Indicator 2.1.2 Prevalence of moderate or severe food insecurity in the population, based on the Food Insecurity Experience Scale (FIES)

Indicator 2.1.2: Prevalence of moderate or severe food insecurity in the population, based on the Food Insecurity Experience Scale (FIES)

Target 2.1: By 2030, end hunger and ensure access by all people, in particular the poor and people in vulnerable situations, including infants, to safe, nutritious and sufficient food all year round

Food and Agriculture Organisation of the United Nations (FAO)

Definition:

The indicator measures the percentage of individuals in the population who have experienced food insecurity at moderate or severe levels during the reference period. The severity of food insecurity, defined as a latent trait, is measured on the Food Insecurity Experience Scale global reference scale, a measurement standard established by FAO through the application of the Food Insecurity Experience Scale in more than 140 countries worldwide, starting in 2014.

Concepts:

Extensive research over more than 25 years has demonstrated that the inability to access food results in a series of experiences and conditions that are fairly common across cultures and socio-economic contexts and that range from being concerned about the ability to obtain enough food, to the need to compromise on the quality or the diversity of food consumed, to being forced to reduce the intake of food by cutting portion sizes or skipping meals, up to the extreme condition of feeling hungry and not having means to access any food for a whole day. Typical conditions like these form the basis of an experience-based food insecurity measurement scale. When analysed through sound statistical methods rooted in Item Response Theory, data collected through such scales provide the basis to compute theoretically consistent, cross country comparable measures of the prevalence of food insecurity. The severity of the food insecurity condition as measured by this indicator thus directly reflects the extent of households’ or individuals’ inability to regularly access the food they need.

Prevalence of food insecurity: Percent (%)

Number of food insecure people: Millions (of people)

Data can be collected using the Food Insecurity Experience Scale survey module (FIES-SM) developed by FAO, or any other experience-based food security scale questionnaires, including:

• the Household Food Security Survey Module (HFSSM) developed by the Economic Research Service of the US Department of Agriculture, and used in the US and Canada,
• the Latin American and Caribbean Food Security Scale (or Escala Latinoamericana y Caribeña de Seguridad Alimentaria – ELCSA), used in Guatemala and tested in several other Spanish speaking countries in Latin America,
• the Mexican Food Security Scale (or Escala Mexicana de Seguridad Alimentaria, - EMSA), an adaptation of the ELCSA used in Mexico,
• the Brazilian Food Insecurity Scale (Escala Brasileira de medida de la Insegurança Alimentar – EBIA) used in Brazil, or
• the Household Food Insecurity Access Scale (HFIAS),

or any adaptation of the above that can be calibrated against the global FIES.

Two versions of the FIES-SM are available for use in surveys of individuals or households respectively, and the difference stands in whether respondents are asked to report only on their individual experiences, or also on that of other member of the household.

The current FIES-SM module include eight questions as in the table below.

The questions should be adapted and administered in the respondents’ preferred language and enumerators instructed to make sure that respondents recognize the reference period and the qualifier according to which experiences should be reported only when due to “lack of money or other resources” and not, for example, for reasons related to health or other cultural habits (such as fasting for religious credos).

The FIES-SM can be included in virtually any telephone-based or personal interview-based survey of the population, though face to face interview is preferred.

Since 2014, the individual referenced FIES-SM is applied to nationally representative samples of the population aged 15 or more in all countries covered by the Gallup World Poll (more than 140 countries every year, covering 90% of the world population). In most countries, samples include about 1000 individuals (with larger samples of 3000 individuals in India and 5000 in mainland China).

Additionally to the GWP, in 2020 FAO collected data in 20 countries through Geopoll® with the specific objective of assessing food insecurity during the COVID-19 pandemic. The countries covered were: Afghanistan, Burkina Faso, Cameroon, Central African Republic, Chad, Democratic Republic of the Congo, El Salvador, Ethiopia, Guatemala, Haiti, Iraq, Liberia, Mozambique, Myanmar, Niger, Nigeria, Sierra Leone, Somalia, South Africa and Zimbabwe. For all these countries, the 2020 assessment was based on Geopoll data.

Other national surveys exist that already collect FIES compatible data.

