Appendix A

Summary, First Meeting, April 16, 2018

A.1. OVERVIEW OF ERS’S VISION AND STRATEGY FOR IMPROVING DATA FOR FOOD AND NUTRITION POLICY RESEARCH

The panel’s first open meeting,¹ held April 16, 2018, consisted of a set of overview presentations by the U.S. Department of Agriculture’s (USDA’s) Economic Research Service (ERS) staff describing current projects in the agency’s Consumer Food Data System (CFDS) portfolio and outlining priorities going forward. Panel members and meeting participants were informed about key program initiatives, including: plans for a possible second iteration of ERS’s National Household Food Acquisition and Purchase Survey (FoodAPS); plans to make greater use of proprietary data sources; and continued development of linkages across multiple data sources (survey and nonsurvey) and of supplemental modules to surveys conducted by other federal statistical agencies. Research highlights emerging from CFDS program initiatives were also summarized.

Following introductory comments by Marianne Bitler, the panel chair, meeting participants were welcomed by Brian Harris-Kojetin, director of the Committee on National Statistics, and Monica Feit, deputy director of the Division of Behavioral and Social Sciences and Education, who also described the National Academies’ study process. During this opening session, ERS leadership provided an overview of the agency’s vision and strategy for improving data for food and nutrition policy research and specified their goals and objectives for commissioning the study. Mary Bohman, admin-

___________________

¹ The meeting agenda appears at the end of this appendix.

istrator of ERS, outlined the blueprint for the current CFDS program—describing its components, organization, and purpose—and its operational context within the agency’s mission: to inform and enhance public and private decision making on economic and policy issues related to agriculture, food, the environment, and rural development.² Within this ERS mission, the role of the Food Economics Division is to

• take stock of contemporary and anticipated food policy and program objectives and market trends and dynamics;
• develop the necessary data and information infrastructure to examine the evolving questions; and
• produce the right products and information for the Administration, the Congress, and the public on consumer food choice behaviors and outcomes such as nutrition and health.

Bohman stated that the mandate for the Food Economics Division is to build a comprehensive, integrated data system focusing on consumer data to efficiently deliver credible evidence for informing policy and to facilitate production of research findings so that they are in place when food and nutrition-related policy and program needs arise.

Bohman further raised the question, central to the panel’s charge, of whether new kinds of data sources perform as needed, and to what extent they might replace, supplement, or complement current data sources, primarily surveys. She stated that the main task for the panel was to help the agency chart its course going forward, particularly how it can most effectively put to use its $80 million investments in research, statistics, and data. She argued that the panel’s report would have a substantial influence on the agency’s ERS research and data strategies.

Jay Variyam, division director, ERS, described the motivation, vision, and action guiding CFDS activity. The motivation: policy needs drive data investments. The vision: to build a comprehensive, integrated data system to efficiently deliver credible evidence for informing policy. And the action: to develop a multipronged data approach to meet research and policy needs.

The most important policy areas and questions facing the Food Economics Division identified by Variyam are these:

• Diets, Nutrition, and Obesity—What foods do households buy, how much do they pay, where do they shop, and what is the nutritional quality of these purchases?
• Food and Nutrition Safety Net—How are USDA’s Supplemental Nutrition Assistance Program (SNAP) participants and low-income

___________________

² See https://www.ers.usda.gov/about-ers.

• households similar or different when taking diet, nutrition, and obesity into account?

• Changing Consumer Preferences—How do consumers respond to price changes, new information, and varying product attributes?
• Food Environment and Affordability—What role does the food environment play in consumers’ food choices? Does ease of access matter for the nutritional quality of purchases?
• Industry Response and Changing Food Supply—How is the food supply changing in response to consumer preferences for convenience, nutrition, and production attributes, and what are the nutritional implications?
• Agricultural Sector Adaptations—How will the agricultural sector adapt to changing consumer preferences, and what are the resource implications?

Echoing Bohman’s comments, Variyam envisioned a multiple-data approach for the agency. Both beyond and in coordination with FoodAPS and other surveys, other forms of data will be involved, including proprietary scanner data, linked administrative data, food store data, and linked nutrition and food acquisition data. The motivation behind this multipronged data approach is that, even given the impressive amount (and quality) of data on food and food program participation that exists across federal statistical agencies, these sources are still insufficient to answer key questions about consumer choices, food acquisitions, industry response, and the role and effectiveness of government programs.

