Elsevier

Measurement

Volume 116, February 2018, Pages 146-152
Measurement

Food security measurement in a global context: The food insecurity experience scale

https://doi.org/10.1016/j.measurement.2017.10.065Get rights and content

Abstract

The ability of households and individuals to access food (one of the key aspects of 'food security') is an important welfare dimension that poses important challenges for objective measurement. This paper describes the Rasch model-based procedures developed to define the eight-item Food Insecurity Experience Scale (FIES) as a contribution towards the establishment of an indicator for global monitoring of food insecurity. Experiential food insecurity survey data, collected by FAO from nationally representative samples of the adult population, once every year in 2014, 2015 and 2016 from 153 countries or territories, are used to develop methods to estimate cross-country comparable prevalence rates of moderate and severe food insecurity. A Rasch model-based scale was estimated separately for each country and data were assessed for consistency with model assumptions. To ensure cross-country comparability, a procedure based on the median normalized severities of each of the eight FIES items was used to define a global reference scale, against which measures obtained in each country can be separately calibrated. Calibration is obtained by equating the mean and standard deviation of the severity parameters of the items that appear to be common between the national and the reference scale, and thus used as anchoring points for the metric. Data showed sufficient consistency with the Rasch model assumptions to produce reliable measures of the prevalence of food insecurity in each country. Calibration was possible using 4 or more items as anchoring points in 151 of 153 (98.7%) of the cases, and 6 or more items in the vast majority of them (121 cases). Concurrent validation of the estimates of prevalence of food insecurity at national level was obtained by comparing the FIES-based indicator with other established indicators of social (under) development. National prevalence rates of moderate-or-severe food insecurity obtained by FAO correlate well with the prevalence of undernourishment and with several widely used indicators of national income, health, and well-being. The proposed calibration method can be applied to other existing experience-based food security scales that use similar items, thus affording the possibility to use data collected with those instruments to produce internationally comparable measures of the prevalence of food insecurity. Pending broader adoption of the FIES or compatible experience-based food security scales worldwide, countries could choose to use the 2014–2016 results obtained using the data collected by FAO as the baseline to monitor progress towards Target 2.1 of the recently established 2030 Agenda for Sustainable Development.

Introduction

Food security is said to exist when all people, at all times, have physical, social and economic access to sufficient safe and nutritious food that meets their dietary needs and food preferences for an active and healthy life [1]. Although food security is inherently multi-dimensional, one critical dimension is continued access to adequate food. The United Nations Food and Agriculture Organization (FAO) has undertaken a project called Voices of the Hungry (VoH) to develop and support a survey-based experiential measure of access to food, called the Food Insecurity Experience Scale (FIES). The approach to measuring households' ability to access food is similar to that of other experience-based food security scales such as the US Household Food Security Survey Module (HFSSM), the Escala Brasileira de Insegurança Alimentar (EBIA), the Escala Latinoamericana y Caribena de Seguridad Alimentaria (ELCSA), the Escala Mexicana de Seguridad Alimentaria (EMSA) and the Household Food Insecurity Access Scale (HFIAS) used in the United States, Brazil, Canada, Mexico and several other countries to monitor food security in the context of large national programs [2]. The innovation brought about by the VoH project is the possibility to calibrate the measures with the FIES or with any of these other scales and the thresholds used for classification, against a standard reference scale, thus ensuring proper comparability of the estimated prevalence rates and the possibility to compute consistent estimates at regional and global level, an essential feature for an indicator to be used in the context of global monitoring frameworks. Following a very broad consultation with many stakeholders, the FIES was chosen as the basis to compile indicator 2.1.2, one of the two indicators included in the global SDG indicator framework put forth by the Interagency and Expert Group on SDG indicators (IAEG-SDG) of the United Nations Statistical Commission to monitor Target 2.1 of the recently adopted 2030 Agenda for Sustainable Development [3].

The FIES measures the severity of food insecurity modelled as a latent trait, broadly conceptualized as the condition of not being able to freely access the food one needs to conduct a healthy, active and dignified life. The measure is based on conditions and behaviors reported by responding to an 8-item questionnaire, the Food Insecurity Experience Scale Survey Module (FIES-SM; see Table 1), resulting from the inability to access food due to lack of money or other resources. These conditions have been selected, among the many possible ones that could be meant to be a direct consequence of the latent condition, as those holding the greater promise to be empirically valid in many different contexts.

