Academic journal article Journal of Agricultural and Resource Economics

A Comparison of Food Demand Estimation from Homescan and Consumer Expenditure Survey Data

Academic journal article Journal of Agricultural and Resource Economics

A Comparison of Food Demand Estimation from Homescan and Consumer Expenditure Survey Data

Article excerpt

(ProQuest: ... denotes formulae omitted.)


Demand models play an important role in the analysis and measurement of consumer preferences and the evaluation of agricultural and food policies. For example, the responsiveness of the quantity of a good demanded to a change in its price is measured by the own-price elasticity, a common output of demand model estimation. Demand models can also be used to analyze policy issues such as the welfare costs of a tax reform (e.g., Banks, Blundell, and Lewbel, 1997) or the effect of health information on demand (e.g., Tonsor, Mintert, and Schroeder, 2010).

Applied demand analysis requires two principal elements: a parametric or semiparametric demand system derived from the theory of consumer behavior (e.g., the Almost Ideal Demand System or the Rotterdam Model) and a dataset, which is used to estimate the model's parameters. An extensive literature focuses on the development of highly flexible demand forms (e.g., Piggott, 2003; Barnett and Yue, 1988). Several types of data have been used for the econometric estimation of the models, including time series, cross-sectional, and panel data; however, only a small number of demand studies have evaluated the quality, statistical properties, or the effects of the data on the final results of their analyses. Nielsen Homescan is one of the few data sources that has been evaluated (Harris and Blisard, 1995; Zhen et al., 2009; Einav, Leibtag, and Nevo, 2010). This proprietary dataset tracks consumers' grocery purchases and is collected by the Nielsen Company. Zhen et al. (2009) compared household expenditures based on Homescan data and the data from the Consumer Expenditure Survey (CEX), collected by the U.S Bureau of Labor Statistics (BLS). They concluded that the datasets report substantial differences in household food expenditures despite having comparable demographic compositions. Einav, Leibtag, and Nevo (2010) compared data recorded by a retailer through its loyalty program with Homescan data for the same group of consumers. They found that recording errors in prices were more prevalent than errors in trip information and product and quantity information, but concluded that the degree of measurement error in those price data was comparable to that found in other datasets commonly used in social sciences. Einav, Leibtag, and Nevo also illustrated the effects of measurement error in two applications: price regressions and demand estimation. In both cases, they found significant differences between the regression results obtained using Homescan data and those based on retailer data. These findings are problematic, as a large and growing number of researchers use scanner data to investigate research questions related to food demand, diet, and health (e.g., Hausman and Leibtag, 2007; Kuchler, Tegene, and Harris, 2005).1

An alternative source of consumer food expenditure data at the household level is the CEX, which has been used widely in the analysis of food consumed away from home for U.S. households (Byrne, Capps, and Saha, 1996; Jensen and Yen, 1996; Stewart and Yen, 2004; Zan and Fan, 2010). One drawback of the CEX data is that they contain information only on expenditures, not on prices and quantities, which precludes their use for estimating conventional demand models, including price and income effects. However, recent advances in the econometric literature have shown that, in some cases, it is possible to overcome this limitation by constructing household-specific price indices (Stone-Lewbel [SL] prices) derived from regional price indices (Hoderlein and Mihaleva, 2008).

The main objective of this study is to evaluate the potential of publicly available datasets and state-of-the-art econometric methods in lieu of the proprietary Homescan data. A secondary objective is to cross-validate the results of food demand analyses (i.e., elasticity values) using alternative data sources. Thus, this study was designed to answer the following questions: Are there any differences between demand model estimates obtained using Homescan data and BLS data (CEX and price indices)? …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.