VOICES HUNGRY Global monitoring of Food Insecurity through experience-based indicators of the FAO Statistics Division
Jan 28, 2016
VOICESHUNGRY
Global monitoring of Food Insecurity
through experience-based indicators
of the
FAO Statistics Division
Introduction
Background information
Concepts
Practical implementation
Expected outcomes
Discussion
Overview of the presentation• Introduction• Background • Concepts• Practice• Outcomes• Discussion
The establishment of a global system allowing fine-grain and timely monitoring of Food Insecurity can no longer be postponed Inequalities within countries, rather than
differences across countries are increasingly recognized as being at the heart of development problems for the near future
• Macro-level national data no longer suffices to understand evolution and to guide intervention
The evolution of globalization and the increased volatility of financial and commodity markets call for increased timeliness of all monitoring efforts
Food Security Measurement• Introduction• Background • Concepts• Practice• Outcomes• Discussion
We need to ensure a globally valid standard to allow proper comparison of situations across countries and across social groups
Current practice with traditional household surveys is problematic for cost effective rapid and consistent monitoring worldwide
Innovation in direct data collection and information processing methods provides the basis for significant improvements • Gallup World Poll is one example of a feasible, global
survey conducted annually with common methodology
• Sophisticated use of the theory of latent trait measurement allows defining a common metric
Food Security Measurement• Introduction• Background • Concepts• Practice• Outcomes• Discussion
We all think we know what food insecurity is, yet operationalizing the definition has been overly challenging, and the results thus far are unsatisfactory Only two official indicators for global monitoring:
• MDG 1.8 (prevalence of children underweight) and
• MDG 1.9 (prevalence of people below minimum level of dietary energy consumption)
They tend to be broadly misinterpreted (i.e., they have been expected to capture the impact of the food price crisis)
It is difficult to make full sense of trends and correlations (i.e., with poverty trends, or economic growth)
What is Food Insecurity?• Introduction• Background • Concepts• Practice• Outcomes• Discussion
The current WFS “mantra” is of limited practical relevance for measurement Literally hundreds of indicators have been
proposed, tested, used, and all criticized for failing to provide a comprehensive picture of food insecurity
Compilation of aggregate indexes (ex. IFPRI’s GHI or the EIU Global Food Security Index) or of dashboards with multiple indicators is problematic• Arbitrariness of weights • Difficult interpretation
Very little is being said in terms of what should be meant by the term insecurity• Is it a feeling? A risk? How does variability of the
past convey information on the prospect for the future?
What is Food Insecurity?• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Dr. Kathy L. Radimer’s Ph.D. dissertation: “Understanding Hunger and Developing Indicators to Assess It”, Cornell University, August 1990
“The lack of an operational definition for hunger has been frequently cited as a barrier to
progress in addressing the problem.”
“Three scales, one each for household, women’s, and children’s hunger, emerged and were found to be valid and reliable indicators for measuring
hunger directly”
(Radimer et al., 1992)
Establishes the concept of food insecurity as an experiential construct
A new avenue opened in the 1990s• Introduction
• Background • Concepts• Practice• Outcomes• Discussion
Quite a history since• Introduction• Background • Concepts• Practice• Outcomes• Discussion
FAO Food Insecurity Experience Scale
consequences
The concept• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Food security
Food insecurity
mild moderate severe
Worries Compromising food quality and variety HungerCompromising
food quantity
Undernutrition (stunting, wasting)
Welfare reduction(Psychological
costs, reduction of other essential
expenses)
Malnutrition (obesity,
micronutrient deficiencies, reduced work
capacity)
StarvationWellbeing
The FIES: a set of questions spanning the range of experiences
experiences
Assuming that: there exist a latent (unobservable) characteristic
or feature to be measured, and that outcomes can be observed that depend on the
latent trait A model can be defined for the probability
of occurrence of the observable outcome as a function of the unobservable trait
A sound statistical procedure can be established to estimate the value of the latent trait for a subject
• Econometrics: discrete choice models in willingness to pay studies (McFadden)
• Psychometrics: Item Response Theory• Food Security measurement (Hamilton et al., Nord)
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Albeit not directly observable, a measure of severity of food insecurity is assumed to exist as a number defined on the real line (a “latent” trait)
Conditions denoting increasing levels of food insecurity severity can be defined (i.e., “items” along the scale)
Individuals are characterized as to possess levels of food insecurity on the same scale
The severity of the items, the location of the individual, and the thresholds to classify respondents can be estimated from the data on the responses to the same set of questions by a (numerous) sample of respondents
The percentages of cases classified in each group are indicators of the prevalence of food insecurity in the population at different levels of severity
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion Respondent h
Thresholds
min maxItem 3Item 2 Item … Item nItem 4Item 1
Item Response Theory
The probability of correctly responding to a test item is a function of the severity of the item and of the competence of the respondent
A dataset of responses to the same set of test items can be used:• To validate the test,
• By measuring the level of difficulty of each item
• By testing the overall fit of the observed response patterns
• To assess each respondent’s ability
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion
0
1
0.5
Item i h
Pr(“yes”;h,j)
Item j
Pr(“yes”;h,i,)
How often respondents answer “Yes” (and, perhaps, to which questions) will be used to establish their position on the scale of food insecurity severity and to classify them as mild, moderate ore severely food insecure
The analytics• Introduction• Background • Concepts• Practice• Outcomes• Discussion
“During the last 12 months, was there a time when, because of lack of money or other resources:
1. You were worried you would not have enough food to eat?
2. You were unable to eat healthy and nutritious food? 3. You ate only a few kinds of foods? 4. You had to skip a meal?5. You ate less than you thought you should?6. Your household ran out of food?7. You were hungry but did not eat?8. You went without eating for a whole day?”
