resilience index measurement and analysis · challenges in resilience measurement rima-ii rima...
TRANSCRIPT
24/4/2018
1
Resilience IndexMeasurement and Analysis II
RIMA-II
Marco d’ErricoResilience Analysis and Policies teamAgricultural Development Economics Division
Food and Agriculture Organization of the United [email protected]
Ou
tlin
eResilience and RIMA-II
What you can get from RIMA
RIMA-II: the descriptive measure
RIMA-II: the causal measure
Stepping up policy influence
1
2
3
4
5
24/4/2018
2
Resilience and RIMA-II
Photo: FAO
RESILIENCE
Latin: “Resilire”“to jump back”
Engineering“…ability to return to a steady
state after a perturbation”
19th Century
naval architecture“…the ability of materials to
withstand severe conditions”
Ecology “the magnitude of
disturbance a system
can absorb before it
redefines its structure...”
Economics & food
security“the ability of a household to keep
with a certain level of well-being
withstanding shocks and
stresses...” (options and ability)
The concept of resilience
24/4/2018
3
•Why resilience?
• Increased risks and uncertainty: natural risks
Mo
tiva
tio
n &
re
se
arc
h q
ue
stio
ns
50
100
150
200
250
300
350
400
1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008
Climatological events (Extreme temperature, drought, forest fire)
Hydrological events(Flood, mass movement)
Meteorological events(Storm)
Asia, 1980-2008
•Why resilience?
• Increased risks and uncertainty: economic risks
Mo
tiva
tio
n &
re
se
arc
h q
ue
stio
ns
SD92-06 = 13,5
SD07-11 = 29,3
24/4/2018
4
•Why resilience?
• Increased risks and uncertainty
•Scholarly debate and policy frameworks
• WB’s Social Protection and Labor Strategy (2012):
“Resilience, Equity, Opportunity”
• Davos 2013 World Economic Forum “Resilient Dynamism”
• IFPRI 2020 Conference (Addis Ababa, 2014) “Building
Resilience for Food and Nutrition Security”
• EU Action Plan for Resilience in Crisis Prone Countries
2013-2020
• FAO SO5: Increase the resilience of livelihoods to threats
and crises
Mo
tiva
tio
n &
re
se
arc
h q
ue
stio
ns
•Resilience vs. early warning
•Rather than forecasting a crisis (EW), resilience
analysis seeks to understand what are the
• the determinants of vulnerability
• the strategies to gain a livelihood, and
• how these strategies are modified to reduce the negative
impact of future shocks
• ⟹ health check of the system at hand
Mo
tiva
tio
n &
re
se
arc
h q
ue
stio
ns
24/4/2018
5
•Resilience vs. vulnerability
•Vulnerability, output-based: asset-income-wellbeing
(Dercon, 2001)
• V = f (exposure to risk, resilience)
• risks faced by the HH
• options available to the HH
• ability to handle risks
•Resilience
• risk reduction and mitigation (ex-ante actions)
• risk coping (ex-post actions)
• short term (e.g. coping) vs. long term (e.g. adaptation) Mo
tiva
tio
n &
re
se
arc
h q
ue
stio
ns
Res
ilie
nce
mea
sure
men
t• Is not observable in nature;
• can be applied to various systems (households; community; nations) and sciences (ecological and economic and architectural);
• is highly context-specific;
• changes characteristics and effects based on the nature and extent of shocks;
• is highly time-dependent; and
• We need to consider the “dynamics” of resilience.
Most adopted approach through a multivariable index (Constas et al., 2016):
𝑅𝐸𝑆𝑖,𝑡 = 𝛼1𝐴1𝑖 + 𝛼2𝐴2𝑖,𝑡 + 𝛼3𝑋𝑖 + 𝛼4𝑆𝑖,𝑡 + 𝜀
Resilience:
24/4/2018
6
Res
ilie
nce
an
d R
IMA
-II
• a context-specific concept with respect to: - specific population of interest- specific outcome of interest- specific shocks
• Linked to an outcome- Resilience is on the right hand of the equation- The Y is in the LHS (food security; consumption)
• Time-dependentImpact on resilience can be measured as change over time; need baseline/end-line data. It is all about time.
How can resilience be measured?
quantitative vs qualitative
big surveys vs lighter surveys
ad hoc vs pre-existing data
Challenges in Resilience Measurement
Res
ilie
nce
an
d R
IMA
-II
RIMA (Resilience Index Measurement andAnalysis) is an innovative quantitative approachthat estimates resilience to food insecurity andgenerates the evidence for more effectivelyassisting vulnerable populations.
