Capacity to undertake research and
understand issues
Scientists and engineers in R&D per million population WB
Note: The data sources are: the World Bank (WB); Human Development Index (HDI); UNEP/GRID-Geneva (GRID); Kaufmann, Kray and Zoido-
Lobaton governance data set; Center for International Earth Sciences Information Network (CIESIN) at Columbia University; United Nations
World Income Inequality Database (WIID).
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 155
countries should examine the scores and rankings of
countries in the individual indicators. Those wishing to
identify highly vulnerable countries for purposes of
adaptation assistance, or as an entry point for case
studies of systemic vulnerability, will find composite
vulnerability indices most useful.
Caution should be exercised when interpreting the key
vulnerability indicators, or when using them to con-
struct composite vulnerability indices. Kaufmann et al.
(1999a, b) discuss problems of use and interpretation for
their governance indicators, for example. There is a high
degree of heterogeneity in the way the KKZ data have
been collected, and they are imperfect subjective
measures of unobservable variables, based on a combi-
nation of surveys of business people and residents within
the countries concerned and the expert judgement of
individuals from outside these countries. The hetero-
geneity in the data means that these data sets can at best
be used to place countries in groups, rather than
compare governance between individual countries.
Kaufmann et al. (1999a, b) also caution against aver-
aging across proxies for a particular country, or
standardising proxies. These cautionary points must be
remembered when incorporating the governance data of
Kaufmann et al. (1999a, b) in any composite indicator,
and constitute general guidelines when using data
representing such ‘‘unobservable’’ variables.
In order to avoid the pitfalls indicated above, a
composite vulnerability index methodology was devel-
oped as follows. For each of the key vulnerability
indicators, the range of data was divided into quintiles,
and each country was assigned to a quintile. Where the
correlation with mortality outcome was positive, a
country in the bottom quintile was assigned a score of
1, and a country in the top quintile a score 5 for the
indicator in question. Where the correlation with
outcome was negative, the scoring system was reversed.
Each country was thus assigned a score of 1–5 for each
key indicator, with a score of 1 representing low
vulnerability and a score of 5 representing high vulner-
ability. Such an approach enables an average score to be
calculated across all the indicators to produce a
composite vulnerability index. However, in order for
the resulting index to have any meaning, we must
address the issue of the relative weights of the different
indicators used to produce the composite index.
3.6. Sensitivity of the composite index to different
weightings
A key issue in the development of indicators is their
robustness, in particular the implications of the use of
different sets of weightings. While the identification of
individual indicators here is based on empirical data, the
analysis does not account for differences in context and
process that mean certain factors (and thus certain
indicators) will be more important than others in
different national (and local) circumstances. The analy-
sis performed here is designed to capture major patterns
in the data, but these patterns will not be representative
of all countries, or of all hazard types.
The application of subjective weightings on the one
hand gives us some indication of how the relative
importance of different factors might vary with context,
and can also tell us how sensitive national vulnerability
ratings are to perceptions of vulnerability in the expert
community. The methodology used to construct the
composite vulnerability index, detailed above, effectively
assigns equal weights to each indicator, an arbitrary
choice in terms of their relative importance. Applying
weights to the individual indicators based on the
strength of the statistical relationship with mortality
outcome would be one way of prioritising certain
variables. However, the key indicators are unlikely to
be independent of each other given the nature of the
shortlisted proxies, and such an approach might over-
emphasise the importance of factors represented by
multiple related variables. Appropriate sets of weight-
ings will also be determined to a large extent by context,
and a reductionist statistical approach will not recognise
this diversity in the nature of vulnerability.
In order to assess the sensitivity of country vulner-
ability rankings to different developmental contexts, the
indicator rankings from the focus group (see above)
ARTICLE IN PRESS
Table 2
Fields of expertise of the focus group participants
Expert Hazard(s) Context(s)
1 All General
2 Drought, crop failure, famine Agriculture, water resources
3 All General
4 Floods, drought, dust, storms, sea intrusion, sea level
rise
Developing countries, coastal management in the UK
5 All Africa
6 Drought, floods Southern Africa
7 Sea level rise, floods UK and developing country coasts, small island
developing states, coastal zones
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163156
were converted to weights, which were then applied to
the key vulnerability indicators. Each set of rankings
from the focus group was thus used to generate an
individual composite vulnerability index, within which
the country rankings were influenced by a set of
subjectively derived weights. Country rankings were
then compared across these different composite vulner-
ability indices in order to assess the impact of different
weightings on the results. A mathematical index
(described in more detail below) was constructed to
provide a quantitative measure of the variation between
different composite vulnerability indices.