The dichotomous (“yes”/“no”) responses to the FIES-SM questions, provide information sufficient to construct a one-dimensional measure, using the Rasch model. Based on the measured severity of food insecurity, each respondent in a representative sample is assigned a probability of being beyond a specified threshold of severity to compile an estimate of the prevalence rate of food insecurity in the reference population. Thresholds used for classification and, thus, prevalence rates of food insecurity, are made comparable across countries by calibrating the measures obtained from estimating the Rasch model parameter separately on each dataset, against a common, global reference scale.

The next sections describe the data used, the statistical modeling and the procedures developed to form the global reference scale and to calibrate the measures, and address validation of the food insecurity prevalence rates estimated in 153 countries for 2014–16.

Section snippets

Data

In proposing the FIES as the basis to compile an SDG indicator, FAO expects that national prevalence rates of food insecurity for monitoring progress toward SDG Target 2.1 will eventually be based on data from national surveys conducted by national statistical agencies in each country, in accordance with the principles that govern the definition of the global SDG indicator framework by the UN Statistical Commission. To develop methods for making prevalence rates across countries comparable,

Statistical modeling of FIES data to produce estimates of the prevalence of food insecurity at comparable levels of severity

The statistical model used for FIES data assessment and scale construction is the single-parameter logistic measurement model, commonly known as the Rasch Model [5]. The Rasch model assumes that the position of a respondent and that of the items can be located on the same one-dimensional scale and postulates that the log-odds of respondent r saying “yes” to item i is a linear function of the difference between the severity of the food insecurity condition experienced by r and the severity of

Development of a global reference and scale calibration

Use of a measure of food insecurity to inform indicators used in a global monitoring framework must ensure that estimated prevalence rates are comparable over time and across countries. To do so, severity thresholds for classification should be defined on a common reference scale and kept constant during the monitoring period, while prevalence rates computed by ensuring that severity measures and thresholds are expressed in the same metric. This can be done either by mapping the national

Scale stability and estimates of food insecurity prevalence rates with small samples

We were concerned that GWP effective sample sizes of non-extreme cases (i.e., after omitting cases that denied or affirmed all items, which provide no information for parameters estimation when using CML) might be too small, in many countries, to provide sufficiently precise parameter estimates. If that is the case, estimation of the scale using data from only one year could be rather unstable for those countries. With data sets from the GWP rounds of 2014, 2015 and 2016, stability over time

Setting thresholds and estimating prevalence rates for global SDG monitoring

For the specific purpose of monitoring progress against Target 2.1 of the SDGs, two thresholds have been set: one that identifies the level of severity beyond which a respondent would be classified as having moderate or severe food insecurity, and one that identifies severe levels only. The definition of a threshold of severity for the latent trait is, to a certain extent, arbitrary, as the only requirement for consistent classification is that whatever threshold is chosen, it is kept constant

Results and assessment of the consistency between FIES-based measures of the prevalence of food insecurity in the world and other development indicators

Based on the above procedure, FImod+sev and FIsev were computed for all countries for which FIES compatible data were available. National prevalence of moderate-or-severe food insecurity in 2014–16 in the adult population ranged from 2.3% to 94%. Severe food insecurity rates ranged from below 0.5% to 83%. Regional prevalence rates were calculated as the population-weighted averages of the prevalence rates of the countries included in each region. Across the continents, food insecurity is found

Conclusions

The analysis of FIES data collected over three years in more than 150 countries worldwide confirms that self-reported evidence on the occurrence of conditions typically associated with the inability to access food due to lack of money or other resources, gathered through simple interviews, can indeed inform the construction of a valid measurement scale of the severity of the food insecurity condition. Rasch model-based analytic procedures, consistent with the item anchoring and scale equating

Acknowledgements

Financial support for the Voices of the Hungry project has been provided by the United Kingdom Department for International Development (DfID) and by the Kingdom of Belgium through the FAO Multipartner Programme Support Mechanism (FMM). The comments, opinions and judgments expressed in this paper are those of the authors and do not imply any official position by FAO, the FAO Statistics Division, or the funding partners.

References (17)

There are more references available in the full text version of this article.

Cited by (127)

Arrow Up and RightView all citing articles on Scopus
View full text