Moving Beyond Indices: A Systematic Approach Integrating Food System Performance and Characteristics for Comprehensive Food Security Assessment

Food security indices are widely used to support decision making and provide a structured assessment of countries' capacities to withstand global environmental and economic crises. However, these indices have inherent limitations, including potential biases in ranking and a lack of structural insights into food system dynamics. This study presents a systematic approach that combines elastic-net regression-based feature selection and two-step clustering to address some of these limitations and equip decision makers with structured procedures for making informed decisions and supporting food system management. The mathematical and operational procedures of the proposed approach were demonstrated through an illustrative example using the EIU dataset of 94 countries. The study investigated the sensitivity of composite indicators to extreme data points, relative weights, and dimensionality reduction. After applying elastic-net regression, 15 indicators were selected for Model 1 (M1) and 9 for Model 2 (M2) from an initial set of 25 indicators. Subsequently, two-step clustering grouped the countries into four distinct clusters, reflecting combinations of food system characteristics and income levels. The results demonstrate that countries with industrialized, consolidated food systems and high per capita income tend to exhibit greater food security. Conversely, countries with rural or traditional food systems and low-income levels are more vulnerable to food insecurity. By incorporating statistical rigor and empirical structure discovery, this methodology addresses key limitations of existing indices. It provides an adaptive, transparent framework that informs targeted policy by linking the structural characteristics of food systems to tangible food security outcomes.