Mark Denbaly, deputy division director for Food Economics Data, ERS, elaborated on the multiple-data-source approach. In an era of increasing costs and decreasing public willingness to spend time completing surveys, proprietary and other alternative data are gaining in importance across the statistical system. Examples of this trend within the CFDS at ERS include use of the following:

• IRI household item-level data on grocery purchases,³
• IRI retail store item-level sales data,
• IRI product descriptions and attributes for about 1 million UPCs,
• Nielsen store characteristics and geocoded locations (TDLinx⁴), and
• NPD restaurant locations and characteristics (ReCount⁵).

___________________

³ For more information on IRI products used by ERS, see https://www.ers.usda.gov/webdocs/publications/47633/57105_tb-1942.pdf?v=0.

⁴ For more information, see https://www.nielsen.com/us/en/press-releases/2017/nielsen-tdlinx-announces-new-channel-classification-for-dining-industry.

⁵ For more information, see https://www.npd.com/wps/portal/npd/us/news/press-releases/2018/total-us-restaurant-count-at-647288-a-drop-from-last-year-due-to-decline-in-independent-restaurant-units-reports-npd.

Integration across data sources is another key element of ERS’s state data strategy. Among data-combining initiatives undertaken by the agency to date are these:

• integrating USDA’s Agricultural Research Service (ARS) nutrient information with IRI scanner data, an interagency effort that also involves USDA’s Center for Nutrition Policy (CNPP);
• using a geospatial data system to provide precise information about food retail environments; and
• linking agency administrative records to survey data—the purpose of the Next Generation Data Platform (as described in Chapter 2, this is an interagency effort with the Census Bureau and USDA’s Food and Nutrition Services [FNS]).

A third element of ERS’s CFDS strategy involves developing supplements to existing surveys. The agency has already had success with this strategy—for example, the Flexible Consumer Behavior Survey, which was added to the Centers for Disease Control and Prevention’s (CDC’s) National Health and Nutrition Examination Survey (NHANES),⁶ the Eating and Health Module added to the Bureau of Labor Statistics’ (BLS’s) Time Use Survey, and the Food Security Module, which was added to CDC’s National Health Interview Survey, and is open to greater use of modules where subject matter synergies arise.

Although covered in greater detail by other presenters, Denbaly provided a brief overview of FoodAPS. Designed in consultation with academic and government leaders and experts (and jointly sponsored with FNS), the survey

• integrates multiple types of information from multiple sources;
• brings together food, economics, nutrition, health, program participation, and environmental factors;
• focuses on food acquisition, not on food intake;
• is more than an expenditure study, as it includes prices and quantities of acquired food at the item level;
• includes acquisition of food items consumed at home and food items consumed away from home, as well as free foods; and
• is the only source of information on food items that participants acquire using program benefits and their own resources.⁷

___________________

⁶ For more information, see https://www.ers.usda.gov/topics/food-choices-health/food-consumption-demand/flexible-consumer-behavior-survey.

⁷ Additional details can be found at https://www.ers.usda.gov/foodaps.

A.2. DISCUSSION OF THE PROJECT STATEMENT OF TASK AND PRIORITIZATION OF TOPICS FOR THE STUDY

Following review by ERS leadership presentations, there was open discussion of the project Statement of Task and prioritization of topics for the study. Panel members provided their perspectives, identifying key issues embodied in the charge and discussing their primary interests related to the study. This discussion yielded the revised Statement of Task (see Chapter 1).

During discussion of the Statement of Task, Mark Denbaly noted the importance of identifying the most important policy questions to be answered; these questions, in turn, would drive how the division invested in its data infrastructure. Many of the thoughts on these issues are encapsulated in the above-referenced white paper produced by ERS for the panel. Denbaly asked that the panel consider the return on investment for various data infrastructure options. Panel members noted that “assessing the value” of different data investments requires identifying a cost/benefit metric—which might consider research-enabled use of data in program administration, amount of staff time, and dollar costs. These metrics are not readily available. All agreed that informing policy was a top priority.

