1. Measuring Vulnerability
START Advanced Institute on Vulnerability
May 11, 2004
Karen O’Brien
CICERO, University of Oslo
Email: karen.obrien@cicero.uio.no
CICERO
Center for International Climate and Environmental Research – Oslo
Senter for klimaforskning
Measuring Vulnerability:
1.
2.
3.
4.
Theoretical issues
Conceptualizations of vulnerability
Practical issues
The use of scenarios
Definitions of Vulnerability
1.
2.
3.
”an aggregate measure of human welfare that integrates
environmental, social, economic and political exposure to a range
of harmful perturbations” (Bohle et al. 1994)
“…the exposure to contingencies and stress, and difficulty in
coping with them. Vulnerability thus has two sides: an external
side of risks, shocks and stress to which an individual or household
is subject; and an internal side which is defencelessness, meaning
a lack of means to cope without damaging loss” (Chambers 1989)
”Vulnerability: the degree to which a system is susceptible to, or
unable to cope with, adverse effects of climate change, including
climate variability and extremes. (IPCC 2001)
Why measure vulnerability?
1.
2.
3.
4.
Identify magnitude of threats, such as climate
change;
Guide decision-making on international aid and
investment;
Prioritize aid for climate change adaptation;
Identify measures to reduce vulnerability.
Can vulnerability be measured?





Vulnerability is a characteristic, trait, or condition; not
readily measured or observable, thus we need proxy
measures and indicators;
Vulnerability is relative, not absolute;
Everyone is vulnerable, but some are more vulnerable
than others;
Vulnerability relates to consequences or outcomes, and
not to the agent itself;
Defining levels of vulnerability that prompt actions or
interventions is a social and political process.
What is the opposite of
vulnerability?


Is there an opposite?
Is it resilience, adaptability, or human
security?
Conceptualizing vulnerability




Vulnerability can be conceptualized in different
ways.
Any conceptualization of vulnerability can be
interpreted in different ways.
Conceptualizations and interpretation of
vulnerability have implications for what is
measured and how it is measured.
Vulnerability measures can have political and
economic consequences; transparency (in both
concepts and methods) is necessary.
Biophysical vulnerability


Focuses on ecological processes, exposure
to processes of physical change;
Indicators include length of growing
season; frost days, intense precipitation,
etc.
Social vulnerability


Focus on social, political, economic and
cultural determinants of vulnerability.
Indicators include education, income, and
other proxy data (social capital,
entitlements, livelihood diversification).
Climate change vulnerability
IPCC vulnerability framework:
V = f(E, S, AC)
E = Exposure
S = Sensitivity
AC = Adaptive Capacity
Exposure


The degree of climate
stress upon a particular
unit of analysis
Climate stress:



long-term climate conditions
climate variability
magnitude and frequency of
extreme events
Sensitivity

The degree to which
a system will respond,
either positively or
negatively, to a
change in climate.
Adaptive Capacity

The capacity of a system
to adjust in response to
actual or expected
climate stimuli, their
effects, or impacts.
The degree to which adjustments in practices, processes,
or structures can moderate or offset the potential for
damage or take advantage of opportunities created by a
given change in climate.
Interpretation 1:




Vulnerability analysis as a means of
defining the extent of the climate problem
Vulnerability = Impacts – Adaptations
Adaptability defines vulnerability
Vulnerability is the end-point of the
analysis
Interpretation 2:





Vulnerability analysis as a means of identifying
what to do about climate change.
Vulnerability is shaped by adaptive capacity.
Vulnerability determines adaptability
Vulnerability is the starting point of the analysis.
Under this interpretation, we need measures of
the social processes that contribute to
vulnerability.
Implications


End point: We need better GCM scenarios,
better process models, and better
quantifications of adaptation;
Starting point: We need better
understanding of coping capacity, adaptive
capacity, outcomes of social processes,
and measures of well-being.
Measuring vulnerability:
Practical challenges




How should indicators be chosen?
Are adequate data available?
How should composite indicators be developed?
How can measures of vulnerability be validated?
Choosing indicators: Deductive approach