RIMA allows explaining why and how somehouseholds cope with shocks and stressorbetter than others do and provides rigorousframework for humanitarian and long-termdevelopment initiatives to build food secure andresilient livelihoods.
24/4/2018
7
Res
ilie
nce
an
d R
IMA
-II
RIMA suits several definitions of resilience:
• The ability to prevent disasters and crises as well as toanticipate, absorb, accommodate or recover from themin a timely, efficient and sustainable manner (FAO, 2013)
• The capacity of a household to bounce back to aprevious level of well-being (for instance food security)after a shock (Alinovi, Mane & Romano, 2009)
• The capacity that ensures adverse stressors and shocksdo not have long-lasting adverse developmentconsequences (Resilience Measurement TechnicalWorking Group, 2014)
RIMA is focused on householdsR
esil
ien
ce a
nd
RIM
A-I
I
• It is the unit within which the most important decisions to manage uncertain events are made
• It is the unit that benefits the positiveeffects of policies and suffers for negativeeffects of shocks
24/4/2018
8
Co
nce
ptu
al f
ram
ewo
rkR
esil
ien
ce a
nd
RIM
A-I
IRIMA-II provides a comprehensive estimation of resilience and clear policy indications.
It estimates household resilience to food insecuritywith a comprehensive pack which includesdescriptive and causal measure as well as long andshort term measurement approaches
Shocks are considered exogenous and included into aregression model for estimating their impact on foodsecurity and on resilience
Food security variables are considered exogenousindicators of resilience capacity
24/4/2018
9
Sh
ock
s
RIMA-II takes into account several types of shocks that can affect households.
Shocks affecting one household such as livestockdeath, job loss and illness of a household member.These shocks are all directly reported byhouseholds in surveys (idiosyncratic shocks)
Shocks affecting an entire community (covariateshocks) which in turn are divided into:
Climate shocks, such as droughts, floods,rainfalls and other natural hazards,registered through GIS;
Conflict-related shocks, such as war,murders and social disorders
Dat
aset
Quantitative data
Existing data (LSMS, MICS, other HH budget survey)
• LSMS-ISA (Niger, Nigeria, Ethiopia, Malawi, Mali, Uganda, Tanzania)
• Kenya Integrated Household Budget Survey 2005
Ad hoc data (LSMS-type, primary data collection through surveys)
• Baseline/final survey for impact evaluation (South Sudan, Sudan, Somalia)
• Sampling; design; training; data collection, entry, cleaning & analysis
Validated and integrated with qualitative data• Focus group, rapid assessment,
other tools
Qualitative data
24/4/2018
10
Dat
aset
Mixed methods approach
RIMA-II: the descriptive measure
Photo: FAO
24/4/2018
11
• Descriptive measure• It provides information on household resilience capacity.
• RIMA-II employs latent variable models to estimate the Resilience Capacity Index (RCI) and the Resilience Structure Matrix (RSM).
• It is a valuable policy analysis tool to inform funding and policy decisions, as it allow to target and rank households from most to less resilient.
Des
crip
tive
mea
sure
Access to basic services (ABS)
Assets (AST) Adaptive Capacity (AC)
Social Safety Nets (SSN)
Household resilience
Resilience pillars
24/4/2018
12
Des
crip
tive
mea
sure
Resilience pillars Definition
Adaptive CapacityAdaptive Capacity is the ability of a household to adapt to a new situation and develop new strategies of livelihood
Social Safety Nets
The Social Safety Nets pillar measures the ability of households to access timely and reliable assistance provided by international agencies, charities, and NGOs, as well as help from relatives and friends.
Assets
Assets comprise both productive and non-productive assets. Examples of indicators include land, livestock and durables. Other tangible assets such as house, vehicle, and household amenities reflect living standards and wealth of a household.
Access to Basic Services
Access to Basic Services shows the ability of a household to meet basic needs, and access and effective use of basic services; e.g., access to schools, health facilities; infrastructures and markets.
