National Academies Press: OpenBook

A Consumer Food Data System for 2030 and Beyond (2020)

Chapter: 4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System

« Previous: 3 Data and Knowledge Gaps
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

4

Strategies to Strengthen the Infrastructure of a Consumer Food Data System

4.1. DESIRABLE CHARACTERISTICS OF A CONSUMER FOOD DATA SYSTEM

This panel was charged with reviewing the Consumer Food Data System (CFDS) program for the Economic Research Service (ERS) and providing guidance for its advancement over the next 10 years. As part of this charge, the panel was asked “to identify data gaps and to anticipate how evolving policy priorities may affect data needs.” Recognizing that the objective of the CFDS program is to advance understanding of food acquisition, behavior, and outcomes, the panel identified characteristics of a CFDS that is effective and useful for research and policy purposes. These include comprehensiveness, representativeness, timeliness, openness, flexibility, accuracy, suitability, and fiscal responsibility. These characteristics are aspirational for the CFDS in toto and may not be met in any one data resource.

Comprehensiveness

A data system that is effective for monitoring the levels and trends in food behaviors and outcomes and for identifying the effects of public programs and policies on those behaviors requires comprehensive data. These data need to come from a variety of sources and to span multiple topics. Surveys are useful in documenting socioeconomic factors that affect food behaviors and outcomes, such as family/household structure, age, gender, race, education, employment, income, health status, (nonfood) consumption,

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

wealth, time use, and geography, among others. Traditionally surveys have also been the main source for data on program participation within the Supplemental Nutrition Assistance Program (SNAP), Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), Temporary Assistance for Needy Families (TANF), and other safety net programs.

However, surveys have been decreasingly reliable for such analyses, owing to rising rates of nonresponse. Further, surveys suffer from respondent error in reporting program participation. (Meyer, Mok, and Sullivan, 2015; Bollinger et al., 2019). When administrative data are linked to surveys, the combination provides improved accuracy relative to surveys alone for measurement of and the evaluation of transfer programs (concerning both participation and benefit levels). Independent of their linkage to surveys, administrative data are useful for purposes of general program monitoring, as well as for certain forms of evaluation such as “leaver” studies.

Because consumer food choices respond to economic, policy, and environmental incentives, an effective food data system also requires access to comprehensive information on food prices, food policies, food outlets, and the spectrum of food choices within those outlets. Some granular data on prices, outlets, and choices can be obtained from surveys of markets, directly provided by food vendors, or from third-party private aggregators such as Nielsen and IRI. Information on food policies at the federal, state, and local level is essential to understanding the constraints and options facing potential recipients and thus is useful in nonexperimental evaluations of food assistance programs. An exemplar of the latter is the SNAP Policy Database, currently collected by ERS.

Representativeness

Data on food behaviors and outcomes are most useful if they are representative of the U.S. population, both nationally and at component aggregations such as states. National-level representativeness is needed to accurately assess aggregate levels and trends. Because many food and health programs and policies vary across states, a data system that is of adequate size and representative of the diversity of households at the state level is desirable. Given ERS’s important focus on rural areas as well as the rest of the country, representativeness along the urban-rural continuum is also desirable. Household surveys that are representative at the substate level are generally cost-prohibitive; however, administrative and scanner data are generally of high value-added at the substate level owing to their very large samples, and administrative data also do not suffer from coverage or nonresponse issues within the population of program participants.

One concern about extant scanner and some privately collected commercial data is their lack of coverage in rural areas. Thus, having a data

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

system that also reflects the food environment for rural and other hard-to-reach populations, in addition to reflecting the rest of the country, should be a goal of an effective CFDS.

In addition to providing comprehensive data, an effective data system would sample the same households, firms, or geographies repeatedly over time. Ideally these data would be longitudinal in that they follow the same households or firms over time without substantial attrition, but repeated cross-sections of households or firms collected from the same geographic areas over time are also well suited for causal research designs with observational data. Administrative and scanner data lend themselves to longitudinal data formats, since individuals and firms can be readily linked over time with unique IDs (e.g., by Social Security number, Employer Identification Number, or proprietary identifiers). Repeated household measures are preferred when there is not substantial attrition or nonresponse. Nevertheless, much can be learned from repeated cross-sectional data, for example by exploiting changes in the policy environment across states and over time. Whether panel or repeated cross-section, the data are most effective for monitoring and evaluation if the questionnaire’s content and structure are stable over time.

Timeliness

To have maximum program and policy impact, an effective data system needs to collect data at regular intervals, and its data metrics must be consistent over time to allow accurate tracking of trends. The interval of data collection will differ depending on the programmatic need. Many monitoring functions, including the measurement of program participation in food assistance in SNAP, WIC, and school meal programs, require data at a monthly frequency, while other monitoring, including the tracking of health and nutrition outcomes such as diabetes and obesity, is more slow-moving and can be sufficiently handled by annual data collection. Many evaluations of behavioral outcomes are also effectively conducted with annual data. Thus, the minimum interval for collecting data on the program policy environment is annual.

Openness

A data system is effective if it is open and accessible to the public and to the policy and research communities, although the degree of openness should vary based on the “need to know.” Because the programs and data are collected with taxpayer funds, some data used to monitor program policies and participation, as well as health and dietary outcomes, should be readily accessible to the general public for the sake of transparency

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

concerning program reach and operations. Generally, such data are currently publicly available, aggregated at the county, state, or national level over time.

For some nonexperimental monitoring and evaluations of food behaviors and outcomes, a de-identified individual-level dataset (at the household or firm level) to which the public has open access for research purposes is desirable. To be most effective, such data should contain identifying geographic information but restricted to a level sufficient to protect respondent confidentiality, such as state of residence or, in some cases, county of residence. This approach permits merging the data with state-level or county-level information from other sources (e.g., the SNAP Policy Database or the Bureau of Labor Statistics’ state and county unemployment rates), which is standard practice in nonexperimental evaluations. Some monitoring and evaluations of food and health outcomes require access to more granular geographic data, such as latitude and longitude of location or the Census block or tract level.

Still other research requires access to the individual or firm IDs, for example to link survey data to administrative data, or else across administrative data sources. In such cases, policies and procedures are needed (and indeed are in place) to ensure that access to the restricted data is limited to qualified researchers while protecting privacy. One model for accomplishing this is that of the Federal Statistical Research Data Centers (FSRDCs)—a partnership between federal statistical agencies and leading research institutions in which secure facilities provide authorized access to restricted-use microdata for statistical purposes only.1 Further examples were proposed by the bipartisan U.S. Commission on Evidence-Based Policymaking.2

ERS offered an alternative to the FSRDC system for those who wished to use restricted versions of National Household Food Acquisition and Purchase Survey (FoodAPS), but access to the IRI data linked to the FoodAPS would have required signing an indemnity clause, which is forbidden for many researchers at state universities, and thus would have failed the open-access goal of a desirable data system. Policies and procedures for access to restricted versions of the various datasets should be established in cooperation with representatives from the user community.