Theory driven: Start from theory or hypothesis;
find indicators that might support or reject the
hypothesis.
Example: Adger and Kelly (2000) hypothesize
that the architecture of entitlements is a key
determinant of vulnerability in Vietnam; thus
they identify income levels, income inequality
and diversity of livelihood as key indicators.
Choosing indicators: Inductive approach


Data driven: Examine lots of data, look for
patterns and examine correlations or statistical
relationships. Generalizations can be used to
develop conceptual models and theories.
Example: Ramachandran and Eastman (1997)
analyzed 92 variables to explain the need for
food assistance in West Africa. Using statistical
methods, they identified the contributions of
different variables to vulnerability.
Reality:


Eriksen and Kelly (submitted) point out that in
most national level assessments of vulnerability,
the selection of indicators is based on a
”rudimentary theoretical appreciation of
vulnerability (which is often, it is only fair to say,
all that is available)”. Few ”inductive” indicator
studies explicitly discuss implications of findings
for vulnerability theory.
Most studies that measure vulnerability are ”not
easily distinguishable as either deductive or
inductive…”
Data



Need for reliable, readily available, and
representative data for desired indicators of
vulnerability.
Compiling national data is difficult. National level
vulnerability assessments often rely on existing
global data sets (FAO, World Bank, UNDP, WRI,
etc.)
More detailed data usually available for subnational assessments (e.g., census data)
Data
“Data are usually treated unproblematically
except for technical concerns about errors. But
data are much more than technical compilations.
Every data set represents a myriad of social
relations.”
(Taylor and Johnston 1995, p. 58)
Social relations exemplified in different
sources of irrigation statistics for India

Irrigation Department


Revenue office


Irrigation data as basis for repayment of water fee to
maintain irrigation facilities
Irrigation data as basis for land taxes--which are
higher for irrigated lands
Agriculture Department

Supposed to survey all land in the district
 No consistency between these sources
V.-Dimension Desired variables
Empowerment
Child sex rate (”missing girls” or excess girl
mortality)
Female literacy level
Literacy level
Fertility level
Share of landholdings by farm size
% Landless agricultural labourers
Technology
Irrigation rate
Infrastructure Development Index (CMIE)
Source of irrigation
Access to safe drinking water
Fertilizer consumption
Poverty
People below poverty line
Infant Mortality Rate
Housing status
Dependency Employment in agriculture
on agriculture
V.-Dimension
Available data
Empowerment
Child sex rate (”missing girls” or excess girl
mortality)
Female literacy level
Literacy level
% Landless agricultural labourers
Technology
Irrigation rate
Infrastructure Development Index (CMIE)
Poverty
Dependency Employment in agriculture
on agriculture
Does the choice of indicators and
index matter?
”In one sense, this is an empirical question. The
analyst should test different formulations—
choices of indicators, transformations, modes of
aggregation, variations in data quality, etc. If the
overall rankings do not differ much, then one
could argue for the simplest formulation.
Compiling an index is not however an end in
itself. The form of the index may itself be part of
the process of getting support for the index and
its policy implications.”
Source: Downing et al. 2001
Dynamics of vulnerability


Vulnerability is dynamic; indicators are
often static.
Snapshots of vulnerability do not tell us
who is becoming more vulnerable (or less
vulnerable) as time goes on.
Creating composite indices



Vulnerability is multi-dimensional; there is no
one indicator that adequately represents
vulnerability.
Composite indices can provide a more complex
measure of vulnerability.
Many potential methods exist for aggregating
indicators (e.g., indiscriminate aggregation,
weighted indicators, targeted indicators,
contingent indicators, dynamic indicators,
heirarchical vulnerability indices, vulnerability
profiles)
Creating composite indices

”Unless a verifiable outcome variable is
available, there is no clear reason to
choose a particular approach. A guiding
principle may be to keep the analysis
transparent and accessible to end users.”
(Downing et al. 2001)
Verifying measures of vulnerability