Des
crip
tive
mea
sure
1) Factor analysis: from observed variables to pillars
2) Multiple Indicators Multiple Causes: from pillars to RCI
Two-step procedure for RCI estimation:
24/4/2018
13
RC
I es
tim
atio
nPre:
• Outliers (avoiding multiple imputation and employing “imputeout”Stata command by variable)
• “Positive” direction of the variables (e.g. inverse distances to services)
• Test of variables correlation and collinearity (corr and alpha command)
Post:
• Use of the iterated principal-factor method for analyzing the correlation matrix
• Use of the Bartlett method for predicting as many factor scores explain the 90% of varibles’ variance
• Generate the 4 pillars ABS, AST, SSN and AC as a linear combination of predicted factors (the weights are the percentages of explained variance)
Step 1 – Factor analysis
RC
I es
tim
atio
n
𝑅𝐶𝐼 = 𝛽1, 𝛽2 , 𝛽3 , 𝛽4 ×
𝐴𝐵𝑆𝐴𝑆𝑇𝑆𝑆𝑁𝐴𝐶
+ 𝜀3 (2)
𝐹𝑜𝑜𝑑 𝑒𝑥𝑝𝑒𝑛𝑑𝑖𝑡𝑢𝑟𝑒𝐷𝑖𝑒𝑡𝑎𝑟𝑦 𝑑𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦
= Λ1, Λ2 × 𝑅𝐶𝐼 + 𝜀1, 𝜀2 (1)
Multiple Indicators Multiple Causes (MIMIC) (Bollen et al., 2010)
• The measurement equation (1) reflects that the observed indicators of food security are imperfect indicators of resilience capacity – and the structural equation (2) correlates the estimated attributes to resilience capacity
• A scale has been defined setting the coefficient of the food expenditure loading (Λ1) as equal to 1
• Fit statistics: Chi2, TLI, CFI.
Step 2 – MIMIC
24/4/2018
14
Dat
aset
• Pre-existing data (LSMS, MICS, NHBS, … )
• Ad hoc data
Sources of data for covariate shocks:
(1) Episodes of violence
collected by Armed Conflict Location & Event Data Project (ACLED): www.acleddata.com
and Peace Research Institute Oslo (PRIO): www.prio.org/Data/Armed-Conflict/UCDP-PRIO
(2) Geo-climatic variables
Normalized Difference Vegetation Index (NDVI):
www.fao.org/giews/earthobservation
Th
e ro
le o
f sh
ock
sRIMA takes into account several types of shocks:
Idiosyncratic shocks, such as livestock death, job lossand illness of a household member. These shocks areself-reported by households in surveys.
Covariate shocks, which in turn are divided into:
Climate shocks, such as droughts, floods, rainfalls andother natural hazards, registered through GIS (FAO-GIEWS);
Conflict-related shocks, such as war, murders and socialdisorders (ACLED, UCDP/PRIO, HIIK), damages (OCHA);
Market shocks, such as input/output price fluctuations (WFP)
24/4/2018
15
Sh
ock
ty
pes
an
d s
ou
rces
Shock code
Z2.2 In the last month, have you or any of your household members experienced [SHOCK]? [what is in Z2.2 cannot be in Z2.5]0. No >> NEXT SHOCK1. Yes
Z2.3. Can you quantify the total loss you suffered?
RMO
Z2.4. What did your household do in response to this [SHOCK] to try to regain your former welfare level? USE COPINGS STRATEGIES CODES ON CODE PAGE.
Z2.5. In the last 12 months, have you or any of your household members experienced [SHOCK]? [what is in Z2.5 cannot be in Z2.2] 0. No >> NEXT SHOCK1. Yes
Z2.6. Can you quantifythe total loss you suffered?
RMO
Z2.7. What did your household do in response to this [SHOCK] to try to regain your former welfare level? USE COPINGS STRATEGIES CODES ON CODE PAGE.
Flood
Drought
Crop disease
Livestock death
Business failure
High food prices
High input prices
Severe water shortage
Crop failure
Loss of land
Accident
Severe illness
Clashes
Death of main earner
Inability to pay loan
Displacement
Storm
Crop damage when stored
Job loss/no salary
Communal/Political crisis
Fire
Fishing Failure
Loss of fising gear
others SPECIFY
Shock module – Triangle of Hope, Mauritania questionnaire
Sh
ock
ty
pes
an
d s
ou
rces
Agricultural Stress Index (ASI) in Senegal, January 2016 Resilience analysis in Matam, Senegal (2016) report
24/4/2018
16
RIMA-II: the causal measure
Photo: European Commission DG ECHO / Flircr
• Causal measure• RIMA-II estimates the main determinants of food recovery and it
moves the resilience analysis in the long term perspective.• The causal measure can be adopted as a predictor tool for
interventions that build and strengthen resilience to food insecurity.
