Temporal Aggregation and Treatment of Zero Dependent Variables in the Estimation of Food Demand using Cross-Sectional Data
Date of Award
Master of Science (MS)
Applied Economics and Statistics
Carpio, Carlos E
Sharp , Julia L
Ward , William A
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 estimation of the demand system of equations is carried out using the EASI demand model of Lewbel and Pendakur (2009) and five years of data (2002-2006) from the Nielsen Homescan program. Two different levels of temporal aggregation, monthly and the average month within a year, referred to as 'annual' were considered. Using the monthly data, I evaluated the performance of two econometric methods to account for zero expenditures in food demand analysis (Shonkwiler and Yen, 1999 and Blundell and Meghir 1987).
I 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). I also find that the more complex two-step censored model does not improve precision of the estimates over the simpler model. This conclusion has implications for consumer demand researchers attempting to account for censored dependent variables.
The new own- and cross price elasticities presented differ from previous estimates in the literature, but not to the extent that the broad conclusions drawn from previous studies are nullified. Marginal effects of demographic characteristics appear to confirm societal beliefs about consumption patterns of different population sub-segments.
Leffler, Kristyn, "Temporal Aggregation and Treatment of Zero Dependent Variables in the Estimation of Food Demand using Cross-Sectional Data" (2012). All Theses. 1321.