”Verification conveys authority and
credibility, but also contributes to
improving the understanding of
vulnerability and hence the representation
of processes in indicator studies” (Eriksen
and Kelly, submitted)
Verifying measures of vulnerability



In the case of deductive approaches, verification
involves assessment of goodness of fit between
theoretical predictions and empirical evidence.
In the case of inductive approaches, the
statistical analysis must incorporate verification
of any results through testing on independent
data.
Unfortunately, such verification has been limited
in existing studies of vulnerability indicators.
Source: Eriksen and Kelly,
article submitted to MASGC
Verifying measures of vulnerability





Is the outcome acceptable?
Does the ranking match what people
expect based on their experience?
Can anomalies be explained?
Who should be the judge?
How can dissenting views be represented?
Source: Downing et al. 2001
Measuring vulnerability: Scenarios


When we are concerned about future conditions
(e.g., under climate change), and we want to
project vulnerability into the future, we need
scenarios.
Focusing on present-day vulnerability to future
climate change can provide a starting point for
actions or interventions to reduce vulnerability;
less useful for assessing the extent of the
climate change problem.
Different types of scenarios:




Climate change scenarios: Generated by general
circulation models (GCMs) or synthetic scenarios (+/10% precipitation, 30 cm sea level rise, etc.);
The output of GCMs depend on assumptions about
greenhouse gas emissions, feedbacks, etc. SRES
scenarios represent emissions according to different
development trajectories;
Vulnerability will depend on social and economic trends
(economic development, population growth);
However, globalization is creating structural social,
economic and political changes, thus extrapolation of
trends into the future may not be sufficient to describe
the future.
Scenarios



How can we incorporate future scenarios
into measures of vulnerability?
What types of uncertainty are added to
vulnerability measures?
How can measures of vulnerability based
on scenarios be validated?
2. Mapping Vulnerability
CICERO
Center for International Climate and Environmental Research – Oslo
Senter for klimaforskning
Why map vulnerability?


Vulnerability can be both socially and
spatially referenced (it is associated with
social and environmental phenomena,
which often have locational components);
Measures of vulnerability can be visualized
through mapping, and patterns can be
identified and analyzed through spatial
analysis (tomorrow’s lecture!).
How to map vulnerability?



Mental mapping
Remote sensing (NDVI)
Geographic Information Systems and
Science (GIS)
Examples of vulnerability maps:
The issue of scale



National scale assessments of vulnerability
(to produce a global map)
Regional vulnerability assessments (e.g.,
West Africa)
Sub-national vulnerability assessments
(e.g., Norway, India)
National level vulnerability maps



Need indicators common to all countries
(comparable time periods, units)
Present coarse generalizations; hide subnational variations and ”pockets of
vulnerability.”
Can be useful for broad comparisons,
correlation with other national statistics
(GHG emissions)
Regional-level vulnerability maps




Represents differential vulnerability across
regions;
Context-specific indicators can be chosen;
Potentially greater availability of data
(from regional institutions, or compiled
from national statistics);
Useful for identification of regional ”hot
spots” and policy analysis.
Sub-national vulnerability maps



Represents variations in vulnerability
within one country, state, county, district,
or village;
Potentially larger amount of data available
(but large data gaps can still exist);
Can be used to develop national
adaptation strategies, aid distribution,
development plans, etc.
Challenges



Integrating raster and vector or
biophysical and social data;
Normalization and weighting of indicators;
Classification
Example of Mapping Approach

Vulnerability of Agriculture to Climate
Change in Norway
Indicators of biophysical
vulnerability: Agricultural sector






Spring rainfall
Autumn rainfall
Length of growing season
Spring frost/thaw
Autumn frost/thaw
Snow depth
Indicators of social vulnerability:
Climate sensitivity

Employment in agricultural sector, %
Economic capacity


Untied public income (taxes and govt. transfers), NOK
Employment growth prognosis, %
Demographic capacity



Dependency rate, %
Aging working population, %
Net migration rate, avg. 91-01 %
How correct are these
indicators?


