A taste of tomorrow: Predicting food demand elasticities under different Shared Socioeconomic Pathways

• We design a novel data-driven approach to project food demand parameters over time. • We train a machine learning model using meta -analysis and socioeconomic data. • The Shared Socioeconomic Pathways drivers are used to project the parameters. • We provide demand parameters for 12 food types in 138 countries under 5 SSP. • An illustration with 20 CGE simulations shows the added value of this new approach. Food policy assessments and food demand projections rely on demand elasticities. The elasticities used, however, often lack granularity and depend on ad hoc adjustments to make them evolve over time. In this study we explore an alternative approach using a meta -analysis database and the XGBoost machine learning algorithm to predict food demand elasticities. Next, we use the Shared Socioeconomic Pathways (SSPs) database to project the elasticities to 2030, 2040, and 2050. The elasticities are then calibrated to comply with theoretical conditions and used to parameterize the demand system in a Computable General Equilibrium (CGE) model. Finally, using the CGE model, we illustrate the implications of the new parameters by simulating four sets of simple scenarios. As output files we provide (1) income, own-price, and cross-price (both compensated and uncompensated) elasticities for 12 food groups, 138 countries, and 5 SSPs, (2) their calibrated counterparts, and (3) the equivalent expansion and substitution parameters for a CDE demand system. These parameters can be applied in a wide range of scenario building and policy assessments.