Data needs for consumer and retail firm studies.

Supplementing the NHANES dataset with administrative records would allow researchers to study the connection between food choices and neighborhood characteristics, particularly for low-income households in urban and rural areas. To the extent that NHANES includes such households, researchers could correlate health and nutrition outcomes with household and location characteristics. A link between NHANES data and information on the location of food stores and eating establishments would also enhance efforts to understand the effects of access on food choices and health outcomes. Information on locations and characteristics of food stores and foodservice establishments can be collected using proprietary sources, such as Spectra[R] and NPD. Linking NHANES to household and local community descriptors in the Census's American Community Survey will help researchers understand how neighborhood characteristics influence food choices and health outcomes.

Improving Datasets

We have a simple and obvious message. With more data, economists could analyze additional, important issues of economic theory and government policies.

Because data lack rivalry (everyone can consume the data), society under-provides data. Relying on commercial vendors is unattractive because these firms charge very high prices, do not fully disclose the nature of their data, provide data for only very short periods, and report only variables that are important for commercial customers and not all variables that are important for researchers.

One approach to ameliorating data shortages for research would be to have government agencies or nonprofit organizations collect the ideal datasets or provide incentives to commercial providers. Fundamentally, researchers need access to unrestricted data based on proper random samples and that include all the relevant variables.

First, to enable unfettered assess, to improve content, and to obtain better prices, it may make sense for university and government researchers and organizations (the AAEA, government agencies, business school organizations, the American Economic Association, and others) to try to negotiate with private purveyors collectively. They might also negotiate to house, at little or no cost, historical IRI and Nielsen data that are now discarded so that longer time series and additional variables can be created. However, such collective action might raise antitrust issues.

Second, these research groups could try to make arrangements with individual firms to supply data. We know of at least two supermarket chains that have been willing to make such agreements in the past. The AAEA could lead efforts to select representative samples of suppliers to collect details of proprietary transaction data and provide them to researchers so that privacy and confidentiality of the data are maintained.

Third, these research organizations could collaborate to collect data on their own. Even discussing this possibility may facilitate negotiations with commercial data purveyors.

On a less grand scale, we have a laundry list of new datasets that would be particularly useful. First, industrial organization and food safety studies require information at both the retail and upstream levels, including information about wholesale prices, food sources, various slotting and tying relations, and government programs.

Second, nutritional studies need datasets that combine information on food-at-home and away-from-home, nutritional content of these various foods, and prices. Because consumer studies find substantial variation in nutritional consumption across demographic groups and neighborhoods, datasets are needed that cover a broad cross-section.

Third, health and nutrition studies would benefit substantially if we could link the intake and health data with administrative food assistance records to add levels and duration of program assistance. Such a link would have to address two challenging issues: (1) privacy and confidentiality conditions under which states collect the administrative data must be met to access the data for linking purposes and (2) variation of data formats across states makes linking these sets to survey data difficult. In addition, given the relatively small effects of price and income on food choices, addressing the obesity epidemic may require collection of new data on consumers' health and nutritional knowledge, attitudes, and available time to shop and prepare meals to undertake economic studies to understand consumer dietary behavior.

References

Bollinger, C., and M. David. 2005. "I Didn't Tell, and I Won't Tell: Dynamic Response Error in the SIPP." Journal of Applied Econometrics 20:563-69.

Kaufman, E 2007. "Food Market Structures: Food Retailing." U.S. Department of Agriculture, Economic Research Service. www. ers.usda.gov/Briefing/FoodMarketStructures/ foodretailing.htm (2007).

Muth, M.K., EH. Siegel, and C. Zhen. 2007. "ERS Data Quality Study Design." Final Report, Research Triangle Institute, Project Number 210153.001.

National Research Council of National Academies. 2005. "Improving Data to Analyze Food and Nutrition Policies." Committee on National Statistics, Panel on Enhancing the Data Infrastructure in Support of Food and Nutrition Programs, Research, and Decision Making, National Academies Press, Washington, DC.

Villas-Boas, S.B. 2007. "Vertical Relationships Between Manufacturers and Retailers: Inference With Limited Data." Review of Economic Studies 74:625-52.

Waldfogel, J. 2003. "Preference Externalities: An Empirical Study of Who Benefits Whom in Differentiated Product Markets." Rand Journal of Economics 34:557-68.

Jeffrey Perloff is Professor, Department of Agricultural and Resource Economics, and member of the Giannini Foundation, University of California at Berkeley. Mark Denbaly is Deputy Director for Data and Web Communication, Economic Research Service, U.S. Department of Agriculture.

The opinions expressed in this paper are those of the authors and not necessarily the United States Department of Agriculture or any of its members. We thank Rui Huang for statistical help.

This article was presented in a principal paper session at the AAEA annual meeting (Portland, OR, July 2007). The articles in these sessions are not subjected to the journal's standard refereeing process. Table 1. Comparison of U.S. Census and IRI Demographic Data Households IRI Census With individuals <18 years old 35.0% 33.9% With income <$10,000 5.6% 7.4% With income >$100,000 1.9% 14.3% White 86.4% 71.9% Black 5.3% 10.8% Asian 1.3% 5.9% Hispanic 5.7% 17.0% Size 2.8 2.6 Notes: Average across all the zip code regions in the IRI data set. IRI data are for 1999 and Census data are from 2000.