Appendix B

Summary, Second Meeting, June 14, 2018

The panel’s second meeting included presentations covering a range of topics integral to addressing the study charge, including the current and potential use of proprietary commercial and other nongovernmental, nonsurvey data sources; users’ perspectives on directions for Economic Research Service’s (ERS’s) National Household Food Acquisition and Purchase Survey (FoodAPS) survey; and the linking of data sources. The topics covered in the four sessions, which align with this summary, were: (i) proprietary data used by (or of interest to) ERS; (ii) combining data sources to advance food and nutrition policy and research; (iii) use of specialized modules added to federal surveys; and (iv) a FoodAPS-2 status update along with the perspectives of data users and stakeholders.

B.1. PROPRIETARY DATA USED BY (OR OF INTEREST TO) ERS

After introductory comments by the panel chair, Marianne Bitler, and by Jay Variyam and Mark Denbaly of ERS, the panel heard from presenters about proprietary data. This session built on the April 16, 2018, presentation by Megan Sweitzer and David Levin (both of ERS) on the same topic. Proprietary data from commercial data sources may supplement (and, in some cases, replace) survey data. Sources of proprietary (purchased) data used by ERS include these:

• TDLinx, Nielsen (2004–2017)—Names and geospatial locations of food stores in the United States with sales greater than $1 million; used in ERS geospatial database;

• ReCount, NPD Group (1998–2017)—Locations and characteristics of restaurants; used in ERS geospatial database;
• IRI Consumer Network (2008–2017)—Household panel data, including scanner data (also includes MedProfiler and RXPulse, two household health surveys), and Nielsen Homescan (1998–2010)—Household panel data, including scanner data used in the ERS Quarterly Food at Home Price Database;
• IRI InfoScan (2007–2017)—Retail scanner data; used in the ERS Quarterly Food at Home Price Database; and
• Nielsen Homescan (1998–2010)—Household panel data, including scanner data.

Kicking off the meeting, Abigail Okrent (ERS) described ERS’s work to use proprietary commercial data and to understand its strengths and weaknesses. Okrent reported that ERS purchases household panel data (including scanner data) because they offer several advantages. The Consumer Network Panel is used by both Nielsen and IRI in their commercial household panel products. The panel has a large sample size, more than 120,000 households, with around half of these households providing purchase data. Additionally, the same households can participate in the panel every year. Researchers are able to append geographic food environment and economic information from other datasets to household-level records so the impacts of the food environment or macroeconomic conditions on household purchasing patterns can be examined. Scanner information is collected at the UPC level, which conveys brand, type, and manufacturer information; and household geographic location can be identified down to the census tract level.

To evaluate the strengths and limitations of proprietary household panel data, ERS has collaborated with colleagues from RTI¹ and academic institutions on a number of studies. In reviewing these studies, Okrent pointed out two common concerns. The first is that households are not randomly selected into the panel and, hence, the sample might not be representative of the population. Second, households that agree to participate in the sample might not record all of their purchases or might not record them correctly.

Okrent summarized some validation studies. Einav, Leibtag, and Nevo (2008) matched Nielsen Homescan households’ purchase records with data obtained from a large grocery retailer. The authors found that 80 percent of food-shopping trips in Nielsen Homescan showed up in the store’s data; the unmatched trips likely resulted from households not reporting all of their trips. For matched trips they found that about 93 percent of the time the two data sources reported the same quantity. The reported expenditure was the same about 49 percent of the time.

___________________

¹ See https://www.rti.org.

A study by Sweitzer and colleagues (2017) compared weighted expenditures in the IRI Consumer Panel with the Consumer Expenditure Survey (CES) and FoodAPS by food subcategories and demographic groups. The results from this study show that expenditures in the IRI data were lower than expenditures in the CES, but the magnitude and variation of these differences varied across food categories, years, and household demographic characteristics. Many of the food categories with the most underreporting are those that contain more random-weight products (e.g., fruits and vegetables that are measured by the pound).