Human subjects’ protections and privacy rules sometimes limit the way data may be shared. Hence, the CFDS should be conceived in a modular

___________________

1 See https://www.census.gov/fsrdc.

2 The Commission was a 15-member group of experts charged by the U.S. Congress and the president with examining how government could better use its existing data sources to provide high-quality evidence for policy and government decision making. The Commission was created in March 2016 by the Evidence-Based Policymaking Commission Act (P.L. 114-140), legislation jointly filed by Speaker of the House Paul Ryan (R-WI) and Senator Patty Murray (D-WA) https://www.congress.gov/bill/114th-congress/house-bill/1831.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

fashion, with each type of data being shared in the most open manner consistent with human subjects’ protections and privacy rules. Personal data are protected under state and local laws, which require agencies to prevent unauthorized access through security controls on the information technology systems that process and store data. Privacy protections also extend beyond security controls. Agencies decide who can use program data (e.g., employees, contractors, and research partners) and for what purposes (e.g., program evaluation, program improvement, research, and compliance reporting). To support uniform, secure access to administrative data, ERS can provide interpretation of federal statutes and data management protocols to streamline data comparisons and linkages. ERS can also provide guidance on reducing privacy risks in published data aggregates and reports, including disclosure avoidance tools and checklists.

Data access should not be limited to groups with close connections to USDA. For example, Nielsen data must be protected, but its price data in aggregate form is shared in the Quarterly Food-at-Home Price Index constructed by ERS. Similarly, FoodAPS data are shared through a data enclave with NORC at the University of Chicago,3 but they are also available in less detail through public-use files.

Access should also be timely and not require a huge financial burden, thus permitting their use by a broader set of researchers, including those with expertise in economics, nutrition, health policy, geographic information systems, and clinical care. Of course, the USDA ERS staff are perhaps the most expert users of some of the data in the CFDS, given their role in creating it, but facilitating more outside access would also be useful for science and policy.

Flexibility

Ideally, investments in food and consumer data go on to support (i) research applications that were planned in advance, (ii) unanticipated applications generated by a broad, entrepreneurial, and inventive community of research users, and (iii) efforts to evaluate unanticipated changes in policy and in food retail markets.

ERS’s development and inclusion of the Household Food Security Module as a supplement to the Current Population Survey (CPS; prompted by a congressional request) was crucial in that it unleashed an entirely new research and policy agenda. This has allowed the research and policy communities to plan, years in advance, for reports on food insecurity to coincide with the annual release of the data. Another example of planned use was the design of FoodAPS, which allowed researchers to study how the

___________________

3 See http://www.norc.org/Research/Capabilities/Pages/data-enclave.aspx.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

SNAP issuance cycle affects food acquisition or diet quality (Smith et al., 2016; Kuhn, 2018; Whiteman, Chrisinger, and Hillie, 2018).

However, in some cases new ideas or policies have emerged that were unanticipated. Similarly, new forms of food acquisition are emerging, such as online delivery. Thus, a desirable data system must be elastic to respond to such innovations.

Accuracy

Accurate measurement and reporting are the foundation of effective evidence-based policy making, so a desirable data system is one that seeks continuous quality checking and improvement. For surveys this entails, among other things, minimizing nonresponse to questions or to the survey itself as well as minimizing reporting error. Linking survey data to administrative data offers the prospect of better measurement of household participation in assistance programs when links are of high quality, but administrative data, which generally originate from state governments, are not devoid of measurement error. Scanner data on persons and establishments, while rich in granularity, also suffer from underreporting of certain items and often lack coverage of certain populations, notably low-income people and those residing in rural areas. They also often fail to include all the outcomes of interest. Thus, a program of ongoing studies to assess the quality, coverage, and comprehensiveness of surveys, administrative records, and scanner data is needed.

Suitability

While some CFDS purposes are descriptive, others require cause-and-effect inference. The CFDS should anticipate the implications that the desire for achieving causal results may have in its data design. These include the collection and sharing of policy variables for use in executing quasi-experimental designs, the use of program data (or surveys that include nonparticipants) as sampling frames for potential program evaluations using random-assignment experimental research designs, and the use of administrative data to improve inference based on faulty self-reports. They also include the use of longitudinal data for statistical analyses that control for certain types of time-constant and location-constant confounding variables in estimating causal effects, or the use of other econometric approaches offering causal insight (e.g., instrumental variables, Regression Discontinuity Design). They also include the curation of data to maintain version control and enable archiving to support replication.

Features of a (nonexperimental) data system that facilitate strong causal research designs include (i) the provision of sampling frames through

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

administrative data that can be used for random assignment or survey purposes; (ii) the provision of comparison data that are nationally representative for use in understanding the study populations through nonexperimental evaluations; (iii) integration with policy information as explanatory variables (as has been emphasized in the SNAP rules parts of this report); (iv) longitudinal or panel structures for use in fixed-effects models that control for unobserved time-constant confounding variables; and (v) inclusion of appropriate administrative data on program participation linked with nationally or regionally representative survey or administrative data on the population of potentially eligible persons.

Fiscal Responsibility

Taxpayer dollars should be spent wisely. This is especially true today in an era of tightening statistical agency budgets. The CFDS should maximize the research value of federal dollars invested in the data system through its combined impact on improved program monitoring, improved monitoring of the nutritional status, food security, and health of the population, and strengthened ability to conduct research estimating the causal linkages between programs and outcomes. ERS’s CFDS strategy should encompass both investments in special-purpose surveys and initiatives to enhance the research value of administrative data, survey data, and other sources of data already being collected for nonresearch purposes, such as proprietary commercial data. Investment into data products should be diversified to allow for unexpected research directions.

Achieving the above-described characteristics in a data system to support food and nutrition research requires taking a multipronged approach involving survey, administrative, and commercial data.4 The 20th century survey-centric federal statistical system is at a crossroads: Declining response rates have led to surveys becoming more costly and the resulting data possibly becoming less accurate or generalizable, while lower-burden complementary or substitute administrative and proprietary data sources have emerged. The report of the Commission on Evidence-Based Policymaking (2017) lays out many of the challenges and advantages of combining different types of data. Among them are (i) the changes in consumer food shopping modes (e.g., increased food shopping online), which will likely continue to elevate the importance to researchers of nonsurvey data sources such as proprietary data and administrative data; and (ii) assessing the quality, coverage, and representativeness or generalizability of these non-survey data sources, which will be increasingly important.

___________________

4 As articulated by Larimore et al. (2018), this has been a stated goal of ERS for several years.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

Broadly, the challenge is to put each type of data source—surveys, administrative data, and proprietary data—to its best use. Administrative data are best for accurately measuring the use of programs. Survey data can provide rich information on outcomes such as nutrition and health measures while also providing nationally or regionally representative population samples with which to merge the administrative data. Proprietary data are best for high-frequency measures, such as purchases in real time, which would be prohibitively expensive and perhaps infeasible to track with surveys. As discussed in Chapter 2, administrative data can be strengthened, coordinated, and integrated with survey data and put to better use than they now are; similarly, proprietary data could be used more extensively, if made more accessible. Sections 4.24.6 detail our ideas for ways ERS can move forward as it continues the development of its multipronged data system combining surveys, proprietary data, and administrative data. We discuss each of these separately, as well as the importance of integration.

A consumer food data system, such as that maintained by ERS, contains information at the individual, household, and firm level from surveys, administrative data systems, and commercial proprietary data that are representative and accurate at the national, state, and local levels, as demanded by the purposes to which they are put. These data, collected from governmental and nongovernmental agencies and organizations, ideally at regularly scheduled intervals, cover food acquisitions, food security, food prices, food assistance program participation and eligibility, demographics, and health and economic outcomes. Data are needed for monitoring purposes on a regular basis, to allow comparisons over time and to support causal research. Some purposes require data that are repeated cross-sections or longitudinal at the individual, household, or firm level.

4.2. SURVEY COMPONENTS OF THE CFDS

As articulated in Chapter 2, surveys have long been a central data source in consumer food and nutrition research. Survey data provide insight into household- and person-level variables about outcomes that frequently are missing in administrative data. Some surveys have the advantage of linkage between food-related variables and diverse other variables of interest. While surveys are comparatively expensive on a per-observation basis, in the past they have provided researchers with representative samples. Nevertheless, this strength is challenged by increasing difficulties with participation rates, the high respondent burden in some surveys, increased misreporting of important variables such as program participation and income, and lack of timeliness.

