Reassessing Food Security: How a Data-Efficient 4As Framework and Machine Learning Uncover Hidden Patterns Across G20 Nations

Abstract

Food security is a global challenge that demands a systematic approach to inform effective policymaking. However, empirical country-level food security studies remain scarce because of data limitations. To bridge this gap, we first develop a data-efficient National Food Security Index (NFSI) by innovatively adapting the 4As framework (availability, affordability, accessibility, and acceptability) of energy security. The weights of indicators in the framework are determined by an expert survey. The index is then applied to G20 members, and a clustering algorithm on the basis of machine learning uncovers several hidden patterns. The main findings of this study are as follows: (1) agricultural productivity, food affordability, and natural resource endowment are perceived as most crucial in determining food security; (2) Australia, the USA, France, the UK, and Germany consistently exhibit strong food security, whereas India, Mexico, Russia, and Indonesia trail behind. EU members demonstrate substantial improvements in sustainability, contrasting with mixed progress patterns observed in other major economies; and (3) five clusters are identified: leading performer (USA), resilient performers (like Canada and Germany), innovative performers (China, Japan, and South Korea), moderate performers (like Saudi Arabia and South Africa), and vulnerable performers (India and Indonesia). Tailored policy recommendations are provided for each cluster.