Forecasting food trends using demographic pyramid, generational differentiation and SuperLearner

This paper aims to predict future food consumption patterns for different social groups, considering generational changes. Using over 20 million Swiss household observations (1990-2017), four forecasting techniques are developed and applied, progressing from linear extrapolations to population-driven models. Each method defines social groups based on household characteristics, projects the group proportions until 2050, forecasts consumption of 75 food items for each group, and weighs these to obtain aggregate consumption. Despite variations across food items and methods, the results converge on a narrow range of future consumption until 2050. The study offers insights into forecasting techniques using big data and future food consumption.

Daria Loginova, Stefan Mann