These comparison studies suggest that researchers should be cautious when using the IRI household data for certain types of studies, such as research focusing on fresh fruits and vegetables or high-income or large households and studies that draw conclusions about the overall composition of consumers’ purchases or diets.

The strength of commercial scanner data, both for households and retail, is the detail they can provide on nutrition facts labeling information (e.g., calories, sodium, and calcium per package); health and nutrition-related claims (e.g., whether gluten-free, type of sugar/artificial sweetener, whole grain claims); and other claimed characteristics (e.g., organic, no preservative, hormone-free, natural).

The strengths of store data, such as InfoScan, are realized from highly detailed information on weekly food purchases for large numbers of stores; expenditures and quantities of UPC and random-weight food products; and location of establishments. Similarly, store characteristics data sources, such as TDLinx, National Economic Time Series (NETS),² and ReCount, offer detail on retail and food service establishments, including location and sales and employment information for each store.

ERS research (Levin et al., 2018) has compared store counts across TDLinx, NETS, and InfoScan.³ For the period of 2008–2012, the authors found that the number of stores and food sales in InfoScan were considerably lower than those in TDLinx and NETS. Comparing these totals to the 2012 Economic Census indicates that the version of InfoScan purchased by ERS covered about half of all sales.

Okrent reported that ERS is currently working on solutions to alleviate shortcomings in their application of commercial data. They are using available data to develop weights or projection factors to use to help make the store-level data more representative and for imputing missing random-weight prices.

___________________

² The National Economic Time Series database is available from Willis and Associates. It is based on establishment information from Dunn and Bradstreet.

³ Okrent pointed out that InfoScan only releases data to ERS for stores that agree to this arrangement, and this is limited to stores that make more than $2 million in sales. Also, some stores only release their sales data for their retail marketing area, which is retailer-defined.

Brian Burke of IRI described the company’s point-of-sale data, which InfoScan collects on a weekly basis from more than 200,000 stores globally. The company’s IRI Consumer Network Panel includes more than 110,000 consumers. The company has some health and wellness data in these databases now and will be expanding that feature in the future. IRI also has a shopper loyalty database with more than 125 million loyalty card holders. Finally, the company offers analytics that leverage its data assets.

Ann Hanson and Louis Lesce of the NPD Group summarized their data products, noting that they have point-of-sale data from retailers, distributors, and food service operators and that they also conduct consumer surveys. NPD Group has a number of food industry databases, and Hanson and Lesce summarized four: National Eating Trends, NPD’s consumer database with 19,000 respondents and food consumption data for at-home and away-from-home eating; Eating Patterns in America, NPD’s annual analysis of the state of food and drink consumption in the United States with long-term and emerging trends; ReCount (used by ERS), a census of food service locations (650,000 restaurants, 130,000 convenience stores, and 450,000 noncommercial locations); and CREST, another consumer database, which has 440,000 buyers and focuses on consumer use of food service establishments for meals, snacks, or drinks. With its food databases, NPD Group provides research on food consumption, restaurants, and commercial food service and eating patterns. The firm is working on adding nutrient intake to National Eating Trends and are working on analysis of local market data using CREST.

Joseph Fortson of Nielsen observed that shopper consumer behavior has changed over the years, and data such as Homescan and TDLinx (both used or formerly used by ERS) can be used to quantify those changes and help firms take advantage of trends. He noted the rise in online shopping and commented that one issue online vendors face is the alignment and coherence (or lack thereof) between federal and state regulations.

During open discussion, panel member Craig Gunderson pointed out the tendency of sources such as Homescan to underrepresent low-income consumers. He suggested that it would be useful to researchers if the number of low-income households could be increased in the Homescan data, especially in some of their longitudinal panels. Fortson pointed out that Nielsen does aggressively recruit in lower-income and diverse areas, because they are the toughest populations to capture. Burke noted that IRI surveys tend to rely on participants “opting in” but that they do target difficult-to-reach groups and the data are weighted to be representative of the population. Lesce stated that NPD has considered redirecting some surveys to specific demographics. For example, they already have a Hispanic consumer survey. Fortson noted that incentives to participate in proprietary surveys in the form of payments (albeit small ones) mean more to low-income families than they do to high-income families.