The (current version of) FIES• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Results from 4 pilots in Africa• Introduction• Background • Concepts• Practice• Outcomes• Discussion
AngolaSeverity Errors Infit Outfit
WORRIED -1.59 0.14 1.18 1.46HEALTHY -0.76 0.12 0.86 1.04FEWFOOD -0.37 0.11 1.07 1.28SKIPPED -0.24 0.11 0.87 0.74ATELESS -0.44 0.11 0.92 0.81RUNOUT 0.25 0.11 0.86 0.72HUNGRY 0.87 0.11 1 0.98WHLDAY 2.27 0.13 1.18 1.63
ValuesMean 0St. Dev. 1.09N (complete, non extreme) 505
EthiopiaSeverity Errors Infit Outfit
WORRIED -0.75 0.11 1.4 2.25HEALTHY -2.17 0.13 0.92 2.19FEWFOOD -2.63 0.14 0.96 1.11SKIPPED -0.02 0.11 0.74 0.71ATELESS -0.6 0.11 0.84 0.85RUNOUT 1.36 0.12 0.76 0.69HUNGRY 1.64 0.13 0.84 0.8WHLDAY 3.16 0.17 1.1 3.7
ValuesMean 0St. Dev. 1.84N (complete, non extreme) 597
MalawiSeverity Errors Infit Outfit
WORRIED -0.49 0.13 1.11 1.17HEALTHY -0.36 0.13 1.16 1.36FEWFOOD -1 0.14 0.84 0.62SKIPPED 0.22 0.12 0.97 0.88ATELESS -0.7 0.14 0.94 0.97RUNOUT 0.12 0.12 0.9 0.81HUNGRY 0.54 0.12 1.03 1.05WHLDAY 1.67 0.12 1.04 1.14
ValuesMean 0St. Dev. 0.79N (complete, non extreme) 423
NigerSeverity Errors Infit Outfit
WORRIED -1.37 0.11 1.24 1.62HEALTHY -0.64 0.1 0.92 0.76FEWFOOD -1 0.11 1.01 0.9SKIPPED 0.8 0.09 1.06 1.05ATELESS -0.49 0.1 0.98 1.02RUNOUT 0.46 0.09 0.79 0.68HUNGRY 0.3 0.09 0.82 0.75WHLDAY 1.95 0.1 1.21 1.33> round(tab2.Nig,2)
ValuesMean 0St. Dev. 1.02N (complete, non extreme) 734
Results from 4 pilots in Africa• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Results from 4 pilots in Africa• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Marginally food insecure
Moderately food insecure
Severely food insecure
(-1.5) (0.0) (1.5)Angola 89.7% 59.9% 22.6%Ethiopia 90.9% 44.6% 7.6%Malawi 87.1% 66.5% 36.8%Niger 89.2% 61.7% 26.3%
(-1.0) (0.0) (2.0)Angola 82.5% 59.9% 13.8%Ethiopia 79.1% 44.6% 2.9%Malawi 81.8% 66.5% 27.6%Niger 82.4% 61.7% 17.2%
(-1.0) (0.5) (2.0)Angola 82.5% 46.6% 13.8%Ethiopia 79.1% 28.6% 2.9%Malawi 81.8% 56.9% 27.6%Niger 82.4% 49.3% 17.2%
The challenge is to use the information provided by the answers to the FIES questionnaires to classify cases into food insecurity classes in a way that is meaningful and comparable over time and across countries and socioeconomic contexts.
Doing so requires: (a) establishing the metric equivalence of the scale and (b) classifying cases into food insecurity categories, taking into account possible differences in the severity of some items in some of the countries.
We are working with Mark Nord (USDA/ERS) on statistical methods for equating severity classifications across cultures.
Equating the scale (continued) • Introduction
• Background • Concepts• Practice• Outcomes• Discussion
Experiences, not “opinions”
Typical experiences are invariably associated with food insecurity• The challenge is to identify the crucial ones
that are meaningful in most cases
Self-reported, not “subjective”
We do not ask if people perceive themselves as being food insecure
Self reporting is not to be necessarily considered less reliable than other ways of collecting information• Ex.: Expenditure, Income, Employment, etc.