A major question raised by Denbaly was what to do with FoodAPS. How can it be improved in regard to reduced burden and improved location and price information? Should future iterations of the survey be pursued and, if so, how should they be specified? And, what are the alternative uses of FoodAPS resources, and could these alternatives provide the same value to researchers and policy makers? To begin framing these questions, the next session consisted of presentations by ERS staff to inform panel members about current CFDS data programs, research activities, and program plans. The session was meant to stimulate panel thinking about what additional kinds of information the panel might need in order to carry out its charge.

A.3. CURRENT CFDS PROGRAMS, RESEARCH ACTIVITIES, AND PLANS

Kicking off a session on “Current data programs, research activities, and program plans,” David Levin and Megan Sweitzer, both ERS economists, described use of proprietary data by the Food Economics Division. One prominent example is the use by the agency of data collected and processed by the company, IRI (https://www.iriworldwide.com/en-US/Company/About-Us). IRI collects proprietary scanner data on consumer purchase transactions, retail point-of-sales (purchase transaction records are collected from store systems), household scanner data for store purchases

(which can be sometimes be linked with household demographics), and product and store information.⁸

Among the advantages of scanner data cited by Levin and Sweitzer is their granularity. Food price data can be pinpointed geographically to individual stores or market areas. Product detail is available at the level of each individual UPC/item, to which descriptions and attributes are attached. Data are often available on a weekly basis. Sample sizes are large, and in some cases long-term panels of households have been constructed.

Scanner data have been integrated more broadly across USDA into cost estimates and evaluations of USDA programs. Projects to estimate the weights of Special Supplemental Nutrition Program for Women, Infants, and Children (WIC) food packages and the retail value of the average Food Distribution Program on Indian Reservations (FDPIR) food package are two examples.

Limitations of scanner data noted by Levin and Sweitzer have to do with their representativeness and accessibility in a documented format. Future work could usefully be conducted on improving the representativeness of the retail data and developing weights for the retail stores. Creating a capacity to link stores in household and retail data would also be valuable, as would creating new identifiers/variables for SNAP and WIC purchases.

Indeed, ERS has plans in place to expand scanner data applications. One example is the Quarterly Food-at-Home Price Database (QFAHPD), which aggregates food purchases from Nielsen Homescan for more than 50 food groups (available to the public on the ERS website). Another is in future iterations of FoodAPS to support product identification and food environment studies. The panel was asked to weigh in on these issues in its report, to provide a sense of the way these different commercial datasets may be used in stand-alone applications by researchers, and in what ways they may be combined with survey or other nonsurvey data in the CFDS.

Andrea Carlson, ERS, elaborated on the project to integrate USDA’s ARS nutrient information with IRI scanner data, an interagency effort also involving USDA’s Center for Nutrition Policy and Promotion (CNPP) and ARS. The challenge is to create prices for foods consumed as collected by NHANES, preferably using automated methods, and to match them with nutrient data. The first step was to create a crosswalk, the Purchase-to-Plate Crosswalk, between purchased foods from scanner data (45,000 codes) and foods found in FNDDS (about 5,000 codes) and develop prices for foods reported on NHANES as consumed (8,000 items). The terms used for foods are different in the two sources. A semi-automatic approach has been tested with data from two time periods.

___________________

⁸ A full description of ERS use of scanner and other kinds of commercial data can be found in Chapter 2.

The project also developed Food Purchase and Acquisition Food Groups (F-PAG)—now called ERS Food Purchase Groups (EFPG)—which assign IRI UPCs to USDA-related food groups, based on ingredients, nutritional content, and convenience to consumer and store aisle. An earlier version of this database is linked to Food-APS for studying the nutritional content of purchased foods and the F-PAG are similar to the groups used to prepare the Quarterly Food at Home Price Database. The project also created the Food and Nutrition Database for Dietary Study (FNDDS)⁹ and the Purchase to Plate Price Tool, which estimates prices for individual foods consumed as reported in What We Eat in America (WWEIA) and NHANES. This tool supports analysis of the relationship between food prices and nutritional content.