B.2. COMBINING DATA SOURCES TO ADVANCE FOOD AND NUTRITION POLICY AND RESEARCH

One example of a combined data source is the USDA Branded Food Products Database, described by Alison Krester of ILSI International, a nonprofit science foundation that is primarily funded by the food and beverage industry, and Kyle McKillop of the University of Maryland. This data source augments the USDA National Nutrient Database⁴ with nutrient composition and ingredient information on branded and store-brand food products provided by the food industry. The Branded Food Products Database project is a public-private partnership initiated by former undersecretary of USDA, Catherine Woteki. The goal of the project is to enhance public health and the sharing of open data.

The USDA Branded Food Products Database (BFPDB)—since 2019, the USDA Global Branded Food Products Database—covers 229,064 branded products from 238 food categories. Data elements include product name and generic descriptor; serving size in grams or milliliters; nutrients on the Nutrition Facts Panel per serving size and on a 100-gram basis, 100 milliliter basis, or fluid-ounce basis; ingredient list (never before captured by USDA); and date stamp associated with most current product formulation.

Linking the BFPDB to specific years of National Health and Nutrition Examination Survey (NHANES) surveys, if possible, could more accurately assess dietary intake within the United States. Having a historical record of branded and private-label foods enables comparisons of current and past consumption. Having a historical record of branded and private-label foods enables comparisons of current and past consumption. The BFPDB is in the public domain and is accessible through an Application Programing Interface (API) or directly through the Internet, where users can search, filter, and export their results.

Krester and McKillop argued that this initiative marks a paradigm shift for USDA—and that the benefit to the research community of gaining a large amount of data from food manufacturers on their food products may be a more efficient and cost-effective way of obtaining data than the usual survey approach. Next steps for the project include continuing to grow the database and creating awareness to increase its use. There are also plans for global expansion, as well as to add restaurant foods and food service products, increase private-label food items, and add foods imported into the United States. Work will also continue to align a standardized, validated algorithm to be used across all food products to determine food groupings.

___________________

⁴ The USDA National Nutrient Database and the USDA Branded Foods Database are part of USDA’s Food Composition Databases. See https://ndb.nal.usda.gov/ndb.

Next, building on the theme of combining data sources, Biing-Hwan Lin of ERS discussed a project linking ERS’s Food Availability Data System (FADS)⁵ to nutrition intake data from the Agricultural Research Service (ARS).⁶ FADS measures food commodity supplies from the farmer to domestic consumption. The central motivation behind the project is to be able to develop value-added data products that can be used to analyze both intakes and density of foods and nutrients by food source and population characteristics, as well as to measure commodity consumption by food source and population characteristics.

The project described by Lin combined food consumption data from NHANES for 2007–2010 with USDA’s Food Patterns Equivalents Database (FPED, formerly known as MyPyramid Equivalents Database) to estimate food consumption by food groups, as specified in the 2010 Dietary Guidelines for Americans. For example, a respondent in NHANES may report having eaten a specific amount of apple pie; this piece of data is then mapped into other measures such as cups of fruit, ounces of grain, grams of oils and solid fat, and teaspoons of added sugars (the commodities that make up apple pie). The consumption and nutrient content are reported by food source and can be summarized according to respondents’ demographic characteristics. In a collaborative effort with ERS and the National Center for Health Statistics (NCHS), ARS has converted the NHANES and USDA consumption intake data into 65 agricultural commodities. This Food Intakes Converted to Retail Commodities Database (FICRCD) includes retail-level commodities that fall into eight categories: dairy products; fats and oils; fruits; grains; meat, poultry, fish, and eggs; nuts; caloric sweeteners; and vegetables, dry beans, and legumes. This information was leveraged to convert ERS’s loss-adjusted food availability data (the “consumption” part of FADS), into the 65 agricultural commodities by age, income, ethnicity, and region for both food at home and food away from home.

