Household Fruit and Vegetable Demand Estimation and Forecasting: A Revealed Preference Approach

Abstract

Applying 2016-2017 household scanner data from market research firm IRI, we combine parametric and nonparametric techniques in estimating demands and forecasting consumption for six aggregated fruit and vegetable categories. The 2016 data is segmented by revealed preference (RP) such that the behavior in each subset of households is consistent with traditional utility theory, and a nonlinear Almost Ideal Demand System (NL-AIDS) model is estimated for all subsets. For comparison, demands are also estimated when the data is segmented separately on geography and household demographics. Own-price and expenditure elasticities generated across RP-consistent subsets indicate a wide range of demand responsiveness, whereas geographic and demographic subsets show similar behavior. Demand is generally more elastic for perishable goods than non-perishable. All methods of segmentation perform similarly when forecasting consumption into 2017.