• It provides new depth and breadth to resilience analysis and permits decision makers and other stakeholders to better understand the dynamics of positive trends in resilience and thus develop strategies that will yield positive results.
24/4/2018
17
Cau
sal m
easu
reFood security trajectory
What you can get from RIMA
24/4/2018
18
Wh
at q
ues
tio
ns
answ
er?
Who is most in need?
Where should investment focus in terms of geographical location?
Which dimensions of resilience need to be supported?
To what extent have interventions increased target populations’ resilience? Was our money well-spent?
What are the main determinants of food security recover?
Des
crip
tive
an
aly
sis
The most important pillars of resilience are Access to Basic Services and Assets (productive and not)
Resilience analysis in the Triangle of Hope (Mauritania)
Regional heterogenity: Brakna is the most resilientregion, followed by Assabaand Tagant. Guidimagha isthe least resilient one.
24/4/2018
19
Wh
at y
ou
can
get
fro
m R
IMA
Resilience maps
Des
crip
tive
an
aly
sis
Urban households have on average higher resilience capacitythan rural households.
Rural vs urban status
24/4/2018
20
Des
crip
tive
an
aly
sis
The urban effect is found within each region (by the t-test on the difference RCI), with the exception of Tagant which is predominantlyrural.
Rural vs urban status
Des
crip
tive
an
aly
sis
Resilience Structure Matrix: correlation pillars - RCI by urban status
Rural vs urban status
24/4/2018
21
Res
earc
hq
ues
tio
n
CONFLICT
FOOD SECURITY RESILIENCE
How are they captured?
• Conflict exposure Self-reported assessment & HH localization
• Resilience capacity & FS __ FAO-RIMA
Eco
no
met
ric
app
roac
hInstrumental Variable regression:
𝐶𝑖 = 𝛽0 + 𝛽1𝐷𝑖 + 𝛽2𝑿𝒊 + ε
𝐷𝑖 = distance (km) locality-boarder * > 1 Km to buffer zone𝑪𝒊, indicator of conflict exposure is a dummy for residence damaged because of aggression;𝒀𝒊, outcome of interest, in different specification RCI; ABS; AST; SSN; AC, food security indicators and pillars’ components.
1) Omitted factors (time-varying) affecting resilience and conflict exposure;
2) Measurement errors in conflict exposure;3) Self-selection.
24/4/2018
22
Fin
din
gs
First-stage regression results: instrumenting the residence damage with the distance to the Israeli border
Controlling for HH characteristics and self-reported shocks
1) RelevanceThe distance to the border is a statistically significant (negative) predictor of the likelihood of being affected by residence damages
2) Exogeneity• Instrumenting conflict with the area of maximum violence intensity is widespread in
the empirical literature (Akresh & de Walkw, 008; Voors et al., 2012; Rohner et al., 2013; Serneels & Verpooten, 2015);
• For the small size and Israeli restrictions living conditions (job opportunities, food availability, market access, etc.) are homogenous as food security levels;
• Differences in the observable varibles (balance test) between affected and non-affected households are not significant on sub-samples of HHS located less than 1 Km from the buffer zone, 1Km to the border, 2Km to the border, up to 9km to the border.
Fin
din
gsSecond-stage regression results: impact of residence damage on RCI,
resilience pillars and food security indicators
Controlling for HH characteristics and self-reported shocks
24/4/2018
23
Co
ncl
usi
on
s
Key message:
• Food security of households in Gaza was not directly affected by the conflict;
• Household resilience capacity that is necessary to resist food insecurity declined as a result of the conflict, mainly due to a reduction of adaptive capacity, driven by a deterioration of income stability and income diversification.
• Conflict increased the use of social safety nets (cash, in-kind and other transfers) and access to basic services (mainly sanitation and school).