Lin concluded by laying out data needs along with the future work plan for the project. One such data need is for farmers to better understand who consumes their commodity, where it is consumed, and how it is served. As noted above, the food consumption survey covers food (e.g., apple pie), but not at the commodity level (apples). For the loss-adjusted food availability database, future work includes building in more timely updates, expanding the number of commodities in the FICRCD, and converting food acquisition to retail commodities databases for FoodAPS (databases that capture purchases rather than consumption.)

___________________

⁵ See https://www.ers.usda.gov/data-products/food-availability-per-capita-data-system.

⁶ Andrea Carlson described another ERS project involving the ARS nutrient databases during the first workshop.

During open discussion, panel member Jim Ziliak asked what the mission directive was underlying the choice of the 65 agricultural commodities. Lin responded that category design was driven by the sample size. If there were a sufficient number of observations, the food was assumed to have been eaten quite frequently in the marketplace, and so it could be included in the commodity list. But if there were not enough observations—for example, to separate out almonds from tree nuts—then the food item remained in the more highly aggregated group.

B.3. USE OF SPECIALIZED MODULES ADDED TO FEDERAL SURVEYS

ERS has actively expanded its Consumer Food Data System (CFDS) by sponsoring or cosponsoring modules on surveys conducted by other agencies. These include the Food Security Supplement, added to many surveys, the Flexible Consumer Behavior Survey (FCBS), which has been added to NHANES, and the Eating and Health Module (EHM), added to the Bureau of Labor Statistics’ American Time Use Survey (ATUS).

Eliana Zeballos of ERS provided an overview of the EHM call out supplement to the ATUS. She said that implementation of the EHM was motivated by the need for information about individuals’ decisions on how to use their time, which can have short- and long-run implications for income and earnings, health, and other aspects of well-being. The EHM collects data to analyze relationships associated with time use, eating behavior, obesity, and other health outcomes for important subpopulations such as SNAP and WIC participants, grocery shoppers, and meal preparers. Module questions fall into the following categories: eating and drinking as a secondary activity, grocery shopping and food-away from home (FAFH) purchases, meal preparation, food sufficiency and food assistance, household income, and height, weight, and general health.

Understanding time-use patterns can provide insight into economic behaviors associated with eating patterns as well as the diet and health status of individuals. Understanding whether participants in food and nutrition assistance programs face time constraints that differ from those of nonparticipants can inform the design of food assistance and nutrition policies and programs.

The EHM has supported a number of studies along these lines. One example of findings from this literature: Zeballos and Restrepo (2018) (see also Zeballos, Todd, and Restrepo, 2019) estimate that 58.2 percent of Americans ages 18 and older reported purchasing FAFH at some point during the week before their interview. About 43 percent of individuals who received SNAP benefits in the past month made a FAFH purchase. Another finding is that about 48 percent of individuals who received SNAP benefits

had consumed a soft drink, which is 16.6 percent higher than the share of low-income, non-SNAP individuals who had done so.

Brandon Restrepo of ERS provided an overview of the FCBS, which has been fielded since 2007. The survey is “flexible” in the sense that it changes according to federal agency needs for timely, policy-relevant data. The FCBS includes a number of economic measures, including monthly income, assets, food expenditures, and participation in food and nutrition assistance programs (SNAP and WIC). It also includes dietary and behavioral measures, including self-assessed diet quality; use of packaged food labels when grocery shopping; importance of price, nutrition, and taste when grocery shopping or eating out; frequency of eating out; use of nutrition information on restaurant menus when eating out; and awareness of MyPlate and knowledge of calorie intake needs to maintain current weight.

FCBS data are valuable for informing policy evaluations of federal regulations on food labeling and its use by and impact on consumers. Restrepo concluded by stating that the goal of the FCBS going forward is to continue as a key add-on to the NHANES capable of providing timely national data to inform food and nutrition policy-making decisions.

Alisha Coleman-Jensen of ERS provided an overview of the Food Security Survey Module, describing different versions that have been used and the federal surveys onto which it has been added.⁷ She also discussed research applications of the module.

