NCCOR Connect & Explore Webinar on FoodAPS — You asked, USDA answered
April 13, 2015
On March 31, NCCOR’s Connect & Explore Webinar Series revealed insights from the U.S. Department of Agriculture (USDA) National Household Food Acquisition and Purchase Survey (FoodAPS) and discussed new research opportunities made possible by the first-of-its-kind survey.
During the webinar, there were a number of questions that were asked. Speakers Jessica Todd, Senior Economist, and Mark Denbaly, Deputy Director for Data – both from the Food Economics Division within the Economic Research Service at USDA – have responded to participants’ questions below. Read the full report.
Participant selection and sampling frame
FoodAPS relied on an area-based random sample. The U.S. Postal Service delivery addresses in 27 states were matched to addresses of Supplemental Nutrition Assistance Program (SNAP) participants, creating two distinct frames: one for SNAP and one for non-SNAP populations. Addresses were then selected randomly from the two frames. Field interviewers went to sampled addresses to conduct a screener to determine eligibility for the survey. If the residence was a seasonal unit, the occupants were not eligible (because they also could have been selected at their primary address). Group homes were excluded as well. The screeners collected information about household size, income, and SNAP participation and used this information to determine a household’s survey eligibility.
Some information on response rates is provided in the FoodAPS User Guide, and ERS plans to post more detailed information in the future. About 83 percent of households that completed the screener and were determined to be eligible initially agreed to participate, and about 76 percent of them completed the data collection.
Information brochures were sent to sampled addresses about one week prior to field attempts to contact the household in person to administer the screener, determine eligibility for the survey, and recruit eligible households. Analysis and comparison of the geographic and/or sociodemographic characteristics of households by survey completion status have not been completed.
Exploring the data
In the first interview, the primary respondents were specifically asked what store they go to do most of their food shopping. The interviewer was assisted with a pre-populated list of stores near the household and the household could select a specific location or fill in something that was not on the list. From that we were able to geocode all but 317 households’ primary stores, which were linked to the SNAP database and also our other retail databases.
After asking about where the household does most of its food shopping and follow-up questions about that store, the respondent was asked where else they shop in a typical month. The documentation refers to this store as an “alternate” store. Participants were then asked whether anyone in the household ever shops anywhere else, even if just for a few items. If yes, they are asked for the number of different places the household shops.
For each shopping event, FoodAPS collected information about which forms of tender were used to pay for the purchase. Multiple forms were often reported. If SNAP benefits were reported as tender, the food book asked the respondent to record the dollar amount of the SNAP purchase. Regardless of which tender types were reported, the respondent was asked to report the total paid, including tax and tip if any. If multiple forms of tender were reported, FoodAPS did not collect information on which food items were paid for by which form of tender.
Four target groups were defined for the sampling plan, and the two poverty break-points for the three non-SNAP groups were 100 percent and 185 percent. FoodAPS used 100 percent rather than 130 percent for the lowest income group for two reasons. First, the range between 130 and 185 percent seemed rather small for a sampling group. Second, and more importantly, we expected that finding and recruiting low-income non-SNAP households for the survey would become more difficult among very-low-income households. By using the lower threshold for the low-income group, we expected that the sample of non-SNAP households would more closely resemble the sample of SNAP households in terms of their overall income distribution.
Data access and availability
No, access is available through remote devices called “thin clients.” Researchers with an approved project agreement who have completed the Confidential Information Protection and Statistical Efficiency Act (CIPSEA) training and confidentiality agreement register for an account at NORC and lease one or more devices. There are fees for the account and devices. Thereafter, the thin-client laptops may be used in a secure environment to access the FoodAPS data. More details are available on the FoodAPS page.
The data are nationally representative and include the sub-populations mentioned in the webinar. However, the capability to link households to very local information is available, but there is limited capability to do local area estimates. Regions might be distinguishable, but it really is not suitable for an analysis of a specific market location.
Implications and connections
Nielsen – which is now similar to IRI data – collects only food-at-home purchases, or food from retail food stores. It does not collect any food-away-from-home purchases. FoodAPS is unique in that we have both food-at-home and food-away-from-home acquisitions by the same households among all household members over an entire week. Not only that, we have acquisitions that were obtained at no cost from food pantries, personal gardens, and hunting and gathering, or other types of acquisitions that wouldn’t be captured from any other measure of food-away-from-home purchases or food-at-home purchases. FoodAPS uniquely includes other information about the food environment in which the respondents reside. This enables researchers to study how the varying local food environment (e.g., proximity and access to healthy and affordable foods) influences our food choices. Nielsen and IRI do not focus on providing information about the food environments of their panelists. In addition, IRI/Nielsen data do not include information on SNAP participation.
FoodAPS utilized SNAP administrative records only. However, we did ask if respondents participated in the WIC program or if reported food items purchased were related to the WIC program. If a researcher has access to administrative WIC records, they may be able to link them to FoodAPS data.
While car ownership may or may not be a marker of income depending on location, FoodAPS asked the same questions about car access and modes of transportation to everyone in the survey. There are a number of questions about vehicle use, ownership, and access to a vehicle when needed.
ERS has started the planning process for a second FoodAPS-like survey. A number of factors, however, will affect that decision. Conducting a complex survey like this requires substantial resources, so a second survey will be possible only if sufficient resources are available. Equally important is how useful the existing FoodAPS data become. If these data are used to conduct a wide range of research projects that provide results that can inform public food and nutrition policy, then the chances of conducting a second survey in the future will be increased.