Below we touch on the need for some data sources that measure the population at risk of specific outcomes or measure participation in

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

programs, which often come from surveys. In this section, we provide guidance for future investments in survey data and then offer more detailed recommendations for selected important data sources, especially FoodAPS. We also offer recommendations for survey data for monitoring food security, for linkages with nutrition data in the National Health and Nutrition Examination Survey (NHANES), for time use, and for program evaluation. Taken together, the recommendations in this chapter create a vision for survey data that, by comparison with current practice, is somewhat smaller in scope, somewhat higher in cost per observation, more focused on selected applications that cannot be served by other data sources, and more integrated with administrative and commercial data.

General Findings and Recommendations about Surveys

Surveys will continue to be important to statistical agencies for the foreseeable future. They provide household- and individual-level data that cannot always be acquired through other means. Due to increasing concerns with data quality and response rates, survey investments must keep up with current best practices in survey design and implementation (Groves et al., 2009).

RECOMMENDATION 4.1: A key task for the Consumer Food Data System is to assess the quality of survey data across sources and over time. This should be done by linking the surveys to auxiliary sources in order to check sample records. For example, work comparing population totals and individual reports of program participation can be done by comparing survey totals to administrative totals and comparing self-reports to administrative records. The level of missing data and the characteristics of those missing data should be catalogued.

USDA should anticipate in advance that investments satisfying these current best practices will be expensive on a per-observation basis. This implies limits on the total growth of federal investments in traditional stand-alone surveys.

RECOMMENDATION 4.2: To make effective use of limited resources for survey investments, the U.S. Department of Agriculture should further exploit both administrative data sources and commercial data sources for applications wherein they can be effectively used.

For example, whereas survey data sources have in the past been an important source for understanding determinants of program participation and for research on entry and exit dynamics (Mabli and Ohls, 2012),

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

the CFDS should plan for increased use of administrative data and reduced use of survey-only data for these purposes (Ribar and Swann, 2014).

In some cases, the expense of survey data collection may require USDA to focus on a few high-priority research applications, recognizing that other desired research topics cannot be addressed with survey investments that are feasible, given budgetary constraints. Two examples of high-priority topics that will continue to require survey investments are the monitoring of household food security outcomes and measurements of the impact of nutrition assistance programs on food insecurity and dietary intakes.

As discussed in section 4.6, blended approaches, in which survey data are combined with administrative and commercial data, hold great promise for creating added value and lowering costs per observation. This can be achieved through use of blending in frame development, sample unit screening, edits and imputations, augmenting by joining additional content, and modeling (e.g., small area estimation and simulations).

Recommendations for FoodAPS

FoodAPS, which is sponsored by ERS and the Food and Nutrition Services (FNS), is currently the most visible component of the CFDS. As described in detail in Chapter 2, FoodAPS is designed to generate data on household food acquisitions for different populations, particularly low-income households, including food-insecure households and those participating in SNAP and other government programs. By collecting data on all the places where people purchase and acquire food, FoodAPS was an improvement on previous options about acquisition of food for home and away from home.

USDA invested heavily in independent assessments of FoodAPS to better understand data quality. These assessments reflect a strong emphasis on accuracy, one of the key desirable data system characteristics noted earlier in this chapter. USDA also invested in understanding the strengths and weaknesses in FoodAPS from the perspective of researchers and data users (Wilde and Ismail, 2018). Results of these assessments and activities are reviewed in detail in Chapter 2. In a bid to introduce FoodAPS to the research community, and consistent with the desirable data system characteristic of openness, ERS and FNS also underwrote numerous projects by external researchers selected through grant competitions hosted by the National Bureau of Economic Research and the University of Kentucky Center for Poverty Research.

FoodAPS will remain useful for carrying out the descriptive and monitoring functions concerning overall food acquisition. Because the greatest strength of FoodAPS for research and policy is in its capacity to generate descriptive and monitoring information on food acquisition habits, which

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

likely change slowly over time, and because it is an expensive survey, it is not practical to envision it as an annual or even semiannual program. That said, there is clear value to conducting the survey on a regular basis, as doing so allows it to contribute to the construction of stylized facts for the monitoring function of CFDS. There are benefits to using a fixed and predictable schedule (e.g., as the Census Bureau does with the Economic Census). Doing so may generate efficiencies and predictability by creating a regular staffing cycle, which is important for ERS in managing the data system and not having other valuable components of the CFDS suffer when FoodAPS’s resource demands are high.

RECOMMENDATION 4.3: The National Household Food Acquisition and Purchase Survey should be conducted on a regular schedule, such as once every 5 years.

The move to a regular schedule will also allow ERS to plan for the integration of new data sources such as administrative data on multiple programs. This aspect of data coordination should be improved, and likely would be in the presence of fixed periodicity and use of similar data acquisition modules. The ordered planning cycle would facilitate continual process improvement and institutional memory about how a national survey is conducted. This approach would also avoid paying the fixed costs of conducting new surveys at uneven time intervals. At the same time, consistent questions over time also improve the usefulness of these data by, for example, allowing for comparability across assessments of time trends.

To the extent that FoodAPS is intended to support research beyond monitoring of food acquisitions and related outcomes, such as longitudinal and causal research, planners can learn from other surveys that match a sample to longitudinal administrative data both retrospectively and prospectively. For example, the Survey of Income and Program Participation (SIPP) Social Security Administration (SSA) Supplement linked data support studies on program participation and take-up for programs administered by the SSA that are critical for government and academic policy simulation and evaluation. CPS linked to longitudinal Social Security payroll tax records permits analyses of earnings over the life course (inequality, volatility, mobility) that would not be possible with the repeated cross-sections of the CPS alone.

Related, future iterations of FoodAPS could sample from the same geographical units—the same primary sampling units (PSUs)—to create a repeated cross-sectional design. This would permit researchers to combine cross-PSU over time changes in socioeconomic conditions, policy choices, and the built environment to assess how economic, policy, and environ-

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

mental factors affect food acquisitions and related outcomes collected in FoodAPS, which is a method employed in many quasi-experimental research studies.

RECOMMENDATION 4.4: The National Household Food Acquisition and Purchase Survey should be reviewed across a set of design dimensions for future iterations. Along with linkages to extant administrative records from other federal and state statistical agencies, the review should assess the efficacy of sampling from the same set of primary sampling units over time to facilitate more rigorous monitoring and evaluation functions.

FoodAPS has been effective in getting appropriate samples of SNAP recipients because of its use of a dual frame, with one frame composed of SNAP recipients and the other of everyone else. However, it has been expensive to get enough eligible nonparticipants in the sample to make detailed comparisons with participants. It may be more efficient in future rounds of FoodAPS to go even further in the use of administrative and commercial data to create the initial frame, which would cut the cost of screening the non-SNAP participant samples. An example of this approach is the National Survey of Children’s Health done by the Census Bureau. Another example is the Health and Retirement Study funded by the National Institute on Aging but with data collection by the University of Michigan Survey Research Center. Future rounds of FoodAPS could consider these alternative techniques. In addition, due to great interest in oversampling WIC households, program planners should consider including a sufficient sample of WIC recipients (and eligible nonrecipients) using a frame of WIC administrative data.