Extensions:
• New waves of the panel dataset to study the persistency of the effects;
• Additional sources of data (e.g. child malnutrition)
Sea
son
alit
y
1) First data collection: Nov-Dec 2015 (post harvest season)Reference period: last 12 months
2) Second data collection: Jul-Aug 2016 (post hot dry season)Reference period: last 7 months
Why seasonality is relevant?- Differences between cropping seasons (post
harvest vs growing season);- Differences between consumption habits
(fresh vs stored food) and asset smoothing;- Differences in subjective well-being (happiness
vs sadness);
Seasonality in the Triangle of Hope (Mauritania)
24/4/2018
24
Sea
son
alit
y
01
02
03
04
05
06
0
Re
sili
en
ce
Ca
pa
city I
nd
ex
Rural Urban
Nov-Dec 2015 July-Aug 2016 Nov-Dec 2015 July-Aug 2016
Source: Authors' own calculation
Resilience over Urban/Rural location
01
02
03
04
05
06
0
Re
sili
en
ce
Ca
pa
city I
nd
ex
Assaba Brakna Tagant Guidimagha
Nov
-Dec
201
5
July-A
ug 2
016
Nov
-Dec
201
5
July-A
ug 2
016
Nov
-Dec
201
5
July-A
ug 2
016
Nov
-Dec
201
5
July-A
ug 2
016
Source: Authors' own calculation
Resilience over regions
Regional differences
• In the first seven months of 2016, resilience capacity decreased with respect to the previous round
• Guidimagha is still the less resilient region
Urban status differences
• Urban household are still more resilient than rural ones
• In general, the level of resilience decreased
Imp
act
eval
uat
ion
• Results show an increase in resilience capacity (23%), obtained through a positive impact on agricultural production, income deriving from livestock, transfers, diversification of income sources and access to infrastructures.
Impact evaluation in Dolow (Somalia)
• Impact evaluation in Dolow, Somalia, is being implemented in the framework of the Joint Resilience Strategy programme launched in 2012 by FAO, UNICEF and WFP.
• It is based on a baseline and on a mid-term datasets.
24/4/2018
25
Per
cep
tio
n
Module developed with Overseas Development Institute (ODI) and implemented in Mauritania, Senegal, Somalia and Uganda (Karamoja).
Generic Shock Drought
H6.1 Absorptive
Capacity
My household can bounce back from any
challenge that life throws at us
H6.11 Absorptive
Capacity
If an severe drought occurred tomorrow, my
household would be well prepared in
advance
H6.2 Absorptive
Capacity
My household is better able to deal with
hardship compared with others in our
community
H6.12 Absorptive
Capacity
If an severe drought occurred tomorrow,
my household could recover fully within six
months?
H6.3 Adaptive
Capacity
If threats to my household become more
frequent and intense, we would still find a
way to get by
H6.13 Adaptive
Capacity
If severe droughts were to become more
frequent and intense, my household would
still find a way to get by?
H6.4 Transformative
capacity
During times of hardship, my household
can change its primary source of income or
livelihood if needed
H6.5 Financial capitalMy household can afford all of the things
that it needs to survive and thrive
H6.6 Social capitalMy household can rely on the support of
family and friends when we need help
H6.7 Social/Political
capital
My household can rely on the support
politicians and government when we need
help
H6.8 Learning
Capacity
My household has learned important
lessons from past hardships that will help
us to better prepare for the future
H6.9 Anticipatory
Capacity
My household is fully prepared for any
future threats and challenges that life
throws at us
H6.10 Knowledge and
Information
My household frequently receives
information warning us about future
extreme weather events in advance
1) Strongly Agree ; 2) Agree; 3) Neither Agree or Disagree; 4) Disagree 5) Strongly Disagree
Perceived resilience
Per
cep
tio
nHow well-being perception and social inclusion indicators are associated with resilience capacity?
Perception of well-being and social inclusion in Matam(Senegal) and the Triangle of Hope (Mauritania)
• The perception of well-being:
“Has the HH during the last week felt (i) cheerful and in good spirit; (ii) calm and relaxed; (iii) active and vigorous; (iv) fresh and rested; (v) that his/her life has been filled with interesting things”
• The perception of social inclusion in the decision-making process:
“Is the current process of decision-making in your community: based on mutual agreement among all men and women (4); based on mutual agreement but with lesser participation of women (3); based on participation but without agreement (2); elite or leader driven (1); don’t know (0)”
• The perception of social inclusion in local services provision:
“To what extent can members of this community influence the public sector to provide better local services: a great deal (4); some (3); a small amount (2); none (1); don’t know (0).”
24/4/2018
26
Per
cep
tio
n
• Endogeneity issue between well-being indicator and RCI: no possibility of causal inference! (confirmed by Hausman test)
• Well-being indicator is the result of a Factor Analysis aggregating the answers coming from a list of variables that assumes a value from 0 (meaning “never”) to 4 (meaning “always”) was used, reflecting whether the HH during the last week has felt:
- (i) cheerful and in good spirit;
- (ii) calm and relaxed;
- (iii) active and vigorous;
- (iv) fresh and rested;
- (v) that his/her life has been filled with interesting things.