3.7. Identifying the ‘‘most vulnerable’’ countries
Rather than simply identifying highly vulnerable
countries based on their position in the equal-weights
index and then assessing whether these results changed
significantly as a result of different weighting proce-
dures, we based our assessment of vulnerability on a
combined assessment of country rankings in multiple
composite indices constructed using a variety of
weighting sets. We define significantly vulnerable
countries as those that occur in the top quintile in at
least one of the composite indices constructed using the
different weighting sets, and the most vulnerable
countries as those that occur in the top quintile for
most or all of the alternative composite indices.
4. Results
4.1. Key indicators of vulnerability
The following vulnerability proxies were found to be
correlated with decadal mortality outcome at the 10%
significance level, and are therefore adopted as our key
indicators of vulnerability:
(1) population with access to sanitation,
(2) literacy rate, 15–24-year olds,
(3) maternal mortality,
(4) literacy rate, over 15 years,
(5) calorific intake,
(6) voice and accountability,
(7) civil liberties,
(8) political rights,
(9) government effectiveness,
(10) literacy ratio (female to male),
(11) life expectancy at birth.
Indicators 1–6 are significant when tested at the 5%
level, and indicators 1–3 are significant when tested at
the 1% level. These indicators can be divided into three
broad categories: health status, governance and educa-
tion. Calorific intake and sanitation are predictive of
health status, while life expectancy and maternal
mortality are diagnostic of health status and of the
efficacy of health care. The focus on literacy rates
among the young is indicative of access to non-manual
employment and to information. The governance
indicators emphasise the ability of citizens to participate
in the political process.
The results demonstrate that it is possible to identify
key indicators of vulnerability from a suite of potentially
important proxies using statistical methods. Of course
these results are open to further interpretation based on
the implications of the data and methodology. For
example, it should be noted that many of the vulner-
ability proxy data are available only for a fraction of the
period over which the associated mortality data are
calculated. The KKZ governance data all represent
1998, towards the end of the period represented by the
mortality outcome data. We might expect this to result
in a decoupling of the governance and outcome data and
an underestimation of the strength of any relationship
between these variables, particularly if there are a
number of countries where high mortality was experi-
enced in the early or mid-1990s as a result of the
occurrence of particular severe climate hazards, and
where structures and institutions of governance subse-
quently evolved significantly over the course of the
1990s. The existence of large and significant correlations
for these data suggests that there is a high degree of
temporal consistency in the vulnerability proxy data,
and that the assumption of relative constancy in socio-
economic and political conditions over the course of a
decade is a reasonable one. An analysis based on shorter
timescales might be expected to indicate even stronger
relationships between these variables.
Of course there are numerous examples of countries
that have experienced dramatic changes in their socio-
political and economic landscapes. While their number
is not sufficient to completely undermine the above
analysis, such discontinuities, and non-catastrophic but
still significant societal change, may still obscure
important relationships between the vulnerability proxy
data and the mortality outcome data.
Counter-intuitively, economic indicators such as
GDP and the Gini-based indicators of income inequality
are not identified in this analysis as significant indicators
of vulnerability. While this may, in part, be the result of
pooling countries with very different socio-economic
profiles, they are clearly not as useful as generic
indicators as the health, literacy and governance data
represented by the 11 variables listed above.
While the results of this stage of the analysis are
plausible, it must be recognised that an analysis of this
type is subject to many caveats and reservations, a
number of which have already been mentioned. The
data sample is short in duration. Data from countries
of very different socio-economic characteristics and
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 157
hazards profiles have been pooled. Most of the national-
level indicators cannot reflect in-country variability in
physical or societal space. This assessment should be
seen as a first step in identifying useful proxies for
vulnerability and requires verification. We present the
results of one such verification exercise later in this
account.
4.2. Results of focus group analysis
The areas of expertise of the focus group participants
are summarised in Table 2. There was a bias towards
developing country experience in the group, and towards
short-term climate hazards, though not exclusively. In
identifying personal areas of expertise, there was some
discussion of the meaning of terms within the group. As a
result, a distinction was drawn between vulnerability,
which, it was agreed would, for the purposes of the focus
group, be taken to mean vulnerability to short-term
hazards, and adaptive capacity, which was related to the
longer-term process of adjustment. It was recognised,
though, that the two concepts are ‘‘difficult to disen-
tangle’’. Indeed, in the rest of this paper, we tend to
subsume the two into a single entity. This distinction did,
however, prove extremely useful in categorising the
selected indicators, as discussed below.