More broadly, the FoodAPS team can seek and apply best practices in survey design to reduce the burden on respondents and overall costs while improving data quality. Examples include: (1) using adaptive survey design and tailoring the survey operations to optimize participation and using data to monitor when to change course; (2) using auxiliary data in frame development; (3) screening (e.g., generating adequate samples of households with incomes above/below program cutoffs); and (4) mixed-mode designs. The survey design should incorporate greater use of administrative and proprietary data in imputing missing data, adding content depth, and adding longitudinal content.

Broadly speaking, FoodAPS should not be seen as a stand-alone centerpiece of the CFDS, but rather as a key contributor to a system that also incorporates other complementary data sources. Importantly, FoodAPS should not be prioritized over other major initiatives that are funded by ERS or for which ERS plays a supervisory role such as the food security modules in the CPS, NHIS, NHANES, and PSID, the Next Generation

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

Data Platform, person and firm-level scanner data, SNAP Policy Database, and efforts to document strengths and weaknesses of all the data products. Given the response rate and participant burden challenges facing not just FoodAPS, but surveys across the entire statistical system, it is always important to look for opportunities to scale back the length of the survey instruments and simplify the data collection procedures. Indeed, along with increasing accuracy, this has been a major motivation behind ERS’s integration of external data sources for food products linked with Universal Product Code (UPC) codes or retail receipt coding. Statistical agencies today are envisioning a future in which there will be much more blending of mixed data types. As explicitly recommended in section 4.3 below, when a major survey such as FoodAPS is designed, the role of administrative data or other data types in the overall design and estimation strategy should be considered, including the coverage, quality, timeliness, accessibility, and cost of those data. This attention to total error in the mixed data system broadens the total survey approach that ERS already practices in its survey data collection.

Use of Survey Modules

USDA will no doubt continue to collect data using the modules already strategically placed on other surveys (the current use of such modules is documented in Chapter 2). Vehicles such as the Flexible Consumer Behavior Survey and the Eating and Health Module, among others, exploit the strengths of surveys and take advantage of the explanatory covariates contained in other data collections.

RECOMMENDATION 4.5: ERS should advocate for continued funding of data collection, and research on food security should be treated as a high priority in the Current Population Survey, National Health Interview Survey, National Health and Nutrition Examination Survey, and the Panel Study of Income Dynamics.

As discussed in Chapter 2, food security is emphasized in many ERS and FNS-funded modules, in part because the agency is mandated to collect data on food adequacy and has done so on a regular basis for many years.5 The Food Security Supplement to the CPS was prompted by the National

___________________

5 An earlier Committee on National Statistics report (NRC, 2006) shifted the focus of household surveys away from hunger and toward the measurement and monitoring of food insecurity. Hunger, the panel concluded, is “a separate concept from food insecurity . . . [and] an important potential consequence of food insecurity” and it is “an individual and not a household construct.”

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

Nutrition Monitoring and Related Research Act of 1990.6 The full module of the CPS contains 18 items, with both 30-day and 12-month reference periods. The National Health Interview Survey (NHIS) contains a shorter, 10-item set of adult-focused questions pertaining to the prior 30 days (as does FoodAPS). NHANES and the PSID contain the full 18-item module for the prior 12 months. The 18-item module with a 12-month reference period is preferred both because of the importance of monitoring child-specific exposure to food insecurity and because most of the survey questions on program participation, income, consumption, health, and other domains refer to the prior 12 months (or prior calendar year) and beyond.7

RECOMMENDATION 4.6: The Economic Research Service should recommend that the 10-item, 30-day measure currently used in the National Household Food Acquisition and Purchase Survey and the National Health Interview Survey should be replaced in future iterations of these surveys with the 18-item, 12-month module.

Another key set of measures for monitoring the healthfulness of American diets concerns food intake. Currently, 2-day food intake is measured in NHANES. Yet, for many purposes, the sample sizes are too small to allow meaningful policy analysis.8 The most direct way to alleviate this shortcoming would be to financially support the Centers for Disease Control and Prevention, which sponsors the NHANES, to expand the sample size of individuals whose intake is measured on NHANES.

4.3. OPPORTUNITIES FROM AND CHALLENGES WITH EXPANDING USE OF ADMINISTRATIVE DATA

This report calls for a balance of survey and administrative data sources, as well as an integration of commercial data (as discussed in section 4.4 below). It is well recognized that a data system sometimes requires surveys to measure outcome variables, such as food intake, and to achieve representativeness of the entire population (rather than, for example, just program participants). Moreover, as described in Chapter 2, administrative data have both strengths and limitations just as survey data do. Several

___________________

6 See https://fns-prod.azureedge.net/sites/default/files/FSGuide.pdf.

7Schmidt et al. (2016) present evidence of an inconsistency in how the social safety net affects food insecurity, finding a significant attenuation with the 12-month measure and no effect using the 30-day measure. They conjecture that the difference may be due to the differential timing of transfer-program measurement (12 month) and the 30-day measure.

8 For example, one expert panel (NASEM, 2017b) determined that the sample sizes of pregnant women on and not on WIC were so small that the panel felt they did not support robust statistical comparisons.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

institutions have carefully defined ways of assessing the quality of administrative data. Mathematica Policy Research has issued a comprehensive report on data quality standards, summarizing the dimensions that should be assessed.9 Statistics New Zealand has also created a framework for viewing these dimensions that may provide a useful starting point and that may integrate well into federal data strategies.10Harron and colleagues (2017) show multiple ways to evaluate linkage across datasets, which are also important when administrative data are not being evaluated on their own.

Data quality issues aside, statistical agencies have a variety of other reasons for investing more heavily in administrative data sources. Administrative data can be used either on their own or in combination with other data. An example of the former is the use of SNAP administrative records to study how SNAP participation increases or declines in response to policy changes. An example of combining administrative data sources is the linking of the Department of Housing and Urban Development’s (HUD’s) administrative records to SNAP administrative records to estimate the number of households participating in both programs. Administrative data may also be used to enhance the value of survey data or in combination with other administrative data in integrated approaches.

Optimizing the Next Generation Data Platform

A further advantage of administrative data, relative to survey data, is that they exist as a byproduct of routine processes within federal, state, and local governments for such programs as SNAP, WIC, school meals, and others. ERS’s Food Economics Division (FED) has improved its capacity to collaborate across agencies using the Next Generation Data Platform (also discussed in detail in Chapter 2) to link administrative data on food assistance programs, survey data, and administrative data on other programs. Through a partnership with the U.S. Census Bureau and sister USDA agency FNS,11 FED has accessed and analyzed detailed SNAP and WIC data. As of 2017, this partnership included 20 state SNAP agencies (including some counties in California) and 11 state WIC agencies.” ERS relies on the Census Bureau’s infrastructure to negotiate, ingest, harmonize, and link records. The agency’s researchers then access de-identified administrative records that may be linked to survey information (e.g., from the American Community Survey) to assess program eligibility and uptake. The

___________________

9 See https://www.mathematica-mpr.com/our-publications-and-findings/publications/transparency-in-the-reporting-of-quality-for-integrated-data-a-review-of-international-standards.

10 See archive.stats.govt.nz/methods/data.../guide-to-reporting-on-admin-data-quality.aspx.

11 For information on FNS participation and counts, see https://www.usda.gov/media/blog/2018/01/05/collaboration-across-agencies-supports-food-assistance-research.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

success of this partnership relies on the attention and availability of staff, so at times other priorities and projects at the Census Bureau may crowd out this project. ERS should continue its efforts to inventory data available for research use, invest in data documentation, improve data linkage methods, and study the representativeness of Next Generation Data Platform data.