Average values by subjective well-being thresholdsSubjective well-being RCI Triangle of Hope RCI Matam
Very Low 48.17 47.02Low 40.88 51.51Middle 42.59 57.55High 48.15 58.37Very High 50.21 65.49
Per
cep
tio
n• Hausman test rejects the hypothesis of endogeneity between the social
inclusion perception indicators and resilience.
• Note that social inclusion is at communitarian level and not at household level as resilience.
RC Ii =α+δ Si +δ Wi+ϑ Xi+ εi (3)
Where RC I is estimated by RIMA for the household i. Si is the vector of all the shocks experienced and reported by the households, while Wi is the vector of the two indicators of perception of social inclusion. Finally, Xi is the vector of control variables and εi the error term.
Impact of perception of social inclusion on RCITriangle of Hope Matam
Perception of social inclusion:
- services provision 1.666 0.224(0.000) (0.918)
- decision-making process 0.533 2.889(0.003) (0.000)
• Participation in the decision-making process enhances household resilience capacity in both Matam and in the Triangle of Hope;
• The community’s influence on achieving better local services has a positive and statistically significant effect in the Triangle of Hope, while in Matam the positive sign of the coefficient is not statistically significant.
Results:
24/4/2018
27
Res
ilie
nce
an
d S
oci
al P
rote
ctio
n
• The CGP is an unconditional cash transfer programme, implemented by the Ministry of Social Development (MoSD), targeting the poorest families with children in: Berea, Leribe, Mafeteng, Maseru and Qacha’s Nek
• Over four years, between 2009 and 2013, around 20,000 households received cash transfer on a regular, monthly basis
• The primary goal of the CGP was to increase well-being of children livingin the poorest households in Lesotho. Encouraged the beneficiaries tospend the received cash on the youngest
• The baseline data include information for 3,054 households• In the follow-up round only 2150 of those interviewed in baseline were
captured.• The attrition rate is equal to 6 percent
Res
ilie
nce
an
d S
oci
al P
rote
ctio
n
In this report Average Treatment Effect (ATE) has been estimated, formally:
ATE = E[𝑌1 − 𝑌0]
Yit = β0 + β1Di + β2Tt + β3 TtDi + εit
RCT DiD
CT project
LT impact on children
ST general impact on hh
Impact on resilience
24/4/2018
28
Res
ilie
nce
an
d S
oci
al P
rote
ctio
n
• Positive effects on household resilience (+2.2%);• Strong effect on food insecure (+0.8%) and borderline (+1.4%);• Stronger effect on MHH then FHH (+3.9%); and• Strong effect on labor constrained (+4.6%).
Limitation of an IE for resilience
• Results can be driven by dynamic variables; • Counfoundness with other projects;• Difficult interpretation of an increase/decrease or RCI; and• Need more evidence on how this works.
Resilience & food security
24/4/2018
29
Resili
ence &
food s
ecurityResilience index as a measure of
HH resilience to food insecurity
where,
loss = 1 if the HH suffers a loss in food security between t and t1
RCI is the HH resilience capacity index in year t
FS is the HH food security status in year t
X is a vector of HH characteristics in year t
F is CDF of the normal distribution
𝑃 𝑙𝑜𝑠𝑠𝑡−𝑡1 = 1 = Φ 𝑅𝐶𝐼ℎ,𝑡, 𝐹𝑆𝑡 , 𝐗ℎ,𝑡
𝑃 𝑟𝑒𝑐𝑜𝑣𝑡1−𝑡2 = 1 = Φ 𝑅𝐶𝐼ℎ,𝑡, 𝐹𝑆𝑡1 , 𝐗ℎ,𝑡
where,
recov = 1 if the HH recovers the loss in food security between t1 and t2
FS is the HH food security status in year t1
Resilience index and food expenditure