Coleman-Jensen said that a typical definition of food insecurity stipulates that the household is unable, at some time during the year, to provide adequate food for one or more of its members due to a lack of resources. In an attempt to measure this, food security survey modules have employed various structures that vary in terms of the number of items/questions (e.g., 6, 10, or 18 items), whether child items are included or not, and by reference period (e.g., 12 months or 30 days). The main federal surveys in the U.S. Household Food Security Monitoring and Research System are the Current Population Survey Food Security Supplement (CPS); the American Housing Survey (AHS); the Early Childhood Longitudinal Surveys (ECLS); FoodAPS; NHANES; the National Health Interview Survey (NHIS); the Panel Study of Income Dynamics (PSID); the Survey of Income and Program Participation (SIPP); the Survey of Program Dynamics (SPD) and a growing number of state, local, and regional studies.⁸

Going forward, Coleman-Jensen stated that ERS continues to do research on the measure. For example, ERS is assessing the Spanish-language translation and assessing comparability for households with and without children. It is also conducting Rasch analyses to assess the measurement properties of the module in all federal surveys.

___________________

⁷ See https://www.ers.usda.gov/topics/food-nutrition-assistance/food-security-in-the-us.

⁸ For more information on these surveys in the context of ERS’s CFNDS, see http://ers.usda.gov/data-products/food-security-in-the-united-states/documentation.aspx.

TABLE B.1 Planned Sample Size and Proposed Caseload for the Full Survey

SOURCE: Data from ERS. Reprinted with permission.

B.4. FOODAPS-2 STATUS; DATA USERS’ AND STAKEHOLDERS’ INPUT

The afternoon session kicked off with an update from Laurie May and Tom Krenzke of Westat, the company contracted to design and field FoodAPS-2. The overarching objective in the planning for this second generation FoodAPS vehicle is to support new analyses, including broader analyses of USDA programs, and to improve data quality. This involves changes in the survey’s sampling plan, instruments, and data collection procedures.

May and Krenzke identified the planned sample size for the full survey, as indicated in Table B.1.

These target figures are roughly comparable to those achieved for FoodAPS-1, which collected data from a sample of 4,826 households and is nationally representative. The survey targeted four groups, defined in terms of participation in SNAP and total reported household income.⁹ Sampling plan changes from FoodAPS-1 involved increasing the WIC domain’s effective sample size and creating WIC/SNAP likely-eligible flags. Other goals included improving data on children (by increasing representation) and implementing year-round data collection.

Planned changes to the survey instrument included the addition of questions covering

• more accurate school meal program information, including degree of daily participation and participation in summer meals program;
• food security (through an 18-question battery);
• subjective food needs;

___________________

⁹ For exact figures, see https://www.ers.usda.gov/data-products/foodaps-national-household-food-acquisition-and-purchase-survey/documentation.

• food sensitivities and health conditions such as diabetes, high blood pressure, and high cholesterol;
• work schedule;
• online food purchasing; and
• improved geographic data, including travel distances to stores and restaurants and geocodes for residences and food places.

May and Krenzke also described planned data collection changes. To improve the overall completeness of data collection, Westat has been working to streamline the food log data input process for respondents, add look-up databases for items, and use reminders, targeted calls, and receipts to reduce underreporting of food acquisitions. To reduce respondent burden, Westat planned to replace hard-copy food logs with electronic food logs and income worksheets. To help achieve sample size goals and reduce nonresponse bias, Westat is planning to capture interviewer observations, implement an adaptive survey design, and improve imputation for missing items.

Parke Wilde and Mehreen Ismail of Tufts University presented findings from their work collecting user feedback from researchers using FoodAPS-1.¹⁰ Wilde and Ismail reviewed 25 publications that presented results from FoodAPS, then surveyed 24 research teams that used the data. The literature review and data user survey demonstrated how FoodAPS has filled data gaps about food access and nutritional quality of food choices. Both information sources revealed that researchers largely were motivated to use FoodAPS for its high-quality, detailed coverage of food acquisitions and purchases, the food retail environment, and SNAP participation. However, it also revealed some limitations in data coverage and data quality. The specific needs identified include improvements needed in documentation, data files, and data access.

