A comparative analysis of hedonic models of nutrition information and health claims on food products: An application to soup products

ABSTRACT

Over time, the quality of data on food purchases and label information has improved such that hedonic analyses to determine the implicit prices of product attributes can be conducted using more detailed data than in the past. With the availability of more extensive data, it is important to understand the characteristics of the data and implications of using different data sources on results of analyses. The purpose of this study was twofold: (1) compare results between two sources of label information and (2) develop a better understanding of the effects of product claims and nutrition information on the value of products to consumers. Trans fat claims, organic claims, private label, package size, and several nutrients were found to influence implicit prices for soup products, and the results between the two data sources are comparable.

KEYWORDS:

• Health claims
• food labeling
• scanner data
• hedonic model
• implicit prices

Acknowledgments

This study was conducted under Agreement 58-5000-3-0069 with the U.S. Department of Agriculture (USDA), Economic Research Service (ERS) and under a third-party agreement with IRI to allow access to the proprietary data. We appreciate comments from Abigail Okrent and participants of the Conference on Applying IRI Store and Household Scanner Data in Food Policy Studies. Any opinions, findings, conclusions, or recommendations expressed in this article are those of the authors and are not attributable to USDA, ERS, or IRI.

Notes

¹ In addition to IRI and Gladson, other commercial suppliers of label data in the U.S. include Label Insight and Mintel (for new product introductions). Nielsen also sells UPC-level purchase data but label data must be linked from one of the other suppliers.

² To be considered a match, an UPC needs to have a non-null value for at least one variable in the nutrition data. However, more than 98% of the nutrition data records have values for 12 or more fields, and 78% of the records have values for 24 or more fields.

³ The full details of the nutrition fields for IRI are in the ERS technical bulletin: Muth, M. K., Sweitzer, M., Brown, D., Capogrossi, K., Karns, S., Levin, D., … Zhen, C. (2016). Understanding IRI household-based and store-based scanner data. Technical Bulletin TB-1942. Washington, DC: U.S. Department of Agriculture, Economic Research Service.

⁴ Prices might not represent the exact value a household paid for an item; instead, IRI assigns prices based on the average price of the item at the retail outlet chain where the item was purchased in the market area the household resides. If the household shops at a store that is not represented in the IRI point-of-sale data, IRI uses the price that households input during the reporting (Muth et al., 2016).

⁵ We calculated a price per ounce rather than a price per serving because the weight of a serving size varies across individual products. For condensed and dried soups, we converted the weight to an as-served weight before calculating the average price. Furthermore, the shelf label for the product indicates price per ounce, which allows consumers to compare the price of products.

⁶ In the IRI data, 1,119 soup UPCs were dropped because they did not have associated nutrient or nutrition claims data. In the Gladson data, 823 soup UPCs were dropped that were not in the IRI data with nutrition information.

⁷ We calculated the factor of 5.7 for dry soups by taking the average of the amount of water added to several dry-soup products based on the package instructions. We calculated the factor of 5.3 for ramen from the package instructions of ramen. We calculated the factor of 3.8 for instant noodle products, which had a mean weight of 2.11 ounces, by assuming these products require the addition of 8 ounces of water.

⁸ We used per-serving nutrition information because the consumer sees this information on the Nutrition Facts Label.

Additional information

Funding

This work was supported by the Economic Research Service [Agreement 58-5000-3-0069].