Probit regression: likelihood of suffering a loss and recovering
Re
sili
en
ce
& fo
od
se
cu
rity
Tanzania Uganda
(1) (2) (3) (4)
Loss btw t and t+1 Recovery btw t+1
and t+2
Loss btw t and t+1 Recovery btw t+1
and t+2
RCI -0.0389*** 0.00366 -0.856*** 0.0227***
(0.00690) (0.00504) (0.150) (0.004)
Per capita food expenditure (log) 2.348*** -1.185*** 16.86*** -0.857***
(0.164) (0.117) (2.842) (0.0665)
Female HH head 0.154** -0.0554 -0.0351 -0.0487
(0.0640) (0.0852) (0.071) (0.089)
Age of HH head 0.000274 -0.00248 0.0012 -0.0067**
(0.00180) (0.00239) (0.0022) (0.00287)
HH size 0.0890*** -0.0419 0.0434 0.0338
(0.0260) (0.0375) (0.0320) (0.0398)
Squared HH size -0.00176 0.000853 -0.000999 -0.00156
(0.00165) (0.00248) (0.00217) (0.00270)
Rural 0.346*** -0.282*** 0.348*** -0.304***
(0.0705) (0.0959) (0.086) (0.106)
Constant -5.680*** 3.344*** -1.114*** 1.201***
(0.310) (0.478) (0.240) (0.310)
Observations 2,866 1,440 2,015 1,341
Log-Likelihood -1551.561 -855.002 -1100.709 -679.7411
Pseudo-R2 0.219 0.115 0.142 0.153
Pearson Chi2
Prob > Chi2
2854.13
0.386
1447.74
0.219
2020.53
0.393
2221.08
0.000
24/4/2018
30
Resilience index and dietary diversity
Probit regression: likelihood of suffering a loss and recovering
Re
sili
en
ce
& fo
od
se
cu
rity
Tanzania Uganda
(1) (2) (3) (4)
Loss btw t and t+1 Recovery btw t+1
and t+2
Loss btw t and t+1 Recovery btw t+1
and t+2
RCI -0.0272*** 0.0108*** -0.0052** 0.0138***
(0.00344) (0.00413) (0.00253) (0.0027)
Dietary diversity 3.031*** -2.466*** 2.096*** -1.940***
(0.144) (0.172) (0.125) (0.118)
Female HH head 0.0330 0.0468 0.177** -0.157*
(0.0628) (0.0880) (0.0761) (0.0832)
Age of HH head -0.000445 0.00208 0.00144 0.00043
(0.00175) (0.00248) (0.0023) (0.0026)
HH size 0.000291 -0.0202 -0.119*** 0.0582
(0.0183) (0.0375) (0.0342) (0.0426)
Squared HH size -0.000306 0.00202 0.0049** -0.00132
(0.000902) (0.00250) (0.0022) (0.00298)
Rural 0.231*** -0.175* 0.107 0.0103
(0.0698) (0.0967) (0.0924) (0.101)
Constant -3.565*** 2.478*** -1.937*** 1.249***
(0.283) (0.476) (0.250) (0.306)
Observations 2,866 1,483 2,015 1,417
Log-Likelihood -1584.558 -842.139 -966.809 -805.850
Pseudo-R2 0.201 0.163 0.2110 0.179
Pearson Chi2
Prob > Chi2
2819.64
0.567
1559.22
0.023
2018.33
0.406
1654.08
0.000
Re
sili
en
ce
& fo
od
se
cu
rityThe role of risks
where,
all variables have the same meaning as before
S is a vector of covariant and idiosyncratic shocks that hit the HH
between t and t1
𝑃 𝑙𝑜𝑠𝑠𝑡−𝑡1 = 1 = Φ 𝑅𝐶𝐼ℎ,𝑡, 𝐹𝑆𝑡 , 𝐗ℎ,𝑡 , 𝐒ℎ,𝑡−𝑡1 , 𝑅𝐶𝐼ℎ,𝑡𝐒ℎ,𝑡−𝑡1
Probit regression with self-reported shocks (LSMS-ISA):
⟹ RCI coefficients same magnitude as in previous models, expected
signs, highly statistically significant
⟹ also other variables show the same behaviour: rural positive and
highly significant, female-headed positive and significant
⟹ self-reported shocks not statistically significant
24/4/2018
31
The role of risks: covariate shocks
Probit regression: likelihood of suffering a loss in food expenditure
Re
sili
en
ce
& fo
od
se
cu
rity
Tanzania Uganda
Loss btw t and t+1 dx/dy Loss btw t and t+1 dx/dy
RCI -0.056*** -0.012*** -0.873*** -0.265***
(0.011) (0.152)
Conflict intensity index -0.233 -0.023 0.00288 0.0005
(0.242) (0.0137)
Rainfall CV -5.468** -0.525** -2.801 0.0118
(2.206) (3.624)
RCI * Conflict intensity index 0.0021 -5.72e-05
(0.003) (0.000209)
RCI * Rainfall CV 0.0544* 0.0695
(0.032) (0.0743)
Per capita food expenditure (log) 2.474*** 16.87***
(0.180) (2.867)
Female HH head 0.176** -0.0351
(0.0692) (0.0713)