Next, Robert Moffitt of Johns Hopkins University added his assessment of the value of FoodAPS, which he called “a tremendous dataset in terms of breadth and the domain of the different types of questions related to the population’s food consumption, nutrition, and health and on programs affecting on them.” Moffitt’s work relevant to FoodAPS has mainly concerned the impact of the SNAP program on various kinds of outcomes. His comments were in part methodological—specifically on how to establish and measure causal effects of participation in SNAP. Currently, FoodAPS, as a cross-sectional data source, is fairly limited in this regard, in comparison to the monitoring functions that a longitudinal dataset can serve. Most of the literature attempting to establish the causal impact of

___________________

¹⁰ See Review of the National Household Food Acquisition and Purchase Survey (FoodAPS) from a Data User’s Perspective at https://www.ers.usda.gov/media/9776/foodaps_datauserperspective.pdf.

food stamp participation on outcomes either uses panel data (where people moving into and out of the program can be observed) or applies some form of difference-in-differences analysis comparing people in different states or where the context is changing (e.g., the SNAP rules, eligibility conditions, or benefit levels, or even just the context of the economic conditions that may be inducing people to go on and off SNAP). This kind of analysis is difficult, because there are many other differences across states at a single point in time other than the food stamp benefit or other topic of analysis.

One of Moffitt’s recommendations was that the FCNDS work toward more effective reconciliation of the SNAP program administrative datasets and the FoodAPS survey datasets. That, he said, would help alleviate difficulties in analyzing reporting errors with FoodAPS-1. He suggested that FoodAPS-2 use the same states as FoodAPS-1, especially those states that have validation data. He also suggested that ERS try to acquire SNAP histories for panel analysis.

Moffitt went on to say that reporting errors have not necessarily affected the bottom line result of policy interest—the question of whether SNAP affects outcomes or not, and by how much. Moffitt’s conclusion was that the findings about the impact of SNAP on diet quality, food expenditure, food insecurity, obesity, the Healthy Eating Index (HEI), and so on are not very different whether the input data are from FoodAPS, administrative data, or some combination of the two sources.

Panel member Diane Schanzenbach suggested that much could be accomplished by improving the CPS the Consumer Expenditure Survey, or NHANES. She asked whether there is anything that suggests how measures of SNAP (or other program participation) could be improved on these surveys. She asked Moffitt why having a separate FoodAPS is better than improving the other larger, long-standing surveys.

Moffitt replied that what FoodAPS has that the other surveys do not have are the outcome measures: the nutritional measures, the diet quality measures, the HEI, the obesity measures and food expenditures. He noted that NHANES has very small sample sizes, perhaps even smaller than FoodAPS.

Someone in the audience commented that NHANES has health and nutrition outcomes but no economic outcomes. The Consumer Expenditure Survey has economic outcomes but no nutrition information. FoodAPS is basically the only survey that can link food, nutrition, economics, and health.

Next, Susan Krebs-Smith of the National Cancer Institute spoke about the use of data from USDA’s Consumer Food Data System for health research. Her main topic was the HEI, designed to assess diet quality by showing how well any set of foods comports with the Dietary Guidelines for Americans. Krebs-Smith noted that part of USDA’s mission is to ensure the healthfulness of the U.S. food supply. USDA is a partner in developing the Dietary Guidelines for Americans.

The HEI consists of 13 food group components (total fruits, whole fruits, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, fatty acids, refined grains, sodium, added sugars, and saturated fats). Weights for constructing the index are derived from the Dietary Guidelines. The index requires information on the quantities of all of the food groups to generate the total score. It captures the balance among food groups, including foods to encourage and foods to reduce.

In addition to a total score, there is a score on each of the components. If the total HEI score is 100, all of the components are at their optimal level. If the total score is zero, every component is zero or very low. If the total score is, say, 55 (a very common score), the significance is not clear—for example, it remains unknown whether the diet is high on meats and low on vegetables, or vice versa.