For Afghanistan, Angola, Armenia, Botswana, Burkina Faso, Cabo Verde, Canada, Chile, Costa Rica, Ecuador, Fiji, Ghana, Greece, Grenada, Honduras, Indonesia, Israel, Kazakhstan, Kenya, Kiribati, Kyrgyzstan, Lesotho, Malawi, Mauritania, Mexico, Morocco, Namibia, Niger, Nigeria, Palestine, Philippines, Republic of Korea, Russian Federation, Saint Lucia, Samoa, Senegal, Seychelles, Sierra Leone, South Sudan, Sudan, Tonga, Uganda, United Republic of Tanzania, United States of America, Vanuatu, Viet Nam and Zambia, national government survey data were used to calculate the prevalence estimates of food insecurity by applying FAO’s statistical methods to adjust national results to the same global reference standard, covering approximately a quarter of the world population. Countries are considered for the year/years when national data are available, informing the regional and subregional aggregates assuming a constant trend in the period 2014–2020, or integrating the remaining years with GWP or Geopoll data in case they were compatible. Exceptions to this rule are: Armenia, Botswana, Burkina Faso, Chile, Costa Rica, Ecuador, Ghana, Honduras, Indonesia, Israel, Malawi, Namibia, Niger, Nigeria, Sierra Leone, Uganda and Zambia. In these cases, the following procedure was followed:

• Use national data collected in one year to inform the corresponding year.
• For the remaining years, apply the smoothed trend coming from the data collected by FAO through the Gallup© World Poll to the national data to describe evolution over time. Smoothed trend is computed by taking the mean of the rates of change between consecutive three-year averages.

The motivation behind this procedure was the strong evidence found in support of the trend suggested by data collected by FAO (for instance, evolution of poverty, extreme poverty, employment, food inflation, among others), allowing to provide a more updated description of the trend in the period 2014–2020.

In Indonesia, Kazakhstan, Kyrgyzstan, Mauritania, Nicaragua, Paraguay, Rwanda, Seychelles, Sudan and United Republic of Tanzania, due to lack of data in 2020, the corresponding subregional trend between 2019 and 2020 was used to inform 2020.

Obtaining internationally comparable data for global monitoring:

To ensure comparability of the FImod+sev and FIsev indicators computed for different populations, universal thresholds are defined on the FIES global reference scale and converted into corresponding values on the “local” scales obtained as a result of application of the Rasch model on any specific population, through a process of “equating”.

Equating is a form of standardization of the metric based on identification of the subset of items that can be considered common to the global FIES and the specific scale used for measurement in each context. The severity levels associated with the common items are used as anchoring points to adjust the global FIES thresholds to the local scales. The standardization process ensures that the mean and standard deviation of the set of common items is the same when measured on the global FIES or on the national scale. Compatibility with the global FIES and the possibility to compile this indicator requires that at least four of the eight FIES items are identified as common.

The Statistics Division at FAO has developed the RM.weights package under R, which provides routines for estimating the parameters of the Rasch model using conditional maximum likelihood, with the possibility to allow for the complex survey design.

National data providers will be the National Statistical Authorities that are responsible for the survey in which the FIES or similar scale is included. FAO will provide data for countries where the FIES or compatible module is not included in any national survey.

An average of less than three minutes of survey time is estimated to collect FIES data in a well-conducted face-to-face survey, which should make it possible to include the FIES-SM in a nationally representative survey in every country in the world, at a very reasonable cost. FAO provides versions of the FIES-SM adapted and translated in each of the more than 200 languages and dialects used in the Gallup World Poll.

When used in the Gallup World Poll, with sample sizes of only about 1000 individuals, the width of confidence intervals rarely exceeds 20% of the measured prevalence (that is, prevalence rates of around 50% are estimated with margins of errors of plus or minus 5%). Obviously, confidence intervals are likely to be much smaller when national prevalence rates are estimated using larger samples.

Compared to other proposed non-official indicators of household food insecurity, the FIES based approach has the advantage that food insecurity prevalence rates are directly comparable across population groups and countries. Even if they use similar labels (such as “mild”, “moderate” and “severe” food insecurity) other approaches have yet to demonstrate the formal comparability of the thresholds used for classification, due to lack of the definition of a proper statistical model that links the values of the “indexes” or “scores” used for classification, to the severity of food insecurity. For this reason, care should be taken when comparing the results obtained with the FIES with those obtained with these other indicators, even if, unfortunately, similar labels are used to describe them.