The specific question asked of the participants con-
cerned which of the 11 indicators were ‘‘most important’’
as indicators of vulnerability at the national level. One
noteworthy insight that emerged during the discussion
was the association of the three indicator subsets, health,
governance and education, with timescale:
I would say they rank quite nicely because the health
ones are going to be very much related to short term
vulnerability. In the medium term, governance is
probably the most important and the very long term,
education. (Expert No. 2, agriculture and water
resources focus, August 2003)
Having established that distinction between the short-
term (cc. months), medium-term (cc. years) and long-
term (cc. decades), with the first category related to
vulnerability to short-term hazards and the latter two to
adaptive capacity, there was a feeling that ‘‘you can
more or less pick one out of each of those three
categories’’ to define the most important indicators.
There were distinctions drawn within each category,
though, in particular, concerning indicators of causal
factors and outcomes with a preference emerging for
indictors linked to causation:
Within the health category, my preference would be
to go with the more fundamental causative ones
rather than those which are sort of outcome
indicators. So I would have thought the population
would have access to sanitation and the calorie
intake, one can clearly see how those are going to
affect people’s vulnerability, whereas the terminal life
expectancy at birth are more just measures of the
effect of that. (Expert No. 6, droughts and floods
focus, August 2003)
Beyond that, most of the indicators were advocated by
one participant or another at some point in the discussion
and none was excluded as unreasonable. To determine
whether or not there was a consensus amongst the group
regarding the most important variables, members were
finally asked to rank the different indicators according to
importance, based on their experience in different areas
of vulnerability assessment. They were given the oppor-
tunity to perform multiple ranking exercises to represent
different hazards and/or contexts. Some participants
chose to undertake separate ranking exercises for
vulnerability and adaptive capacity.
Twelve sets of rankings resulted from the focus group
assessment. Three of the participants addressed vulner-
ability and adaptive capacity separately, and two
participants submitted different sets of rankings for
different vulnerability hazards and contexts. One of
these participants separated vulnerability rankings into
those for flood hazards and those for coastal zones in
general. The other submitted rankings for all hazards in
one case, and specifically for the coastal zone in the
other. Other participants submitted general rankings for
all hazards, or specified that their responses were to be
interpreted as relevant to a particular hazard or set of
hazards, either generally or within a particular context.
The aggregate scores, standardised as some partici-
pants ranked a number of indicators in the same
position, are shown in Table 3. Within the three
categories, the key indicators, according to the focus
group, are: for health, sanitation and life expectancy; for
governance, government effectiveness; for education,
literacy 15–24.
There was general endorsement of the indicators that
had emerged from the statistical analysis:
I see them all as proxies in a way for something vague
you might call development. Developing as such isn’t
just a single measure, there are lots of things that
contribute to it. The thing that is omitted from that
group is what people regard as the most important
aspect of development which is poverty but that list
covers the rest of it quite well, but I think, to describe
development properly which I suspect is the really
critical thing, you do need at least one from each
quarter. (Expert No. 4, developing countries focus,
August 2003)
Not that for a country to be developed should be
taken to mean that ‘‘it can automatically adapt and I’ve
seen certainly a lot of the literature make that sort of
assumption implicitly and I think that loses, if you like,
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163158
the diversity and agency that people have and it
constructs the developing areas as victims and I think
that is very dangerous’’.
4.3. Variability in country vulnerability rankings
resulting from different weightings
The focus group exercise generated 12 sets of
responses, each of which was associated with a set of
ranks placing the 11 key vulnerability indicators in order
of subjectively defined relative importance. The ‘‘most
important’’ indicators were ranked as 1, the ‘‘least
important’’ as 11 and equal ranks were also permitted.
These ranks were then converted to weights. A rank of 1
was assigned a weight of 11, a rank of 2 a weight of 10
and so on. For each response, a composite index of
vulnerability and/or adaptive capacity was constructed.
The country scores for an individual indicator (ranging
from 1 to 5) were multiplied by the weight assigned to
that indicator. A country’s score in the composite index
was then calculated as the mean of the weighted scores
which span the range from 1 to 55.