Unfortunately, the usual application process for using the FSRDCs does not give the academic and policy research community easy access to component administrative data and merged administrative and survey data from the Next Generation Data Platform for the SNAP and WIC programs. Existing Census-ERS-FNS data were created with ERS funding, but this was accomplished under the Census Bureau’s Census Act authority, so any project using these data must generate a direct benefit to the Census Bureau. This limitation means that some data projects that would be of value specifically to ERS and FED do not qualify, and outside researchers cannot always access these data or know what is available.

In planning for their specific research, policy, and monitoring needs, investment by FED in the Next Generation Data Platform should take into account the planned implementation of the Foundations for Evidence-based Policymaking Act of 2018 (hereafter the Foundations Act), which will require agencies to identify data that can be used to “facilitate the use of evidence in policymaking” and to create for each agency a chief evaluation officer to coordinate evidence-building activities and a chief data officer to oversee “lifecycle data management.”12 Despite the Foundations Act, it is currently unclear when or how federal resources will support the development of a stable, reliable data sharing infrastructure at the federal or state level. The Foundations Act states that “the head of an agency shall, to the extent practicable, make any data asset maintained by the agency available, upon request, to any statistical agency or unit for purposes of developing evidence.” This raises the question of which data assets are maintained by FED as part of the CFDS and subject to this new law. Section 3564(f) of the Foundations Act notes that nothing in it preempts applicable state laws regarding the confidentiality of data collected by the states. It is expected that the Office of Management and Budget (OMB) and the statistical agencies will gather, interpret, and deconflict laws and regulations related to data access.

RECOMMENDATION 4.7: To aid ERS in expanding the Next Generation Data platform, intergovernmental coordination is needed to maximize the impacts of infrastructure changes made by the Farm Bill (the Agricultural Improvement Act of 2018) and the Foundations for Evidence-Based Policymaking Act. States and localities should share their administrative data, including SNAP and WIC case records,-

___________________

12 This language is contained in the Summary of the Act, which may be found at https://www.congress.gov/bill/115th-congress/house-bill/4174.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

with USDA. USDA should optimize use and access through data intermediaries, including but not limited to the Census Bureau. ERS should develop specifications for their process whereby researchers access administrative and commercial data, and for how researcher-provided data can be brought in and linked to other data.

ERS has authority to request information on programs funded by USDA, though ERS has no compelling legislation that forces state and local agencies to share their data. ERS can encourage these agencies to participate in the Next Generation Data Platform by offering technical assistance for data management and analysis or tools that help agencies improve program monitoring and administration.

The Farm Bill states that the Secretary of Agriculture shall provide guidance and direction for interested states on how states should form longitudinal databases supporting research on participation in and the operation of SNAP, including the duration of participation in the program. The Farm Bill further specifies that the guidance will include standard features for the databases, including database formats, data security, and privacy protections; a directive to establish unique identifiers that provide relevant information on household members receiving benefits; direction on funding the establishment and operation of such databases; and a description of the documentation that research users must provide to gain access to the databases. The law advises USDA to consult with states who have built such databases and with the Census Bureau. Implementation guidelines and technical assistance are needed to help states build databases that are interoperable across state lines as well as with other federal program data. One critical factor necessary for promoting state partnerships is to support the development of data documentation and standard schema across existing state and federal administrative sources to improve data harmonization and interoperability.

As described above, the Foundations Act should make data from other agencies available for federal statistical purposes. This could bring information on workforce, housing, justice, and education issues from administrative data into FED studies of program participation. Along these lines, there are other useful models for merging agency data with surveys as well. One example involves linking data on health outcomes from major agencies and programs—including HUD, the SSA, the National Death Index (NDI), and Health and Human Services’ (HHS’s) Medicare and Medicaid programs—with data from existing surveys conducted by the National Center for Health Statistics. The latter include NHIS, NHANES, the Longitudinal Survey of Aging (LSOA), and others.13 These types of linked

___________________

13 For information about these data linking efforts, see https://www.cdc.gov/nchs/data-linkage/mortality.htm?CDC_AA_refVal=https%3A%2F%2Fwww.cdc.gov%2Fnchs%2Fdata_access%2Fdata_linkage%2Fmortality.htm.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

data have been used, for example, to identify how NHIS parental reports about child asthma correspond to Medicaid data on use of services due to asthma (Zablotsky and Black, 2019). Another example comes from Simon and colleagues (2013). They used linked NHIS-Medicaid data to see what population-level participation by children in Medicaid looks like over a 5-year period, and found that 41 percent of children in the United States were enrolled in Medicaid at some point over 5 years, as compared with 33 percent in a single year.

Numerous administrative sources—such as the Store Tracking and Redemption System (STARS) data on store participation in SNAP and redemptions, and The Integrity Program (TIP) data on WIC store participation, redemptions, and sanctions—are currently reported to FNS but not widely used in research. The same is true of the underlying raw data used to create the SNAP Quality Control datasets and the WIC Participant and Program Characteristics data. It is admirable that some public-use, aggregate versions of these data are available, yet such aggregated data do not contain detail sufficient for some research purposes. Ideally, access to micro versions of WIC Participant and Program Characteristics data, at levels below the state level, would be hosted at ERS or at the FSDRC.

Conclusions about Effective Use of Administrative Data

To fully analyze program participation, eligibility, and take-up through changing social, economic, and policy conditions, administrative data alone are insufficient. Instead, for these purposes administrative data are best used in combination with other kinds of data. Data from surveys and commercial sources can provide more comprehensive information, whether on households or retailers, that can be linked to these administrative data. These sources provide health and nutrition outcome variables that can be used to analyze the effects of participating in programs. They are crucial to analyses of population subgroups, such as veterans, that could not separately be analyzed with administrative data alone. And, by combining administrative data on the use of programs with survey data used to model eligibility, researchers can study take-up patterns and program use among those eligible for the programs. The promise of these types of links can be seen in the two studies using linked NHIS/NHANES-Medicaid data cited above.

For integrating surveys with administrative data, and possibly commercial data, the FED should anticipate data uses in the spirit of small-area estimation. For example, to understand how local labor market conditions affect use of and eligibility for SNAP, researchers need accurate data on participation, income, and other characteristics at the local labor market level to determine eligibility. Similarly, to understand in which communities

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

the SNAP and WIC programs are reaching more eligible people, similar fine geographic data on participation (from administrative sources) and eligibility (derived from survey and possibly administrative data) are needed.

The case for expanded and better coordinated use of administrative data is especially clear for the purposes of program evaluation and improvement, including for SNAP, WIC, school meals, and other program policies. A number of investments will be required to overcome current barriers to the use of administrative data and to make it easier for new states to participate in partnerships with the Census Bureau and for their data to be incorporated.

4.4. OPPORTUNITIES FROM AND CHALLENGES WITH EXPANDING USE OF COMMERCIAL DATA

A forward looking CFDS must anticipate changes in food acquisition among specific groups and be capable of measuring new patterns of food acquisition. This requires thinking about the impacts of emerging food shopping modes such as Amazon and other home food delivery, “grab and go,” and the blurring between store-prepared meals and eating at home. In such an environment, the role of data gathered organically in commercial sectors will become increasingly useful for measurement purposes.

The Changing Nature of Commercially Available Data

The past decade has witnessed the emergence of many innovative techniques and approaches that make use of naturally occurring data to measure population characteristics or predict future behavior. The exponential pace of technology change is having a ripple effect on the availability of data in nearly every sector of research and evaluation, including research in the areas of markets, consumer choices and food security, and health and nutrition. This is important, as even the nature of people’s behavior and attitudes toward food and purchasing has undergone a radical change. For instance, in some markets there has been a move away from an emphasis on packaged foods, microwavable meals, and shopping the “center aisles” toward an embrace of fresh ingredients, deli-prepared foods, healthier alternatives (e.g., shopping the “edges” where the fresh produce is often located), “grab and go” prepped meals, and transparency of ingredients and calories (Fortson, 2018).