A few misconceptions• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Linguistic and cultural adaptation: Getting the wording right
• Inadequate linguistic adaptation and translation of the scale questions may cause their meaning and how they are understood by respondents to differ across cultural and linguistic groups
Proper analytic treatment of data• There are most likely true differences in the way
people experience food insecurity, suggesting that a unique scale may not work universally even if the translation is accurate and the analysis is appropriate
• Items may not be equally discriminating on all context
Classifying respondents’ food security status using the raw score may not be the most appropriate approach• Use probabilistic assignment of cases to food
security classes
The Challenges• Introduction• Background • Concepts• Practice• Outcomes• Discussion
The probability of reporting on a certain food insecurity experience is a function of the severity of the experience and of the level of food insecurity of the respondent
A dataset of responses to the same questionnaire by a sample of individuals can be used: To validate the questionnaire, by revealing
the severity associated to each experience, in the specific context
To locate each respondent’s food insecurity level on the scale spanned by the set of experiences
To link food insecurity to other respondent characteristics (an opportunity still largely unexplored)
Experience-based food security• Introduction
• Background • Concepts• Practice• Outcomes• Discussion
To include a Food Insecurity Experience Scale in the Gallup World Poll
Using a sound and common survey methodology
On nationally representative samples With face-to-face or phone interviews,
conducted in the respondents’ preferred language After proper cultural and linguistic adaptation Covering 150+ countries, including all crucial
emerging and developing countries
To use the data collected to inform the compilation of a set of food security indicators at country level
The Voices of the Hungry project• Introduction
• Background • Concepts• Practice• Outcomes• Discussion
A valid measurement tool produces measures that are both right (that is, it produces on average the correct measure over repeated applications) and precise (each of the produced measures is quite close to the true magnitude of the “thing” that one aims at measuring).
While it is desirable that measures be as precise as possible, excessive focus on precision may sometimes lead to the risk of preferring instruments that are … “precisely wrong” over those that are “approximately right,” a risk that we propose should be adamantly avoided.
Cross culture validity of the FIES• Introduction• Background • Concepts• Practice• Outcomes• Discussion
While ensuring that people understand the questions as they are intended, through careful linguistic adaptation, it may be that food insecurity is not experienced in exactly the same way around the world. Some items may indicate more severe food insecurity in one location compared to another.
Differences in item severity across countries do not necessarily imply rejection of the underlying concept of an experience-based scale or of the capacity to compare prevalence rates based on such measures across countries.
Comparability of the measures
Equating the scale
• Introduction• Background • Concepts• Practice• Outcomes• Discussion
Validated Food Insecurity Experience Questionnaires in local languages to be used in all countries in the World to
collect the data to inform the new indicators on the severity of food insecurity
Annually produced datasets on Food insecurity experiences of individuals from 140 countries as well as country-level indicators on the severity of food insecurity made publicly available through web based
Food Security portals and official publications Research opportunities using the entire
GWP datasets
What the VOH will produce • Introduction• Background • Concepts• Practice• Outcomes• Discussion
1) Data on gender equality and empowerment in relation to food security;
2) Capacity develop for sub-national nutrition and food security planning and programming through better data; and
3) Monitoring hunger reduction and the human right to adequate food.
VOH will accomplish this both by collection of annual data through the Gallup World Poll and in collaboration with national institutions to promote for inclusion of the FIES in national surveys.
Contributions to development• Introduction
• Background • Concepts• Practice• Outcomes• Discussion
Within the GWP, food insecurity questions will be asked in reference to individual respondents rather than households
This provides valuable insights into differences in the lived experience of food insecurity between men and women
Important for advocating for better efforts to empower women with respect to
• land holding, • employment and income, • household budgetary decision making, and • political involvement
All of which have implications for nutrition as well as food security
Importance of looking at individualsIntroduction
Background ConceptsPracticeOutcomesDiscussion
The VOH promotes use of the FIES in national surveys designed to allow disaggregation at sub national levels
Thus enabling governments to produce their own statistics as part of national food security and nutrition information systems
This information contributes to better understanding of characteristics that increase vulnerability to food insecurity as well as a range of outcomes with potentially detrimental effects for individuals and for society as a whole
Consistency of the method used by FAO globally and by national governments will promote comparability of results between the national-level data collected through the Gallup World Poll and more in-depth food security analyses within countries
Capacity developmentIntroductionBackground ConceptsPracticeOutcomesDiscussion
The VOH data, provided annually, can play a significant role in monitoring of the Zero Hunger Challenge and food security targets considered for the post 2015 Development Agenda.
The percentage of people experiencing severe food insecurity could form the basis for monitoring progress towards a new food security target in the post 2015 development agenda
Promote use of FIES by local governments, non-governmental organizations and advocacy groups to monitor food insecurity locally or regionally, engaging diverse stakeholders in the process, and building bridges between people of different backgrounds
This may in fact be where its greatest potential lies to effect change and contribute to guaranteeing the human right to adequate food
Advocacy Introduction
Background ConceptsPracticeOutcomesDiscussion
Thanks!
http://www.fao.org/economic/ess/ess-fs/voices/en/