The result of this exercise was 13 alternative
composite indices of country-level vulnerability (includ-
ing the equal-weights index). An initial assessment of
variability in country rankings across these indices was
performed by plotting the mean rank of a country in all
13 indices against the difference between its highest and
lowest rank (Fig. 1). A small number of countries
exhibit very large variations in rank, exceeding 100
places. However, most countries vary in rank by less
than 50 places across the various indices, with a large
number varying by fewer than 25 places, or some 12%
of the possible range (out of a total of 205 countries
represented).
In order to assess the divergence between the various
composite indices constructed using different sets of
weights, we calculated the average difference qin
country rankings between pairs of composite indices:
@m;n
i¼Xjrm
irn
ij
=N,
where rm
iis the rank of country iin index m, and Nis the
number of countries represented in each index (in this
case 205). With 13 indices (including the equal-weight-
ings index), there are 78 possible pair combinations; the
78 corresponding values of @range from 4 to 26. The
greatest value of @occurs between indices 3 and 12,
constructed from sets of weightings chosen to represent
adaptive capacity with respect to drought, crop failure
and famine, and vulnerability to sea level rise within the
context of small island developing states and coastal
zones, respectively. The smallest value of @occurs
between indices 5 and 13, representing generalised
adaptive capacity and generalised, aggregated vulner-
ability and adaptive capacity, respectively. Fig. 2 shows
the ranks in index 3 plotted against those of index 12
(most dissimilar indices) and ranks in index 5 plotted
against those of index 13 (most similar indices). In the
former case, the majority of countries and territories
exhibiting large differences in rank are associated with
missing data in a number of categories represented by
the 11 constituent variables. Many of these are small
island states or territories.
Fig. 2 illustrates that there is a degree of consistency
in terms of country rankings between the different
indices, even where the results are most divergent.
However, assessments of vulnerability or adaptive
capacity based on individual country rankings are
generally not appropriate, due to the variation in rank
across indices and for reasons discussed above. A more
appropriate method of assessment is one based on
placing countries in vulnerability categories, for example
as represented by quintiles of country scores or
rankings. Here we divide the rankings into quintiles to
produce categories containing equal numbers of coun-
tries, across all indices. A further test of robustness
across different sets of weightings is the extent to which
countries fall consistently in a particular quintile, or
vulnerability category, across the different indices. Fig. 3
ARTICLE IN PRESS
Table 3
Scores after standardising the rankings
Indicator Score Rank
overall
Rank within category
Sanitation 0.85 3 Health 1
Literacy 15–24 0.87 5 Education 1
Maternal mortality 1.33 11 Health 4
Literacy 15+ 1.08 7 Education 2
Average calorie intake 1.31 10 Health 3
Voice/accountability 0.70 2 Governance 2
Civil liberties 1.13 8 Governance 4
Political rights 0.94 6 Governance 3
Government
effectiveness
0.69 1 Governance 1
Literacy ratio 1.26 9 Education 3
Life expectancy 0.85 3 Health 1
Note: A low number indicates a high preference.
Fig. 1. Sensitivity of individual country rankings to different weight-
ings of individual indicators.
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 159
shows the % of countries common to the top quintile of
the constituent indices of each possible index pair,
plotted against the corresponding value of @: In most
cases, over two-thirds of the countries represented in the
upper quintiles of a pair of indices are common to both
indices.
4.4. Identification of the ‘‘most vulnerable’’ countries
While there is variation between composite indices
according to the weightings assigned to the variables
from which they are constructed, the distribution of
country scores is not systematically and radically
different, suggesting that the indices are relatively robust
in their representation of vulnerability, yet sensitive
enough to emphasise its different aspects. Countries that
score consistently highly, i.e. that occur in the upper
quintile of multiple indices, may be interpreted as being
highly, and systematically, vulnerable within the frame-
work considered here (i.e. one that emphasises human
mortality). Table 4 lists the countries occurring in the
upper quintiles of any of the 13 indices, along with the
number of occurrences. We define countries occurring in
the upper quintile of one or more index as moderately to
highly vulnerable, and those occurring in the upper
quintile of 11 or more of the 13 indices as the ‘‘most
vulnerable’’ countries. The most vulnerable countries
are nearly all situated in sub-Saharan Africa: of the 59
countries and territories listed in Table 4, 33 are sub-
Saharan African nations, five are small island states or
territories and many have recently experienced conflict.
The results in Table 4 are broadly consistent with
subjective expectations, with the possible exception of
the presence of Estonia, Brunei Darussalam and Qatar.