These new sources of data—most often available through commercial research organizations—provide information that can help address critical questions in areas such as (i) diets, nutrition, and obesity; (ii) food security and safety nets; (iii) changing consumer preferences in response to price changes, new information, or product attributes; (iv) the food environment,

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

including the availability of stores and restaurants, food prices in an area, and community characteristics; and (v) industry responses and agricultural sector adaptations to these many changes (Larimore et al., 2018).

It is critical, however, that as these new sources of data become available for use, food researchers and evaluators have a deep and nuanced understanding of the nature of these data and the strengths and weaknesses of each source. Despite their attractive qualities, the value of commercial data can be limited by access issues, coverage and representation bias, lack of documentation or transparency of methods, limited scope of variables, and privacy concerns.

As a general class, such commercial data can be thought of as “organic data” as opposed to “design data” (Groves, 2011). Traditional sources of data—from surveys, censuses, and evaluation studies—often involve a significant design element. That is, the researcher is the one who determines the specific population, how that population is (or is not) to be sampled, the data elements to be collected, and how those data elements will be used in the analysis and to draw insights and conclusions. In short, design data are those over which the researcher has much influence and control.

In contrast, organic data arise out of the broader information ecosystem, that is, they emerge from or are used to drive a process. In this respect, such data are not designed for research purposes, but are more the creation of engineers and computer programmers tasked with running a system or platform; examples include retail UPC scanning systems and an online restaurant or home-food-delivery websites.

These data can have great value in the following ways. First, they tend to be massive. For example, IRI InfoScan provides weekly food purchase data from around 48,000 stores, based on more than 6.6 billion observations annually (Levin and Sweitzer, 2018). Second, they provide measurements linked to the events happening in “real time.” Retail scanner data, for example, capture the exact time and date of each scanned transaction. Third, they can provide an unobtrusive (or passive) way of measuring phenomena, without the need to directly engage with the subjects of the research or evaluation. For example, store-level retail scanner data are captured as part of the natural store checkout process. However, because these data are not designed or controlled in any way by the researcher, a number of cautions and data evaluation steps should be considered before these data are used (see below for more details).

As noted in a prior National Academies of Sciences, Engineering, and Medicine report (NASEM, 2017a), organic data can vary greatly in their degree of structure, that is, the degree to which data are in a fixed and readily available format for analysis, as contrasted with those data—termed “unstructured data”—which need to undergo some kind of transformation before they can be analyzed (such as images, videos, social media, or

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

TABLE 4-1 Types of Organic Data by Degree of Structure

Structured Data–Administrative Records Other Structured Data Semi-Structured Data Unstructured Data
Definition Data with a fixed format easily exportable to a dataset for analysis with minimal scrubbing required Highly organized data easily placed in a dataset but requiring additional scrubbing or transformation before analysis Data that may have some structure (but not complete structure) and cannot be placed in a relational database; require substantial scrubbing Data that have no standard analytic structure and must have usable data extracted from them and transformed before use
Examples
  • Government programs
  • Commercial transactions
  • Credit card/bank records
  • University/school records
  • Medical records
  • e-commerce transactions
  • Mobile phone GPS
  • Computer logs
  • Text messages
  • Email
  • Wearable sensor data
  • Internet of Things data
  • Social media content
  • Images/videos
  • Drone data
  • Satellite/radar information

SOURCE: Adapted (and tailored to food security-related items) from NASEM (2017a).

satellite data). Table 4.1 provides an overview of these categories of data, running from the most to the least structured, as they relate to various sources of potential use in understanding food issues, access, and security. These distinctions are important for differentiating the utility of administrative records versus other types of data that may be commercially available.

Types of Commercial Data, and How They Can Be Used by USDA

Commercial organizations provide an array of structured and unstructured data that are especially useful for improving information about the food environment, such as details about stores and how food is laid out within them; ways of acquiring food; information about prices, quantities, and nutritional values; and other characteristics of individual food items (Burke, 2018). As reviewed in Chapter 2 and summarized in Table 4.2, ERS routinely draws from commercial databases in its ongoing research and evaluation work on consumer food, nutrition, and health. Such data have been used to increase the granularity or timeliness of information and to fill data gaps while, in some cases, also reducing costs and respondent burden.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

TABLE 4-2 Commercial Data Sources Routinely Used by ERS

Source Description
IRI Household Item-level grocery purchases
IRI Retail Item-level sales
Nielsen TDLinx Store characteristics and geocoded location
Nielsen Homescan Household scanned price and quantity information for package goods
NPD Recount Restaurant location and characteristics

Sources such as Homescan and retail scanner databases can provide longitudinal measures of consumer choice that, even if imperfect, can be linked to policy or food environment data (Okrent, 2018). These data can provide details about geographic distribution across individual stores or markets, as well as product-level details such as brand, size/weight, and type of package, health and nutrition claims (e.g., gluten-free, type of sugar added, and whether good for reducing risk of heart disease or diabetes). For example, retail scanner data has been linked with nutritional data to create a crosswalk for understanding the relationship between food prices and food purchases (Carlson, 2018). In the near future, such data will also likely provide information about added or supplemental vitamins and minerals, use of non-genetically modified organism (GMO) ingredients or with specialized farming or animal care procedures, and other detailed claims of characteristics about which consumers may care (Burke, 2018). Other potential sources, although not available routinely or systematically, may include data on farmers’ markets, food banks, or school lunch suppliers.

ERS is combining IRI InfoScan sales data with sales data from the 2012 and 2017 Economic Census to augment their analytic capacity in a variety of ways. For some chains, IRI only reports sales at the level of the retailers’ marketing area so as not to reveal individual store sales, which might help competitors. Yet many uses of interest to ERS require sales to be linked to specific locations. The Economic Census data can be used to help impute disaggregated sales for those stores that report at a metropolitan-area level. Also, many retailers do not report private-label sales (for competitive reasons); nor do they always report random-weight items and perishables, which are important metrics for understanding the food environment. By matching InfoScan stores to stores in the Economic Census, the quality of imputed sales for these items can be improved. Of course, imputations should also be assessed for quality.

Firm-originated data can also be leveraged to evaluate information on food away from home in the IRI Consumer Panel. While the detailed Consumer Panel data are a crucial input for many studies of consumer choice and the food environment, they do not have information on food away from home that is not acquired at food stores. Government survey data can

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

be used to impute information about food consumed away from home to be used in tandem with the Consumer Panel and firm data, again with care to assess the quality of the imputations.

The breadth and depth of commercial data available vary greatly across geographic areas, populations, and programs. Even so, the geospatial aspect of these data can be used, perhaps in tandem with other data, to enable fine geographic assessment of poverty, hunger, and food accessibility (Allard, 2017) and to assess how the food environment can affect consumer choice (Ver Ploeg, Larimore, and Wilde, 2017). For example, NDP Recount provides a near census of data on the locations of food-service operators (both commercial and noncommercial), which can be used to draw insights into (i) restaurant density within a particular area; (ii) the penetration of different types of restaurants within a community; and (iii) comparisons of restaurant revenues across geographies (Hanson and Lesce, 2018). These could also be linked with the other data described above to get a more complete picture of the food environment, to help understand issues related to population changes in an area, or to help understand the impact of the age or racial composition of local neighborhoods. Important features generated from combining the IRI Consumer Panel with data on firms include metrics on the locations of stores used and the distance to the nearest store (accessibility), the assortment of foods sold, the costs required to get to a store, and so on (Bonanno, 2018).

