15 September 2021. In 2013, FAO launched the “Voices of the Hungry” project, establishing a new globally valid tool called the Food Insecurity Experience Scale (FIES). The latter provides an approach for measuring the severity of people’s food insecurity condition by looking at their own experiences, allowing to hear the voices of the people who struggle daily to have access to safe and nutritious food.
The objective of this RUFORUM webinar was to introduce FIES as a tool for measuring food insecurity at different levels and raise awareness among the participants on FAO’s work linked to food security data and the SDGs.
2. Measuring food insecurity in the SDG era
• Where we were, back in 2013
• Where are we now
• How did we get here
• What is the FIES and why it stands out compared to other
household food insecurity measurement systems/tools
• What still needs to be done
4. “While the last 20 years have seen a deepening understanding of the
concept of food security, its measurement has lagged behind.
At the global level, there are no direct estimates of the number of
food insecure people. The most widely-cited indirect measure is the
‘prevalence of undernourishment’ (POU), constructed by the United
Nations Food and Agriculture Organisation (FAO)[…]
These estimates give no sense of the severity of hunger”
HLPE, 2012. Social protection for food security.
A report by the High Level Panel of Experts on Food Security and Nutrition
of the Committee on World Food Security, Rome 2012.
12. From PoU to FIES
• PoU was clearly inadequate, as it can hardly be sufficiently timely,
reliable and disaggregated as needed for the ambitious SDG universal
agenda and for policy guidance
• Improving it, as suggested/implied by Bill Gates, is impossible
• Among existing alternative indicators, many were equally problematic
• The best option seemed to reside in the concept of experience-based
food security measurement
• In 2014 we had the opportunity to collect data worldwide, which
allowed us to develop and validate the FIES
13. Problems with the PoU
• It depends on aggregate food supply data and on infrequent,
heterogenous, household consumption surveys
• Food consumption in caloric term at the individual level
• is estimated only as an approximation
• cannot be matched to the individual characteristics that determine energy
requirements
• It cannot be reliable when the prevalence of undernourishment is
relatively low (5% or less)
• Was it appropriate as an indicator for the MDG Goal
14. Problems with alternative indicators
• Food Consumption Score
• Not clear what it measures. Concept of “food consumption” is very vague (is it
quantity? Quality? Diversity?)
• Problems in comparability over time/across countries
• Coping Strategy Index
• Reflect intensity of shock and coping ability in a way that is difficult to
separate
• Share of food expenditure
• Needs very detailed data to be reliable. Not clear what the optimal value
should be. Difficult to compare across countries.
15. What is the FIES and why it stands
out compared to other household
food insecurity measurement
systems/tools
16. The concept of experience-based food
security measurement
• Food insecurity as a measurable “latent trait”
• The severity level is linked to possibly different conditions or
experiences
• By asking people to report the occurrence of those experiences, we
can infer the position of the respondent on the same scale of severity
• Advantages compared to
17. The Food Insecurity Experience Scale
Food
security
Food insecurity
mild moderate severe
Undernutrition
(stunting, wasting)
Welfare reduction
(Psychological costs,
reduction of other
essential expenses)
Malnutrition
(obesity,
micronutrient
deficiencies,
reduced work
capacity)
Starvation
Wellbeing
The FIES: a set of questions spanning the range of experiences
Worries
Compromising food
quality and variety
Hunger
Compromising
food quantity
18. The Food Insecurity Experience Scale survey
module
During the last 12 MONTHS, was there a time when, because of a lack of
money or other resources:
1. You were worried you would run out of food? [YES/NO/DK]
2. You were unable to eat healthy and nutritious food? [YES/NO/DK]
3. You ate only a few kinds of foods? [YES/NO/DK]
4. You had to skip a meal? [YES/NO/DK]
5. You ate less than you thought you should? [YES/NO/DK]
6. Your household ran out of food? [YES/NO/DK]
7. You were hungry but did not eat? [YES/NO/DK]
8. You went without eating for a whole day? [YES/NO/DK]
19. The Food Insecurity Experience Scale survey
module – extended version
During the last 12 MONTHS, was there a time when, because of a lack of money or
other resources:
1. You were worried you would run out of food? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
2. You were unable to eat healthy and nutritious food? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
3. You ate only a few kinds of foods? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
4. You had to skip a meal? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
5. You ate less than you thought you should? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
…
20. The Food Insecurity Experience Scale survey
module
During the last 12 MONTHS, was there a time when, because of a lack of
money or other resources:
…
6. Your household ran out of food? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
(if YES): Was it rarely [only 1 or 2 times], sometime [YES/NO]
7. You were hungry but did not eat? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
(if YES): Was it rarely [only 1 or 2 times], sometime [YES/NO]
8. You went without eating for a whole day? [YES/NO/DK]
(If YES): Was it during the last 4 weeks? [YES/NO]
(if YES): Was it rarely [only 1 or 2 times], sometime [YES/NO]
21. Household Food
Insecurity
Access Scale
HFIAS
U.S. Household Food Security
Survey Module
USA, 1995; Canada, 2004
ELCSA
Guatemala, 2011
EMSA
Mexico, 2008
EBIA
Brazil, 2004
FIES
A global reference
standard
Colombia Venezuela
FIES genealogy
HHS
22. The Rasch model
𝑃𝑟𝑜𝑏 𝑋𝑖,𝑗 = 1 =
exp 𝑎𝑖 − 𝑏𝑗
1 + exp 𝑎𝑖 − 𝑏𝑗
• The probability to affirm an item is increasing in the distance between the
item and the respondent
• Example:
• The more “competent” is a student, the more likely it is that she will answer correctly any
item
• “Easier” items will be answered correctly more often than “difficult” ones
• Measures are defined/produced on an interval scale, not a ratio one
• As the probability depends only on the difference between measures, the model is
defined up to an arbitrary constant
• Maximum likelihood principles can be applied to estimate the values of 𝑎
and 𝑏, given a set of data {𝑥}.