Bangladesh is less vulnerable than might be expected
ARTICLE IN PRESS
Fig. 2. Scatter plots of country rankings for the indices with the
greatest (top) and least (bottom) divergence, measures in terms of the
mean difference in individual country rankings or ‘distance’, @:
Fig. 3. Overlap between upper quintiles of indices constituting each of
the 78 possible index pairs (% of countries occurring in both indices
for each pair, plotted against mean difference in rank or ‘distance’
between pairs, @).
Table 4
Countries whose rankings occur in the upper quintile of multiple
vulnerability indices (from a total of 13 indices) constructed from
different sets of weightings of constituent variables, with the number of
indices for which they occur in the upper quintile
Most vulnerable countries Moderately to highly
vulnerable
Afghanistan 13 Cote d’Ivoire 10
Angola 13 Qatar 10
Burundi 13 Kenya 9
Central African Rep. 13 Laos 9
Democratic Republic of
Congo
13 North Korea 8
Yugoslavia 7
Eritrea 13 Nigeria 7
Ethiopia 13 Benin 6
Equatorial Guinea 13 Turks and Caicos
Islands
6
Gambia 13 Bosnia Herzegovina 5
Guinea Bissau 13 Congo 5
Haiti 13 Mali 5
Mauritania 13 Guadeloupe 5
Mozambique 13 Senegal 5
Niger 13 Tonga 5
Pakistan 13 Nepal 4
Rwanda 13 Djibouti 3
Sierra Leone 13 Zimbabwe 3
Somalia 13 Azerbaijan 2
Sudan 13 Puerto Rico 2
Togo 13 Bangladesh 1
Turkmenistan 12 Bhutan 1
Chad 12 Estonia 1
Gabon 12 Cambodia 1
Iraq 12 Uganda 1
Liberia 12 United Arab Emirates 1
Malawi 11 French Guiana 1
Brunei Darussalam 11 Morocco 1
Burkina Faso 11 Wallis and Futuna
Islands
1
Guinea 11
Yemen
Note: Countries in continental sub-Saharan Africa are highlighted in
red, small island states or territories in blue.
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163160
according to this analysis, although the factors leading
to large negative outcomes from climate-related disas-
ters in Bangladesh are as much geographical as social
and political, and may therefore be under-represented
by the indices used here. Furthermore, high mortality is
indicative of high risk, which is determined by levels of
climate hazard as well as socio-economic vulnerability;
countries such as Bangladesh may be subject to high
outcome risk as a result of exposure to very high levels
of hazard or event risk, even if their vulnerability is
relatively low when compared with other countries
subject to less significant hazards. A death toll of
138,000 from a tropical cyclone in 1991 was followed by
hugely reduced mortality from a similar event in 1997,
suggesting that Bangladesh has addressed its vulner-
ability with some considerable success. According to
Alam (2003), ‘‘Bangladesh has set a pioneer example in
disaster management during the cyclones of 1991 and
1997. The role of the government and non-government
organizations during the pre and post disaster periods
helped shrink the number of deaths and damage. The
initiatives were appreciated and recognized worldwide.’’
The vulnerability of small island states is likely to be
under-represented as they constitute a minority ‘‘special
case’’; the particular factors that lead to large negative
outcomes in islands (small size, low elevation, isolation,
etc.) are not characteristics of the majority of the
countries assessed here, and their effects will therefore
not lead to a significant statistical signal in the analysis.
However, countries that are exposed to frequent severe
climate extremes are likely to be more prepared than
those that are not, and are likely in many cases to have
reduced their vulnerability through adaptation to
recurrent climate hazards. It is perhaps worth reiterating
here that vulnerability as defined and assessed in this
study only leads to severe negative outcomes when
coupled with climate hazards; conversely, relatively low
vulnerability is no guarantee of safety when it is
countered by extremely high levels of hazard. The
purpose of this study is to develop a method for
identifying vulnerability even where it is not thrown into
sharp relief by frequent and severe extremes, an exercise
that is particularly useful when we are considering the
potential impacts of climate change, which is likely to
lead to the emergence of new hazards, associated with
changing patterns of climate-related risk.