The above examples are suggestive of the potential of commercial data to add new dimensions to ERS’s CFDS. It is important that the agency continue its work in this area.

RECOMMENDATION 4.8: The U.S. Department of Agriculture (USDA) should exploit new ideas for integrating commercial data into the Consumer Food Data System. For example, to produce a long “time series” of data on Supplemental Nutrition Assistance Program (SNAP) participation, food insecurity status, and the location of all stores in the immediate environment of the respondent, USDA could facilitate matching restricted-access Food Security Supplement data (with respondents’ locations) with TDLinx data on stores, state data on SNAP and other program participation, and Store Tracking and Redemption System data on stores that redeem SNAP.

As commercial data sources are increasingly used to assess food-related issues, a number of priorities for advancing their use have emerged. These include the following:

  • Documenting and improving the overall representativeness of retail data.
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
  • Developing weights for the retail stores to make them representative of the geographical areas covered.
  • Imputing prices for random weight purchases in the household data.
  • Merging prices of products not chosen by consumers as outside options when formally modeling consumer demand systems and choices.
  • Imputing prices and/or sales for individual stores and private labels where they have been suppressed and documenting their suppression.
  • Linking stores listed in the household Consumer Panel with data generated by those establishments.
  • Acquiring new data from vendors, if feasible, on SNAP and WIC variables that are less restricted in use than existing consumer and firm household data.14
  • Extracting from commercial data sources a variable on the payment method to infer usage of cash, credit, coupons, and SNAP or WIC benefits.

CFDS could productively undertake these project ideas.

Assessing the Quality of Commercial Data

The promise of data beyond surveys and structured administrative data is large, but the benefits are only just beginning to be realized. As with survey and administrative data, commercial data must be evaluated for quality. Like survey and administrative data, these data often suffer from a variety of issues that affect a researcher’s ability to truly understand their nature, representativeness, and quality.15 Also lacking is a set of agreed-upon techniques for assessing the validity, reliability, and robustness of the inferences made from such data. Assessing the quality of data requires a level of transparency. For example, CFDS stewards (and researchers) need to be able to deal with changing platforms among proprietary providers

___________________

14 Current use of the retail data from Nielsen is available for a 5-year window via the Kilts Center for Marketing at the University of Chicago. Data use agreements for these data prevent authors from determining detailed geography, limiting the usefulness of these data. IRI data access via licenses from ERS is limited to users approved by ERS; more open access to these data would increase knowledge. While FNS has data on SNAP redemptions and individual states have data on WIC redemptions of food instruments (in paper-voucher states) and electronic data (in EBT states), these data are used for research only lightly if at all. And these data on redemptions do not inform researchers about what else consumers buy when they are using SNAP or WIC or what they buy on trips when they do not use the vouchers.

15 See NASEM (2017a, Ch. 4) for a comprehensive discussion of the use of private-sector data for federal statistics.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

and to document changes in internal algorithms, an issue especially relevant to proprietary datasets.

RECOMMENDATION 4.9: As with survey and administrative data, commercial data in the Consumer Food Data System should be continually reviewed for accuracy. Data checking, including comparing proprietary commercial data with other sources, such as the Census of Retail Trade, is an essential part of data acquisition, data processing, and vetting. It is important to document coverage of these auxiliary data in terms of geography, the distribution of retail outlets across types, and the amount of purchases captured. It is also important to construct weights to make the population of participants demographically representative of the national population.16

Widespread use of commercial data as a replacement for well-designed, representative surveys and more robust and accessible administrative data is still some distance in the future. Nonetheless, ERS has an admirable tradition of using commercial data while also comparing findings, totals, and coverage with other sources.

Given the need to ensure data quality and transparency for research or evaluation purposes, a framework is needed for evaluating potential sources of bias and error. Having such a framework would allow a more systematic and standardized way of assessing datasets before use.

One such framework builds on the concept of Total Survey Error (TSE), which parses potential sources of bias and error broadly into sampling and nonsampling errors (Biemer, 2010). The TSE framework attempts to break down the potential sources of error and variance originating from (a) the sampling process, including errors that occur because only a sample of the population of interest, rather than all of it, is surveyed, and (b) nonsampling components of the survey process, such as frame construction, data collection, data processing and estimation approaches (Biemer, 2010).

A similar yet more expansive framework is needed to assess newer types of nonsurvey data. The outlines of such an approach, which builds on and expands the TSE framework and may be thought of as the Total Data Error (TDE) approach, were described by Japec and colleagues (2015). The TDE framework includes the more traditional sampling error assessments but expands the sources of nonsampling error to include measures of error capturing how commercial or organic data are generated, extracted, transformed, loaded, and ultimately analyzed. The approach attempts to account for a variety of potential errors, such as observation-level errors of omission

___________________

16 Some sources, such as the IRI Consumer Panel, include weights that are provided to ERS as part of the data purchase. Other sources, such as InfoScan data, do not come with weights.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

(when relevant cases are excluded due to data selection procedures—similar to survey noncoverage); duplication (depending on how data are captured there may be multiple observations for a single individual or firm); or cases may be included that do not actually match the population of interest).

The potential sources of error extend far beyond those normally encountered in more design-based surveys. As such, it is critical that a framework be developed to help evaluate these potential sources of error in current and future commercial data sources used in food research. Commercial or organic data may have problems related to concept error (when data provided do not actually measure the concept the researcher thinks is being covered) or variability across datasets in the way similar concepts are measured or stored. Likewise, throughout the collection and transformation process, these data may be subject to errors related to how data are extracted, transformed for analyses (particularly in the case of unstructured data), imputed, or analyzed. A TDE framework will account for these newer sources of error to allow for easier and more meaningful assessment of the quality of the data and insights generated from it.

RECOMMENDATION 4.10: The Economic Research Service should develop and use a Total Data Error Framework—which includes the assessment of traditional sampling error and expands on the traditional sources of nonsampling error—to aid in evaluations of the quality and utility of existing and future potential data sources, ranging from commercial or other “organic data” sources to data from surveys and administrative sources. This framework should consider aspects of data origin, generation, extraction, transformation, loading, and analysis in addition to the preceding recommendations for assessing data quality. Standards should be identified and adhered to for gauging the quality of stand-alone data and linkages and to assess privacy risks associated with all components of the Consumer Food Data System.

Data Use and Access

Commercial data are purchased by statistical agencies with the intention that they can be effectively used in a strategy that improves the accuracy or breadth of information, reduces costs or survey burden, or both.

RECOMMENDATION 4.11: The Economic Research Service should continue to invest in efforts to overcome barriers to the use of proprietary data. One element of the strategy should be to negotiate an improvement in terms for Nielsen TDLinx data-sharing agreements to increase the ability to link these data at fine geographic levels and across sources.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

Specific examples of the use of such proprietary data include linking the Consumer Panel to the TDLinx data and linking data from either of these sources to detailed Census data on local characteristics or to data on local policies. ERS should act to ensure that other proprietary data also have use terms similarly specified.

Giving researchers access to data is also essential in order to generate value from the investment in it. One challenge with using proprietary commercial data is that access to them is limited. For example, non-ERS-affiliated users can obtain access to retail data from Nielsen through the Kilts Center for Marketing at the University of Chicago’s Booth School of Business, but users of these data through Kilts face limits on determining detailed geography. IRI data access via licenses from ERS is limited to ERS-approved users. More open access to these data would increase knowledge.

