Temporal Aggregation and Treatment of Zero Dependent Variables in the Estimation of Food Demand using Cross-Sectional Data

This study analyzes U.S. consumers' demand for eight food commodity groups: Cereal and Bakery goods, Meat and Eggs, Dairy, Fruits and Vegetables, Nonalcoholic Beverages, Fats and Oils, Sugar and Sweets, and Miscellaneous goods. The data used in this study is Nielsen Homescan data for the period 2002-2006. Two different levels of temporal aggregation, monthly and the average month within a year, referred to as "annual" were considered. We conclude that the models using monthly data closely approximate the underlying annual expenditure elasticities, but do a poor job of estimating own- and -cross price elasticities and marginal effects. This finding is true for both the uncensored model of Blundell and Meghir (1987), and the two-step censored model of Shonkwiler and Yen (1999). We also find that the more complex two-step censored model does not improve precision of the estimates over the simpler model.