5. Discussion
We have identified 11 key indicators of vulnerability,
broadly defined as integrated over time and incorporat-
ing elements of adaptive capacity. The results from the
focus group exercise emphasise the importance of
governance indicators (voice and accountability and
government effectiveness), sanitation and life expec-
tancy. The participants broadly agreed that the indica-
tors could be separated into variables that were
essentially indicators of instantaneous vulnerability
(i.e. at the time of onset of a hazard event), and those
that represented the capacity to adapt over time. The
exercise emphasised the importance of governance, civil
and political rights, and literacy as indicators of
adaptive capacity. It is notable that GDP is not
identified as a significant indicator of vulnerability. This
does not indicate that poverty is unimportant in
determining vulnerability, rather that GDP does not
capture the aspects of the economic environment that
make people vulnerable. Nonetheless, all the indicators
presented here are likely to be jointly determined, and
related to indicators not significantly correlated with
mortality outcomes; in other areas of research, for
example, human capital and economic growth have been
shown to be intimately related (Krugman and Venables,
1995;Knack and Keefer, 1997). While national eco-
nomic wealth plays a key (but not exclusive) role in
shaping the developmental environment, it appears that
non-economic indicators are more directly representa-
tive of the vulnerability of national populations in terms
of mortality associated with climate-related hazards.
As the indices developed here are based on indicators
of vulnerability related to mortality outcomes, the
indicators of adaptive capacity represent the capacity
to reduce mortality from climate hazards, and should be
used within this context. They may be less appropriate
for assessing economic vulnerability, or for identifying
the factors that enable societies to ameliorate non-fatal
outcomes. The implicit adaptation goal, built in to this
analysis, is one of reducing mortality outcomes from
climate-related disasters (Haddad, 2005).
This analysis indicates that reductions in mortality
outcomes may be achieved through increasing govern-
ment effectiveness and accountability, civil and political
rights, and literacy. While these factors mask the more
complex processes that lead from climate hazards to
high mortality, they also underlie them. A literate
population will be better able to lobby for political
and civil rights, which in turn will allow it to demand
accountable and effective government. Where such
rights exist, governments are more likely to become
accountable for reducing the impact of successive high-
mortality disasters, and are thus more likely to address
vulnerability (Wisner et al., 2003).
It is interesting to compare the 11 key indicators with
other indicators developed for sub-national scale con-
texts. The World Food Programme uses indicators
representing the following variables to assess vulner-
ability to food insecurity in Kenya at the district level
(Haan et al., 2001):
Life expectancy Mean vegetation
Adult literacy Vegetation variation
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 161
Stunting Education
Wasting Gender development
Livelihood diversification Non-agricultural income
Access to safe water Proximity to markets
Livelihood fishing HIV/AIDS incidence
High potential land Civil insecurity
Some of these indicators are identical to those
identified in the analysis described in this paper, and
indicators of health, education and governance (in the
form of security) are represented in both sets. The sub-
national WFP indicators also include variables specific
to the local context, representative of livelihoods and
environmental conditions. These contextual indicators
are extremely important for the natural resource
dependent communities with which the WFP is con-
cerned in this case, but are not universally relevant, and
would be inappropriate or meaningless in certain other
contexts. This comparison of indicator sets across scales
demonstrates that, while the national-level indicators
presented in this paper capture certain important
elements of vulnerability operating at the scale of the
nation state, for a full description of vulnerability sub-
national scale factors representing local contexts must
also be considered.
The results confirm the vulnerability of sub-Saharan
Africa, with this region making by far the largest
contribution to the set of highly vulnerable countries.
Not only does environmental stress make countries
prone to conflict (see e.g. Stewart, 2002), but conflict
exacerbates vulnerability and reduces the adaptive
capacity of a country and its population to deal with
environmental stress. Research on the causal mechan-
isms between conflict and vulnerability suggests that it is
the lack of functioning government, the breakdown of
common management institutions and direct impacts of
conflict on health and well-being that undermine the
capacity to adapt (Barnett, 2005). Many of the most
vulnerable countries, notably including nations outside
of sub-Saharan Africa such as Iraq and Afghanistan,
were experiencing or recovering from conflict during the
period represented by the data.
6. Conclusions
This paper has presented a novel methodology for
assessing vulnerability to climate-related mortality,
based on empirical analysis rather than subjective
identification of indicators, and which addresses the
sensitivity of vulnerability assessments to different sets
of subjective weightings. Our approach provides a firmer
foundation for the identification of highly vulnerable
countries than a simple equating of vulnerability with
poverty, and offers a robust methodology for the
assessment of vulnerability that is not based on a single,
subjective index.
The results of this study confirm the extreme
vulnerability of sub-Saharan Africa, although the nature
of the analysis probably results in the vulnerability of
small island states being underestimated. While the
results broadly support existing views of vulnerability,
they produce some unexpected results, such as the
absence of GPD from the list of key indicators, the
relatively low vulnerability of Bangladesh, and the
relatively high vulnerabilities of some wealthier nations.
