The Influences of Portion Size, Context and Package Characteristics on Snack Food Consumption: Evidence from a U.S. Rolling Cross-Sectional Survey

ABSTRACT

Using a rolling cross-sectional sample of over 40,000 U.S. snack food consumers during 2004–2010, we build a three-dimensional model to study how portion sizes, consumption context, and package characteristics jointly affect the amount of food consumed. As with prior research, we find that consumption does increase with larger portions. However, this effect diminishes significantly when accounting for the consumption context and packaging characteristics as consumption drivers. Most important is how the snack is presented to consumers at the point of consumption (type, size, and quantity of packages). We also find that snacking on impulse, alone, with a beverage, or while watching television increases the amount consumed, whereas socializing or snacking at home decrease consumption. We test our model over seven snack types (potato chips, crackers, cheese puffs, nuts and seeds, tortilla chips, pretzels, popcorn), and find that our results hold even for products with different cross-category granularity, palatability, and satiety.

KEYWORDS:

• Portion size
• food marketing
• consumption norms
• packaging
• obesity

Notes

¹ http://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight.

² http://www.healthdata.org/news-release/nearly-one-third-world’s-population-obese-or-overweight-new-data-show.

³ http://www.cdc.gov/obesity/data/adult.html; Ogden, Cynthia L., et al. “Prevalence of childhood and adult obesity in the United States, 2011–2012.” JAMA 311.8 (2014): 806–814.

⁴ Every participant was trained to record his/her daily snacking via a survey pre-programmed in an iPhone-like mobile device. At the end of the 2-week participation, the participant mailed the device back, and the data was downloaded and validated. The survey design and administration, as well as the management of data collection, was performed jointly by the snack food company and a major market research provider.

⁵ To account for possible trend or seasonality variations in snacking, the 9,000 participants were selected daily at an even rate across each year. Both the presence of an individual and the time at which the individual is included in the sample were determined via random selection. Each participant recorded his/her real snacking over a 2-week period. The data we secured for this research only lists the first snacking occasion recorded by each participant.

⁶ Participants were trained on how to determine portion size and amount consumed by reading the nutritional label information (e.g., there are nine chips in a serving and three servings in a small package), counting the number of packages in front of them, and estimating the share of the package they consumed in increments of 1/8 of package. Parents did the measuring and reporting for children up to 12 years of age. Data was validated through follow-ups with random participants. No consistent or significant mis-reporting or mis-estimation was found.

⁷ Participants self-reported their height and weight and from this we calculated their BMI. See: http://www.cdc.gov/obesity/adult/defining.html .

⁸ Other categorization schemes modeled show consistent and similar results.

⁹ To quantify the effect of each predictor on amount consumed, we follow the standard calculation in linear regression models with log-transformations as shown by Greene (2003), where the percentage effect is given by exponentiating the linear regression coefficients.

¹⁰ Available at: https://health.gov/dietaryguidelines/2015/guidelines/ .