One advantage of the HEI is that scores can be constructed at different levels in the food supply chain, from the commodities produced by farmers to the foods consumed by ultimate consumers and anything in between. In order to examine the HEI at different levels of the food chain, the foods or commodities at that level need to be identified and classified into the 13 categories of the HEI. Earlier during the day’s meeting, Biing-Hwan Lin described translating the data from the ERS food supply system, FADS. With his apple pie example, he explained the difference between foods as eaten (like apple pie), commodities (apples), and the nutrients associated with them. For some levels of the food chain, that linkage can be done by using databases maintained and updated by ARS in collaboration with others. Constructing the HEI for FADS requires data from both the Nutrient Availability Database and the U.S. Salt Institute. Constructing the HEI for food consumption data from NHANES requires the Food Patterns Equivalent Database and the Nutrient Database. Crosswalk databases are still needed for food processing establishments and for the community food environment.

With the appropriate crosswalks, the HEI can be used to evaluate the “diet quality” associated with grocery store purchases, grocery store circulars, where food is obtained (e.g., different kinds of restaurants or fast food outlets), schools, food pantries, and so on. Also, in addition to using the index for comparison and monitoring, it could be used to analyze the relationships between diet patterns and health outcomes.

Krebs-Smith noted that the HEI provides a standardized measure across multiple levels and thinks this has real advantages. Making food comparable and available at all levels will involve overcoming some infrastructure challenges, which she thinks would be most useful to do.

Melissa Abelev of the Food and Nutrition Service (FNS) noted that FNS administers the USDA hunger programs: SNAP, WIC, the Child and Adult Care Food Program (CACFP), and the School Lunch and School Breakfast programs. FNS analysis groups provide cost data for budget analyses,

analyzing how policy changes might impact the cost of food in FNS programs, how they might affect participation in the program, and how they might impact the overall cost of the program. They look at food security, the alignment of diets with nutritional standards, and program integrity and operations. They use a range of ERS and other data sources, including those of the Census Bureau and Bureau of Labor Statistics.

FNS has helped to fund a number of ERS initiatives, including FoodAPS and adding the Food Security Module to a variety of surveys. Abelev noted that FNS has not used FoodAPS data, however, primarily because the data are not user-friendly. FNS hopes to be able to use these data in the future, or to be better able to use the data from FoodAPS-2.

Following up on many of these themes, during open discussion participants discussed the strengths and weaknesses of using FoodAPS, and ideas for making future iterations more powerful. Panel member Eric Rimm pointed out that by the time FoodAPS-2 comes out, the online purchasing environment will have changed and become even more complex. Online shopping will become an important form of food acquisition. Laurie May and Tom Krenzke agreed, noting that they think in 2 years most food items will be available online. They stated that the online questionnaires had not yet been finalized, but agreed that nuances needed to be taken into account. At the event level, they will likely ask whether the purchase was online, delivered, or picked up in a grocery store. They will also collect information on items purchased.

Panel member Amy O’Hara asked about plans for broader use of data matching, specifically matching FoodAPS-2 to administrative data such as SNAP, WIC, Medicare, and Social Security data. May replied that Westat plans to use WIC and SNAP as part of the sampling plan. They are not currently planning to bring in other datasets. Mark Denbaly of ERS noted that simply getting ahold of SNAP and WIC data and matching them requires a heroic effort. O’Hara noted that ERC could buy Medicaid data and should consider how linkages might be facilitated several years down the road.

B.5. MEETING AGENDA

Panel on Improving USDA’s Consumer Data for Food and Nutrition Policy Research
Second Meeting, June 14, 2018

The National Academy of Sciences Building, Room 120
2101 Constitution Ave NW, Washington, DC

Meeting Goals: The panel’s second meeting will include a number of presentations geared toward informing the panel as it considers its charge and begins shaping a strategy for producing a report that fully addresses it.

Topics of interest for this meeting are the current and potential use of commercial and other non-government, non-survey data sources; users’ perspectives on directions for ERS’s FoodAPS survey; and linking data sources. Meeting #3 will follow up further on some of these topics. During closed session, the panel will review its charge, begin shaping a report outline, and identify key topics to address during its Fall meeting.

Day 1, June 14: Open Public Sessions