These findings in particular suggest potential avenues
for research that may further enhance our under-
standing of vulnerability.
The results from the analysis presented here, and their
interpretation, highlight some of the caveats that must
be borne in mind when using spatially and temporally
aggregated data. Within countries, vulnerability is
geographically and socially differentiated, and processes
that mediate the outcomes of hazard events operate at
the local scale. Ultimately, it is people not countries that
are vulnerable. National-level indicators must therefore
be complemented by locally contextual indicators to
yield a full picture of vulnerability. The analysis
presented here can provide information regarding
leverage points in a country’s institutional environment
for the promotion of resilience, but must be comple-
mented by studies of the distribution of vulnerability,
including the identification of vulnerability ‘‘hotspots’’.
Prospects for adaptation will be improved by addressing
issues of health, education and governance, but specific
measures and technologies for the promotion of
adaptation will also be required. The nature of these
measures and technologies will be determined by local
contexts, and they should be targeted at specific
localities, groups and sectors.
The results of this analysis are important in that they
provide an empirical basis for assessing vulnerability
rather than risk; adaptation through vulnerability
reduction is of particular relevance where climate
projections are unavailable or where such projections
are associated with a high degree of uncertainty, for
example in monsoon regions in sub-Saharan Africa
(Brooks, 2004;Conway, 2005).
References
Adger, W.N., 1999. Social vulnerability to climate change and
extremes in coastal Vietnam. World Development 27, 249–269.
Adger, W.N., Kelly, P.M., 1999. Social vulnerability to climate change
and the architecture of entitlements. Mitigation and Adaptation
Strategies for Global Change 4, 253–266.
Alam, M., 2003. Bangladesh country case study for National
Adaptation Programme of Action (NAPA) Workshop, 9–11
September 2003, Bhutan, http://www.unitar.org/ccp/bhutan/11-
%20Bangladesh%20Case%20Study.pdf, accessed 22 November
2003.
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163162
Barnett, J., 2005. Climate change, insecurity and justice. In: Adger,
W.N., Huq, S., Mace, M.J., Paavola, J. (Eds.), Justice in
Vulnerability and Adaptation to Climate Change. MIT Press,
Cambridge in press.
Bell, S., Morse, S., 1999. Sustainability Indicators. Measuring the
Immeasurable. Earthscan, London.
Brooks, N., 2003. Vulnerability, risk and adaptation: a conceptual
framework. Working Paper 38, Tyndall Centre for Climate Change
Research, University of East Anglia, Norwich. Available at:
www.tyndall.ac.uk.
Brooks, N., 2004. Drought in the African Sahel: long-term perspectives
and future prospects. Working Paper 61, Tyndall Centre for
Climate Change Research, University of East Anglia, Norwich.
Available at: www.tyndall.ac.uk.
Brooks, N., Adger, W.N., 2003. Country level risk indicators from
outcome data on climate-related disasters: an exploration of the
EM-DAT database. Ambio, submitted for publication.
Conway, D., 2005. From headwater tributaries to international river:
observing and adapting to climate variability and change in the
Nile Basin. Global Environmental Change 15, +++–+++
Downing, T.E., Butterfield, R., Cohen, S., Huq, S., Moss, R.,
Rahman, A., Sokona, Y., Stephen, L., 2001. Vulnerability Indices:
Climate Change Impacts and Adaptation, UNEP Policy Series,
UNEP, Nairobi.
Easterling, D.R., Meehl, G.A., Parmesan, C., Changnon, S.A., Karl,
R., Mearns, L.O., 2000. Climate extremes: observations, modeling,
and impacts. Science 289, 2068–2074.
Ebisuzaki, W., 1997. A method to estimate the statistical significance
of a correlation when the data are serially correlated. Journal of
Climate 10, 2147–2153.
Frich, P., Alexander, L.V., Della-Marta, P., Gleason, B., Haylock, M.,
Klein Tank, A.M.G., Peterson, T., 2002. Observed coherent
changes in climatic extremes during the second half of the twentieth
century. Climate Research 19, 193–212.
Haan, N., Farmer, G., Wheeler, R., 2001. Chronic vulnerability to
food insecurity in Kenya—2001 a WFP pilot study for
improving vulnerability analysis. World Food Programme. http://
www.wfp.org/operations/vam/country_experience/kenya_vam/in-
dex.asp, accessed 22 November 2004.