Case studies must be carried out to verify
the indicators selected, and identify
factors that shape vulnerability in
Norwegian municipalities.
Stakeholder dialogues: Voss and Oppdal
Mapping Vulnerability to
Multiple Stressors: Climate
Change and Globalization in
India
Karen O’Brien1, Robin Leichenko2, Ulka Kelkar3, Henry
Venema4, Guro Aandahl1, Heather Tompkins1, Akram Javed3,
Suruchi Bhadwal3, Stephan Barg4, Lynn Nygaard1, Jennifer
West1
1CICERO
2Rutgers
University
3TERI
4IISD
Indian agriculture




Agriculture is the dominant economic sector
(employs 68% of the population)
Highly vulnerable to climate variability and
climate change
Undergoing rapid economic changes,
presently threatened by globalization
(especially import competition, removal of
domestic subsidies)
Appropriate example for investigation of
vulnerability to multiple stressors
Mapping Vulnerability to
Multiple Stressors
1) develop a regional vulnerability profile for
climate change
2) develop a regional vulnerability profile for an
additional stressor (in this case
globalization)
3) superimpose the profiles to identify
districts that are “double exposed;” and
4) investigate double exposure at the local level
via case studies
Step 1: Develop Profile of
Vulnerability to Climate
Change


Operationalized the IPCC-based definition
of Vulnerability (McCarthy et al. 2001)
Vulnerability to climate change is a
function of adaptive capacity, sensitivity,
and exposure
Defining Adaptive Capacity,
Sensitivity and Exposure



Adaptive capacity: the ability of a system to adjust to
actual or expected climate stresses, or to cope with
the consequences (a function of current socialeconomic-technological conditions)
Sensitivity: the degree to which a system will respond
to a change in climate, either positively or negatively
(we based this current climatic conditions)
Exposure relates to the degree of future climate
stress upon a particular unit of analysis (we based
this on projected climatic change)
Operationalizing Adaptive
Capacity


A function of a combination of social, economic
and technological factors
 Social: literacy, gender equality
 Economic: agriculture share of labor force,
land ownership
 Technological:quality of infrastructure and
availability of irrigation
Additive index, normalized and scaled: higher
adaptive capacity implies lower vulnerability
Adaptive Capacity
Operationalizing Sensitivity
and Exposure



Sensitivity: function of dryness and monsoon
dependence under normal climate
Exposure: Alter the sensitivity index using
climate change scenarios (downscaled
HadRM2 model)
Additive index, normalized and scaled so that
highest sensitivity under exposure implies
highest vulnerability
Sensitivity and
Sensitivity Under Exposure
Climate Change
Vulnerability


Summed adaptive capacity with sensitivity
under exposure
Reveals current vulnerability to future
climate change
Climate Change Vulnerability
Step 2: Develop Profile of
Vulnerability to Globalization



Agricultural trade liberalization a key
dimension of globalization for Indian
agriculture
Focus on import competition
Used IPCC typology of adaptive capacity,
sensitivity and exposure
Operationalizing Globalization
Vulnerability


Adaptive capacity: same definition as used
for climate change adaptive capacity
Sensitivity (and exposure) to import
competition: crop productivity, production
patterns and distance to ports

low productivity, high shares of production in
import competing crops, and close proximity
to ports make an area more sensitive to
competition from international imports
Globalization vulnerability
Step 3: Identify areas of
double exposure


Overlay climate change and globalization
vulnerability profiles to identify areas that
are double-exposed
Use the information to inform policy and
to suggest areas for case study research
Double Exposure
Summary




Our approach reveals relative distribution of
vulnerability to multiple stressors
Areas of double exposure need special
attention from policy makers
Vulnerability concept applies to a wide range
of stressors -- human dimensions work
informs other social science research
Need to combine macro profiles with locallevel investigation
Key findings relevant to
vulnerability mapping



Need to ”ground truth” the maps;
Not all factors contributing to vulnerability
can be captured in quantitative indicators
(e.g., institutional factors, policies);
Vulnerability changes over time.
Social Vulnerability in India
Two components of vulnerability
How useful are vulnerability maps?



