Simultaneous Prediction of Multiple Dimensions of Food Security

Lucie Maruejols

No 397911, 100th Annual Conference, March 23-25, 2026, Wadham College, University of Oxford, Oxford, UK from Agricultural Economics Society (AES)

Abstract: Program officers are faced with difficult modelling choices when monitoring the state of diverse yet related dimensions of household food security. We compare results of single- and multi-output frameworks, applying both econometric and machine learning models, to jointly predict three interrelated outcomes of food security: prevalence of undernourishment (access), dietary diversity (utilization), and food market dependence (stability). Using the 2013–2022 Kyrgyz Integrated Household Survey, the results show that accounting for interdependencies among outcomes significantly improves accuracy, with the Generalized Structural Equation Model (GSEM) outperforming both single-output regressions and single- and multi-output neural networks. The asymmetry between caloric adequacy, dietary diversity, and market exposure explains that multi-output modeling enhances predictive power over single outputs. The findings can guide program officers navigating the trade-offs between prediction performance, interpretability and feasibility, showing that linear models can deliver robust predictions even when outcome dependencies are jointly addressed. These findings advance understanding on the interrelations between multiple dimensions of food security and have important implications for designing monitoring strategies that recognize the multidimensional nature.

Keywords: Food; Security; and; Poverty (search for similar items in EconPapers)
Pages: 41
Date: 2026-03
References: Add references at CitEc
Citations:

Downloads: (external link)
https://ageconsearch.umn.edu/record/397911/files/L ... al_paper_Jan2026.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:ags:aes026:397911

DOI: 10.22004/ag.econ.397911

Access Statistics for this paper

More papers in 100th Annual Conference, March 23-25, 2026, Wadham College, University of Oxford, Oxford, UK from Agricultural Economics Society (AES) Contact information at EDIRC.
Bibliographic data for series maintained by AgEcon Search ().