Haddad, B. 2005. Ranking the adaptive capacity of nations to climate
change when socio-political goals are explicit. Global Environ-
mental Change 15, +++–+++.
Intergovernmental Panel on Climate Change 2001. In: McCarthy, J.,
Canziani, O., Leary, N., Dokken, D. and White, K. (Eds.), Climate
change 2001: Impacts, Adaptation, and Vulnerability. Cambridge:
Cambridge University Press.
Jones, R., Boer, R., 2003. Assessing current climate risks. Adaptation
Policy Framework: a Guide for Policies to Facilitate Adaptation to
Climate Change, UNDP: http://www.undp.org/cc/apf.htm.
Kaufmann, D., Kraay, A., Zoido-Lobato
´n, P., 1999a. Governance
Matters. World Bank, Washington, DC.
Kaufmann, D., Kraay, A., Zoido-Lobato
´n, P., 1999b. Aggregating
governance indicators. Policy Research Working Paper 2195,
World Bank, Washington, DC.
Kelly, P.M., Adger, W.N., 2000. Theory and practice in assessing
vulnerability to climate change and facilitating adaptation.
Climatic Change 47, 325–352.
Knack, S., Keefer, P., 1997. Does social capital have an economic
payoff? a cross-country investigation. Quarterly Journal of
Economics 112, 1251–1288.
Krugman, P., Venables, A.J., 1995. Globalisation and the inequality of
nations. Quarterly Journal of Economics 110, 857–880.
Lawn, P.A., 2003. A theoretical foundation to support the index of
sustainable economic welfare, genuine progress indicator, and
other related indices. Ecological Economics 44, 105–118.
Leichenko, R.M., O’Brien, K.L., 2002. The dynamics of rural
vulnerability to global change: the case of southern Africa.
Mitigation and Adaptation Strategies for Global Change 7, 1–18.
Morgan, D.L., 1996. Focus groups. Annual Review of Sociology 22,
129–152.
Moss, R.H., Brenkert, A.L., Malone, E.L., 1999. Vulnerability to
Climate Change: a Quantitative Approach. Pacific Northwest
National Laboratory, Battelle Institute, Washington, DC.
Neumayer, E., 2001. The human development index and sustainability:
a constructive proposal. Ecological Economics 39, 101–114.
Nordhaus, W.D., 1994. Expert opinion on climatic change. American
Scientist 82, 45–51.
O’Brien, K.L., Leichenko, R., Kelkar, U., Venema, H., Aandahl, G.,
Tompkins, H., Javed, A., Bhadwal, S., Barg, S., Nygaard, L., West,
J., 2004. Mapping vulnerability to multiple stressors: climate
change and globalization in India. Global Environmental Change
14, 303–313.
Smith, K., 1996. Environmental Hazards. Routledge, London.
Smit, B., Burton, I., Klein, R.J.T., Street, R., 1999. The science of
adaptation: a framework for assessment. Mitigation and Adapta-
tion Strategies for Global Change 4, 199–213.
Smit, B., Pilifosova, O., Burton, I., Challenger, B., Huq, S., Klein, R.,
Yohe, G., 2001. Adaptation to Climate Change in the Context of
Sustainable Development and Equity. IPCC Third Assessment
Report, Working Group II.
Stenchion, P., 1997. Development and disaster management. Austra-
lian Journal of Emergency Management 12, 40–44.
Stewart, F., 2002. Root causes of violent conflict in developing
countries. British Medical Journal 324, 342–345.
UNDHA, 1992. Internationally Agreed Glossary of Basic Terms
Related to Disaster Management. United Nations Department of
Humanitarian Affairs, Geneva.
Vaughan, D.G., Spouge, J.R., 2002. Risk Estimation of Collapse of
the West Antarctic Ice Sheet. Climatic Change 52, 65–91.
Wisner, B., Blaikie, P., Cannon, T., Davis, I., 2003. At Risk: Natural
Hazards, People’s Vulnerability and Disasters, second ed. Routle-
dge, London.
Yohe, G., Tol, R.S.J., 2002. Indicators for social and economic coping
capacity—moving toward a working definition of adaptive
capacity. Global Environmental Change 12, 25–40.
Further reading
O’Brien, K.L., Leichenko, R.M., 2003. Winners and losers in the
context of global change. Annals of the Association of American
Geographers 93, 89–103 (more of a theoertrical discussion).
ARTICLE IN PRESS
N. Brooks et al. / Global Environmental Change 15 (2005) 151–163 163