Data at the individual or household level is collected by applying an experience-based food security scale questionnaire within a survey. The food security survey module collects answers to questions asking respondents to report the occurrence of several typical experiences and conditions associated with food insecurity. The data is analysed using the Rasch model (also known as one-parameter logistic model, 1-PL), which postulates that the probability of observing an affirmative answer by respondent i to question j, is a logistic function of the distance, on an underlying scale of severity, between the position of the respondent, $a_i$, and that of the item, $b{}_j$.

$Prob\left\{X_{i,j}=Yes\right\}=\frac{\exp \left(a_i-b_j\right)}{1+\exp \left(a_j-b_j\right)}$

Parameters $a_i$ and $b_j$ can be estimated using maximum likelihood procedures. Parameters $a_i$, in particular, are interpreted as a measure of the severity of the food security condition for each respondent and are used to classify them into classes of food insecurity.

The FIES considers the three classes of (a) food security or mild food insecurity; b) moderate or severe food insecurity, and (c) severe food insecurity, and estimates the probability of being moderately or severely food insecure ($p_{mod+sev}$) and the probability of being severely food insecure ($p_{sev}$) for each respondent, with $0\<p_{sev}\<p_{mod+sev}\<1$. The probability of being food secure or mildly food insecure can be obtained as $p_{fs}=1-p_{mod+sev}$.

Given a representative sample, the prevalence of food insecurity at moderate or severe levels (FImod+sev), and at severe levels (FIsev) in the population are computed as the weighted sum of the probability of belonging to the moderate or severe food insecurity class, and to the severe food insecurity class, respectively, of all individual or household respondents in a sample:

$\left(1\right)\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }F{I}_{mod+sev}=\mathrm{\ }{\sum }_i{p_i}_{mod+sev}\times w_i$

and

$\left(2\right)\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }\mathrm{\ }F{I}_{sev}=\mathrm{\ }{\sum }_i{p_i}_{sev}\times w_i$

where $w_i$ are post-stratification weights that indicate the proportion of individual or households in the national population represented by each element in the sample.

It is important to note that if $w_i$ are individual sampling weights, then the prevalence of food insecurity refers to the total population of individuals, while if they are household weights, the prevalence refers to the population of households. For the calculation of the indicator 2.1.2, objective is to produce a prevalence of individuals. This implies that:

if a survey is at household level, and provides household sampling weights, they should be transformed to individual sampling weights by multiplying the weights by the household size. This individual weighting system can then be used to calculate the individual prevalence rates in formulas (1) and (2)

If the survey includes only adults, then the adult weights applied to the probabilities in formulas (1) and (2) provide the adult prevalence rates ($F{I}^{Adults}$). In this case, to calculate the prevalence in the total population, then the proportion of children who live in households where at least one adult is food insecure must also be calculated. This can be done by dividing the adult weights by the number of adults in the household and multiplying those approximate household weights by the number of children in the household. Once the approximate child weights are obtained, the prevalence of food insecurity of children who live in households where at least one adult is food insecure ($F{I}^{Children}$) can be calculated by applying these weights to the probabilities of food insecurity in formulas (1) and (2). The prevalence of food insecurity in the total population is finally calculated as:

$F{I}_{mod+sev}\mathrm{\ }=\mathrm{\ }\frac{F{I}_{mod+sev}^{Adults}\times N^{Adults}+F{I}_{mod+sev}^{Children}\times N^{Children}}{N^{Adults}+N^{Children}}$

and

$F{I}_{sev}\mathrm{\ }=\mathrm{\ }\frac{F{I}_{sev}^{Adults}\times N^{Adults}+F{I}_{sev}^{Children}\times N^{Children}}{N^{Adults}+N^{Children}}$

Where $N^{Adults}$ and $N^{Children}$ are the adult and children populations in the country.

When applied to the country total population, the prevalence of food insecurity in the total population provides the number of individuals who live in food insecure households (or in households where at least one adult is food insecure) in a country, at different levels of severity ($N_{mod+sev}$ and $N_{sev}$). In the database, the number of food insecure people are expressed in thousands.

2024-07-29

Food and Agriculture Organisation of the United Nations (FAO)

2.1.1, 2.2.1, 2.2.2, 2.2.3

Comments:

Links with Target 2.2, to the extent that hunger may lead to malnutrition, and Target 2.2 may not be achieved if Target 2.1 is not achieved.

2