The tools have been used to augment publicly available data from existing surveys. External users have restricted access to them. New external data products that resulted include these:

• ERS Quarterly Food at Home Price Database (QFAHPD): https://www.ers.usda.gov/data-products/quarterly-food-at-home-price-database/ERS
• ERS Fruit and Vegetable Prices (FVP) data product: https://www.ers.usda.gov/data-products/fruit-and-vegetable-prices.aspx
• Food Consumption and Nutrient Intakes 2007–2010, based on NHANES data: https://www.ers.usda.gov/data-products/food-consumption-and-nutrient-intakes/
• Commodity Consumption by Population Characteristics 1994–2008, using FADS and NHANES: https://www.ers.usda.gov/data-products/commodity-consumption-by-population-characteristics/

Next, Michele (Shelly) Ver Ploeg, chief of the Food Assistance Branch, ERS, presented options for linking survey data to food environment data. The motivation for these kinds of data linkages is to be able to address research questions concerning whether Americans’ diets are out of balance with dietary guidelines and, if so, to what extent it is due to a lack of access to healthy and affordable foods and to what extent to the larger food environment (e.g., availability of stores and restaurants, variation in food prices, food policies, and community characteristics) that may influence food choices and diet quality.

To pursue the measurement of these accessibility issues, ERS has invested in proprietary food retailer and restaurant data, which have been combined

___________________

⁹ FNDDS is a database maintained by Agricultural Research Service (ARS) that contains information on foods, their nutrient content values, and the weights of portions. It is used to analyze the nutrient content of foods consumed as reported in WWEIA/NHANES.

with survey and administrative data (e.g., population and food assistance program data) to produce food access and food environment indicators. Ver Ploeg described a number of products that have emerged from this initiative. One of these, the Food Access Research Atlas, provides a spatial overview of access to supermarkets, supercenters, and large grocery stores. Another, the Food Environment Atlas, distills more than 200 indicators of a community’s ability to access healthy food and its success in doing so. The Food Environment Atlas includes county-level detail for most indicators, as well as indicators of store and restaurant availability, food assistance use, food prices and taxes, local foods initiatives, and health and physical activity.

ERS’s data and mapping tools are used in research, in policy, and by planners. The FoodAPS geography component adapts the Atlas information in the construction of its survey instruments and to characterize the food retail environment in a given primary or secondary sampling unit. These data and mapping tools have also been linked to other surveys, including the IRI Consumer Panel, Panel Study of Income Dynamics (PSID), the Health and Retirement Study (HRS), NHANES, and CPS. Policy applications include the Healthy Food Finance Initiative (Treasury, HHS, USDA) and SNAP store authorization regulations. And, among community planners and local governments, the Atlases have consistently been among the ERS products with the greatest number of web views.

Mark Prell, ERS, presented information about the Next Generation Data Platform, which is a strategic partnership among ERS, Food and Nutrition Services, and the U.S. Census Bureau (Census) to promote record linking for research purposes. The platform is a long-term effort to acquire state-level administrative microdata for SNAP and WIC that can then be linked to Census survey data and administrative files from other federal agencies in a secure data environment to support research on USDA programs.

As articulated by Prell, the Next Generation Data Platform project goals are to inform policy makers, managers, and the public on (i) who participates in USDA food assistance programs; (ii) how participation affects people’s lives; and (iii) who does not participate and why. The Census Bureau brings to the project the data infrastructure expertise to inform decisions regarding the use of surveys (including the 2020 Census), data-linkage processes and regulation, and linkage agreements with other federal agencies.

Among the benefits of SNAP administrative data are that they include information on the universe of SNAP participants in a given state and that these data are known for their completeness and accuracy. Among the benefits of American Community Survey (ACS) data are that they are a random sample of both SNAP and non-SNAP participants and include demographic information as well as annual income data—used to model SNAP income eligibility. The benefit of linking SNAP and ACS data is that the strength of each data source can be leveraged.

During the day’s final open session, Elina Page of ERS presented on information about FoodAPS and plans for future iterations of the survey. She began by outlining the needs for a data source such as FoodAPS. The primary objective of the program is to inform policy on diet-related health issues, an important policy issue. The economic burden of diet-related diseases reaches into the trillions of dollars each year for the nation. For 2016, the cost of obesity and overweight conditions alone—in terms of direct expenditures and lost productivity—was estimated to be $1.42 trillion (Milken Institute, 2016; Benjamin et al., 2017). Cardiovascular diseases ($316 billion) and type 2 diabetes ($320 billion) are the two costliest condition categories.