24. yes yes yes yes yes no no no
less severe
(a – b) > 0
Prob(“yes”) > 0.5
more severe
(a – b) < 0
Prob(“yes”) < 0.5
a
b4
b5
b3
b2
b1 b6 b7 b8
𝑃𝑟𝑜𝑏 𝑋𝑖,𝑗 = 1 =
exp 𝑎𝑖 − 𝑏𝑗
1 + exp 𝑎𝑖 − 𝑏𝑗
25. yes yes no no no no no no
b4
b5
b3
b2
b1 b6 b7 b8
What if we only observe the responses?
(making inference on the unobservable position of the respondent, based on the observed responses)
a
The most likely position is the one that maximizes the joint probability of responses
(an application of the maximum likelihood principle)
26. Computing a FIES-based indicator
𝑃𝑇
𝑖
= 1 − Φ 𝑇 ; 𝜇 = 𝑎𝑖; 𝜎 = s.e. 𝑎𝑖
𝐹𝐼𝑇 =
𝑖=0
𝐾
𝑃𝑇
𝑖
× 𝑤𝑁𝑖/𝑁
where:
𝑃𝑇
𝑖
= Probability of being food insecure, given a certain threshold of severity, 𝑇.
F(.) = Normal Distribution function
wNi/N = weighted proportion of respondents in a representative sample of the population, with
raw score RS = i, (with i from 0 to K, the number of valid items)
𝑎𝑖 = Rasch model respondent parameters for a respondent with raw-score RS = i
T = severity threshold to define the class of food insecurity
27.
28. Two conceptually separate steps
• Measuring the probability of being food insecure for each respondent
• Requires a statistical analysis of the data to transform the “YES/NO” answers
into a probabilistic measure (for which there is a point estimate and a margin
of error)
• Computing the prevalence of food insecurity in the population
• It is obtained as the weighted average of the probability of being food
insecure, across the entire sample
• Use of sampling and post-stratification weights are essential to refer the
estimates to the reference population
29. Estimating the Rasch Model
• RM.weights package in R
• Raschtest (http://fmwww.bc.edu/RePEc/bocode/r) +
fitstat_ers (http://fmwww.bc.edu/RePEc/bocode/f) in STATA
• Rasch extensions in SPSS
30. Uses of the FIES
1. SDG monitoring
2. Short term, quick assessments of recent food insecurity in periodic surveys (e.g., in IPC acute
food insecurity analyses)
3. Detailed, disaggregated assessment of the state of food insecurity at country level, for internal
policy monitoring and guidance (e.g., by urban/rural, by regions, by socio-demographic group,
etc.)
4. Assessment of the impact of shocks on the food insecurity status of specific population groups
in specific areas (e.g., for the impact of COVID)
5. Monitoring and Evaluation of projects that intend to improve food security (e.g., within GAFSP
M&E system)
6. Projection of food insecurity in the future (e.g., in the “world hunger clock”
https://worldhunger.io/ )
7. Research on the contribution of access to food to nutrition (e.g., the many dose-response
analyses often referred to by Perez-Escamilla and others)
8. Use within multi-dimensional poverty assessments to capture the food deprivation dimension
(e.g., as in what CONEVAL does in Mexico with the EMSA)
9. …
32. Major improvements
• Training analysts on the principles of measurement for social sciences
• Highlight the importance of proper assessment of measurement error
• Promote better statistical literacy (sampling theory, estimation,
statistical inference)