Age of HH head 0.0010 0.000929
(0.0019) (0.00221)
HH size 0.10*** 0.0424
(0.029) (0.0323)
Squared HH size -0.0023 -0.000966
(0.0018) (0.00218)
Rural 0.310*** 0.362***
(0.0764) (0.0925)
Constant -4.006*** -0.492
(0.759) (0.839)
The role of risks: covariate shocks
Probit regression: likelihood of suffering a loss in dietary diversityR
esili
en
ce
& fo
od
se
cu
rity
Tanzania Uganda
Loss btw t and t+1 dx/dy Loss btw t and t+1 dx/dy
RCI -0.031*** -0.0075*** -0.008 -0.0015**
(0.0079) (0.0169)
Conflict intensity index 0.193 0.013 0.00843 0.002
(0.223) (0.017)
Rainfall CV -1.214 0.181 -2.944 -0.609
(2.046) (3.761)
RCI * Conflict intensity index -0.0024 1.21e-05
(0.0027) (0.00025)
RCI * Rainfall CV 0.0312 0.0131
(0.0293) (0.0753)
Dietary diversity 2.987*** 2.110***
(0.155) (0.127)
Female HH head 0.0238 0.178**
(0.067) (0.0766)
Age of HH head -0.00042 0.00159
(0.00188) (0.00233)
HH size 0.00194 -0.121***
(0.0192) (0.0344)
Squared HH size -0.0003 0.00501**
(0.0008) (0.00226)
Rural 0.238*** 0.139
(0.0755) (0.0992)
Constant -3.610*** -1.319
(0.722) (0.865)
24/4/2018
32
Wh
at y
ou
can
get
fro
m R
IMA
An infographic visually explains the process step by step and the
results of the analysis.
Evidence-based policy choice
A brief addressed to government policy-makers summarizes the results of the resilience analysis and formulates policy recommendations.
Cas
e st
ud
ies
Resilience analysis Karamoja (Uganda)
• The region of Karamoja, located in the northeast of Uganda, is the poorest and least developed region in the country
• In 2015, UNICEF, FAO and WFP developed a resilience strategy for the region in order to improve food security and nutrition
• RIMA-II was adopted as the tool to measure resilience capacity to food insecurity in the region
• Results and policy recommendations have been published in a resilience analysis report. They can also be visualized through an interactive infographic
In collaboration with IGAD, the Office of the Prime Minister, the Bureau of Statistics, UNICEF &WFP, FAO is conducting a food security and resilience analysis among refugee and host community households in North Uganda.
24/4/2018
33
RIM
A a
nal
ysi
s in
th
e w
orl
d
RIMA – Finalized AnalysisSenegal, Niger, Burkina Faso, Mali, Sudan, South Sudan, Kenya, Somalia, West Bank and Gaza Strip, Nigeria, Uganda, Tanzania and Malawi
RIMA – Ongoing AnalysisSenegal, Mauritania, Chad, Ethiopia, Lesotho, West Bank and Gaza Strip
Mauritania
Chad
Ethiopia
Tanzania
Malawi
Lesotho
Senegal
MaliNiger
Sudan
South Sudan
Kenya
Nigeria
West Bank & Gaza Strip
Stepping up policy influence
Photo: FAO
24/4/2018
34
Ste
pp
ing
up
po
licy
in
flu
ence
• Resilience markerpilot in West Bank and Gaza Strip
• Integration/harmonization with other toolse.g. USAID-TANGO, UNICEF, WFP, IFAD
• Global resilience indexe.g. future development for global comparison
• C-RIMApilot in Somalia
• Broadening RIMA analytical capacitiese.g. gender (FAO); shocks (IFPRI; Cornell and TUFTS University)
Ste
pp
ing
up
po
licy
in
flu
ence
• Strict collaboration with:1) Regional initiatives (CILSS/IGAD)
2) National Bureau of Statistics and other significative ministries
3) FAO country regional and sub-regional offices
4) Universities for enhancing local capacity building
• Re-thinking resilience analyses and communications tools under a policy-oriented perspective
24/4/2018
35
THANK YOU!
Contact me…
… and sign up to our newsletter!
www.fao.org/resilience/background/tools/rima