Page pointed out that the high level of program spending to improve the population’s health provides another pressing policy impetus for collecting accurate and timely information. In 2016, Medicare and Medicaid spending reached $672 billion and $566 billion, respectively. For 2017, expenditures on other key programs were as follows:

• all food assistance programs, $99 billion;
• SNAP, $68 billion (for an average monthly participation of 42 million);
• WIC, $6 billion (for an average monthly participation of 7 million);
• National School Lunch Program, $14 billion (for an average daily participation of 30 million); and
• National School Breakfast Program, $4 billion (for an average daily participation of 15 million (Oliveira, 2018).

Table A.1, from Page’s presentation, lists key data sources for documenting these expenditures and their effectiveness at fulfilling their stated goals.

As indicated by gaps in Table A.1, it is clear that current information needs are met only incompletely. FoodAPS was designed to address these shortcomings by collecting comprehensive data on household food purchases and acquisitions; foods from food-at-home (FAH) retailers; food-away-from-home (FAFH) establishments; and foods obtained for free. Information is reported by all household members over a 7-day period during the period in which the survey was fielded (from April 2012 to January 2013).

Page described the FoodAPS sample as nationally representative of U.S. households with four target populations: (i) SNAP households, (ii) non-SNAP households with income < 100 percent of the federal poverty guideline, (iii) non-SNAP households with income ≥ 100 percent and < 185 percent of the federal poverty guideline, and (iv) non-SNAP households with income ≥ 185 percent of the federal poverty guideline. Table A.2 describes the composition of the FoodAPS participants in more detail.

Goals for future iterations of FoodAPS would be to capture higher quality data, reduce nonresponse bias, reduce respondent burden and reporting fatigue, and reduce processing time. A key strategy for accomplishing these

TABLE A.1 Data Sources

	SIPP (Survey of Income and Program Participation)	NHANES (National Health and Nutrition Examination Survey)	CE (Consumer Expenditure Survey)	Proprietary Consumer Panels
	Census	NCHS CDC	BLS	Nielsen Homescan or the IRI Consumer Network
Food-at-home purchases
Food-away-from-home purchases
Free food acquisitions
Household unit
Food assistance program participation
Demographics

= included

= included with limitations

SOURCE: Presentation to the panel by Elina T. Page, April 16, 2018. Reprinted with permission.

goals is to continue exploring data linkage options. FoodAPS survey data are already linked to the extant data to reduce respondent burden and enhance data analysis. As described above, proprietary scanner data were used to create item descriptions and weights. SNAP administrative records were used in sampling frame and data quality checks. Thirteen data sources were used to enhance the FoodAPS geography component—specifically, to fill in details (location and density of retailers, measures of access to these retailers, local food prices, and area demographics) about the local food environments. Finally, the USDA food nutrient databases were used to add micro- and macro-nutrient content and food pattern equivalents to the FoodAPS-generated micro record.

Page concluded her presentation by posing the following questions for the panel to consider as it deliberates on its charge and, hopefully, to answer: Where is FoodAPS headed? Is FoodAPS worth the investment? And should FoodAPS be a permanent data collection effort?

A.4. MEETING AGENDA

Panel on Improving USDA’s Consumer Data for Food and Nutrition Policy Research

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

Meeting Goals: The panel’s first meeting consists of a set of high-level overview presentations by ERS staff about current projects and priorities of the agency’s Consumer Food Data System (CFDS) program. These presentations will inform panel members, and meeting participants more broadly, about key developments—exploiting proprietary data, developing linkages across disparate data sources, adding supplements to exiting surveys, and continued planning for FoodAPS-2—pushing forward the CFDS data infrastructure. Research highlights emerging from various CFDS program initiatives will also be summarized. During this afternoon closed session, the panel will review (and, if necessary, refine) its charge, attend to institutional requirements, and finalize planning of an open session for the project’s second meeting in June.

Open Public Sessions, 9:00 a.m.–3:00 p.m.