Developing vulnerability measures and maps moves
”vulnerability science” forward; they force us to clarify
concepts; address methodological challenges;
interrogate assumptions, hypotheses, and the processes
that contribute to vulnerability;
They provide a means of depicting differential
vulnerability;
The output maps can be dangerous if the concepts and
methods are not transparent, and if they are taken as
reality, rather than as one representation of reality.
”All maps state an argument about the
world” (Brian Harley)



Know your concepts
Know your data
Know your case
Hands On Exercise: Mapping
Vulnerability in India
START Advanced Institute on Vulnerability
May 11, 2004
CICERO
Center for International Climate and Environmental Research – Oslo
Senter for klimaforskning
When mapping vulnerability, how
do we define and assign different
levels of vulnerability?
Issue that we will address:



Normalization
Weighting
Classification
Normalization

HDI method (UNDP): Normalization to
the range
( xi  xmin )
100 
( xmax  xmin )

But to which range?
Fixing of ”goalposts” for
max and min values

Comparison in space


Who should we measure against?
Comparison in time


Retrospective: What has happened in earlier
periods?
Prospective: What are projections for the
future?
(reference: Anand and Sen 1994)
Goalposts: two alternatives
1.
Use the actually occurring range
or
2.
Use predefined maximum and minimum
values
Goalposts: actual range or
predefined?
Occuring range
[1991,2001]
Independent min
and max
6,6% - 94,7%
0% - 100%
Agricultural labourers
0,06% - 88,25%
0% – 100%
Literacy
13,7% - 95,7%
10%– 100%
Female literacy
4,2% - 93,97%
0% – 100%
”Missing girls”
43,2% - 48,5%*
40,0% - 48,5%*
Indicator
Agricultural
dependency
Normalization: range (2) vs predefined max and min (3)
Normalization: range (2) vs predefined max and min (3)
- impact on ranks
Weighting

We gave equal weighting to the three
components of adaptive capacity; and an
equal weighting between adaptive
capacity and sensitivity/exposure. Other
weightings were tested, but without a
priori reasons for weighting one index
higher than another, it was considered
best to keep it simple.
Classification

Can exaggerate non-significant differences

Can hide significant differences
Data distribution for social
index, 1991
SV91_2
80.00
70.00
Vulnerability score
60.00
50.00
40.00
30.00
20.00
10.00
0.00
1
13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193 205 217 229 241 253 265 277 289 301 313 325 337 349 361 373 385 397 409 421 433
Districts
Data distribution for social index, 1991 –
natural breaks (minimized variance within groups)
Data distribution, norm to range
80.00
70.00
Vulnerability score
60.00
50.00
40.00
30.00
20.00
10.00
0.00
1
13 25 37 49 61 73 85 97 109 121 133 145 157 169 181 193 205 217 229 241 253 265 277 289 301 313 325 337 349 361 373 385 397 409 421 433
Districts
Data distribution for social index, 1991 –
quantiles (groups are equal size, 20% of pop)
Data distribution, norm to range
80,00
70,00
Vulnerability score
60,00
50,00
40,00
30,00
20,00
10,00
0,00
1
15 29 43 57 71 85 99 113 127 141 155 169 183 197 211 225 239 253 267 281 295 309 323 337 351 365 379 393 407 421
Districts
Classification: natural breaks (nb) vs quantiles
(qnt)
Exercise

We will map climate change vulnerability in
India, using different weightings and
classifications. The first objective of the exercise
is to explore how sensitive or robust vulnerability
maps are to such decisions. The second
objective is to change the composition of the
index and try to create a map that depicts the
eastern coast of India (including Orissa) as
highly vulnerable.
Files:


Excel spreadsheet with district-level
indicators for India;
Shape files for district and state
boundaries;
The following questions
should be answered:




How sensitive is the map to weighting and
classification methods?
How easily can indicators be manipulated
to show whatever you want to show?
What are the policy or political
consequences of these findings?
How can we prevent misuse of
vulnerability maps?
GIS

Introduction to ArcGIS