Experience-Based Food Insecurity Scales, a Non-Aggregative Approach to Synthesis of Indicators

Food security is a complex and multidimensional phenomenon, access to food is one of its dimensions and experience-based food insecurity scales have been, in the last decades, the main tool for investigating it. When the phenomenon is considered at the household level, then these scales are built from a set of dichotomous items reflecting experiences related to access to food that the household possibly had to deal with in the previous months. From the point of view of the statistical treatment, these scales are mainly tackled using two different approaches: the counting approach and the Rasch model. The first one is mainly adopted at a national level to compute prevalences of food insecurity at different levels of severity. On the other hand, the Rasch model approach is adopted by the Food and Agriculture Organization of the United Nations (FAO) with the aim of monitoring access to food at a global level by producing comparable prevalences of food insecurity across countries. Although following different statistical steps of the analysis, both the counting and the Rasch model approach consider the vector of responses of a household to a number of dichotomous items and condense all information into one value only, the range of which will determine the category of food insecurity the household belongs to.In this way, two households with the same value for the final indicator could have potentially affirmatively answered different items, with a consequent lost of information that would instead help differantiate the two households and better suggest strategies for policy-makers. The objective of our work is to provide, starting from the indicators of each domain of the access to food, synthesis adopting a non-aggregative approach, namely the Partial Order Set Theory (Poset). The resulting composite indicator, in contrast with the case of both the counting approach and the Rasch model, is not a number anymore but a Directed Acyclic Graph (DAG) called the Hasse diagram. This graph represents the set of partial comparabilities that can be established among different profiles households belong to. At the core of this approach is the idea that not all profiles resulting from answers to a set of dichotomous items can be directly and unambigously compared. Therefore, it can be of practical relvance to rely on a methodology that more realistically reflect the ordinal quality of the data. Analysis of this work concerns data from the food security section of the “National Survey on Life Conditions” in Guatemala in 2014. The non-aggregative approach allowed us to highlight differences in the eight regions of Guatemala that would otherwise not show up if adopting an aggregative approach to the synthesis of indicators.