RECOMMENDATION 4.12: The commercial data in the Consumer Food Data System (CFDS) should also be made more accessible to outside researchers and the policy community while preserving privacy. The U.S. Department of Agriculture should ensure that qualified researchers have access to proprietary data from Nielsen, TDLinx, and other commercial providers in CFDS. Legal barriers—such as indemnity clauses that prevent access to researchers, especially those employed by land-grant and other state-assisted institutions, which are forbidden by state law from entering such agreements—should be eliminated from current and future contracts; or, alternatively, means of data access should be explored while maintaining data privacy and security.

4.5. CREATING COMPREHENSIVE POLICY DATABASES

Policy evaluation should be an important consideration in data collection design. There are important policy questions at each level: program questions at the state level, agency questions at the system level, and outcomes questions at the client level. Often, data belong to the states, and they and local governments administer the policies governing the data’s use, but the federal government is paying for some or all of it. Nevertheless, comprehensive databases tracking important policy choices do not exist, and FED faces restrictions on how much and what they ask states, retailers, clinics, and other entities to limit the burden they put on the public.

The SNAP Policy Database and SNAP Distribution Database are model resources for the handling of administrative policy data, allowing research to be carried out about how program choices made by different governmental entities affect outcomes in their localities. These two databases have led

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

to the publication of papers (e.g., Kuhn, 2018; Heflin et al., 2019; Beatty et al., 2019; Ganong and Liebman, 2018) studying the effects of program participation on participant outcomes and on the SNAP cycle.

RECOMMENDATION 4.13: The Supplemental Nutrition Assistance Program (SNAP) Policy Database and the SNAP Distribution Database should be updated annually by the Economic Research Service’s (ERS’s) Food and Economics Division. Similar cross-state over-time policy databases on additional food assistance programs, such as Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), the School Breakfast Program, the National School Lunch Program, and the Child and Adult Care Food Program should be established and updated annually by ERS. Data that measure rules affecting participating retailers (e.g., stocking requirements) and other entities (e.g., reimbursed foods in school meals programs) should also be collected and made available. Data should be made available about the geographic location of benefit offices (e.g., the city, county, state, latitude, and longitude of locations where participants apply and recertify for assistance, including schools, SNAP offices, and WIC clinics). Finally, administrative data on store participation in SNAP (through the Store Tracking and Redemption System) and WIC (through The Integrity Profile) should be made available with geographic locations for participating retailers; the possibility of making redemption data available should also be explored.

Ideally, data would be included on cash purchases and SNAP or WIC redemptions for the same individuals and sales and redemptions at the same stores so complete acquisitions could be studied.

There have been some successful linkages across survey, administrative, and program-rule databases that have enhanced knowledge. For example, the Census Bureau’s American Community Survey has been combined with administrative SNAP records and the SNAP rules database. That linkage is at the household level, which has allowed researchers to answer questions such as, what is the impact of SNAP policy changes on program participation and employment outcomes? This has been successful, but at present easy access is limited to those with internal Census projects, and access only applies to data for the states that participate in the Next Generation Data Platform.

4.6. COMBINING DATA SOURCES AND DATA ACCESS

We conclude with some overarching guidance that applies to more than one aspect of the CFDS as ERS continues to enhance data products through more expansive contracts with proprietary data, states and localities, and

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

links to other federal administrative data. Most critical to this process is continued assessment of the quality of all data, whether survey, commercial, or administrative. Estimates based on different sources should be compared with one another where possible, and the quality of any linkages should be assessed.

Standards are available for gauging the quality of stand-alone data and of linkages and for assessing the privacy risks associated with all components of the CFDS. Some of the sources of these standards operate within the statistical agencies, such as the Federal Committee on Statistical Methodology.17 Others—such as the Inter-university Consortium for Political and Social Research (ICPSR), a group of more than 750 academic institutions and research organizations that “provides leadership and training in data access, curation, and methods of analysis for the social science research community”18—reside outside of government. Elsewhere, a group of researchers at University College London provides guidance for information about linking datasets;19 and the Harvard Privacy Tools Project seeks to advance “a multidisciplinary understanding of data privacy issues and build computational, statistical, legal, and policy tools to help address these issues in a variety of contexts.”20

The quality of data can only be thoroughly assessed through regular use of the data by researchers.

RECOMMENDATION 4.14: The Economic Research Service’s (ERS’s) Food Economics Division should create a process for hosting restricted-use data through a secure platform, such as the Federal Statistical Research Data Centers network. Data for publicly funded programs should be made available for research at granular levels, including individual-level de-identified and linkable data, while still addressing privacy concerns. This should include information generated in activities funded or sponsored by ERS and the Food Nutrition Service, including the food assistance programs and other programs whose output is included in the Consumer Food Data System.

The FSRDC network has a well-established set of enclaves hosting sensitive data. However, costs and conditions for hosting data in the FSRDC are not transparent. Timelines and processes for getting multi-agency projects approved also lack transparency and stability over time, resulting in fragile arrangements often reliant on an agency champion or insistent

___________________

17 See https://nces.ed.gov/fcsm.

18 See https://www.icpsr.umich.edu/icpsrweb.

19 See https://academic.oup.com/jpubhealth/article/40/1/191/3091693.

20 See https://privacytools.seas.harvard.edu.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

investigator. ERS FED can work on improving conditions in the FSRDC network. FSRDC processes are evolving to streamline requests across many sources of government data (driven in part by the Evidence Act), building upon long-standing processes that supported Census and health agency data access. For example, the FSRDC network has a demonstration project testing secure remote access for approved researchers, potentially aligning the network with other providers who already offer remote research access (e.g., NORC Data Enclave, New York University’s Administrative Data Research Facility).

The following issues have been observed in earlier data-sharing efforts and will need to be addressed by ERS FED to increase data access.

  1. A broad framework should be created specifying who should get access, including all qualified researchers suitably defined, rather than making data available only to specific subsets of the research community (such as cooperative researchers only, only those with USDA funding, or only for Intergovernmental Personnel Assignment Act reassignments or Schedule A Federal Employees). This will require considering issues like institutional attachment, citizenship, and vetting/background checks. It may also need to vary by researchers’ attachment to the government.
  2. A broad array of data should be made available at disaggregated but de-identified levels that also contain clear and complete metadata.
  3. Costs should be able to be covered by funding from USDA or from external sources (e.g., from reputable federal, state, or nonprofit sources).
  4. Automated data provisioning should be used to minimize delays and errors caused by people manually moving files to secure workspaces.
  5. Application processes for use of any non-Census Act data at the FSRDCs (via ResearchDataGov) or at an ERS-hosted location should be clearly laid out, following the models of other agencies, such as the Bureau of Labor Statistics, National Center for Health Statistics, and Agency for Healthcare Research and Quality.
  6. Best practices for using data collected for other purposes should be delineated.
  7. Careful consideration should be given concerning who is allowed to execute linkages of external data. Data-linkage protocols, including use of trusted third parties, should be established.
  8. State administrative data used by ERS researchers should follow output review protocols that ensure adherence to project scope and sensitivity while maintaining academic freedom and access to publicly funded data.
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
  1. Proprietary data, such as ERS versions of Nielsen and IRI data, should be housed in such a setting for joint academic research. This may require innovation in the FSRDC network or ERS FED can pursue secure hosting for such data by other service providers.

To ensure that no proprietary, confidential data are being released, FoodAPS requires content and disclosure reviews for projects. Decisions about disclosure should be based on protecting respondents, not based on reviewing the research message—that is, they should be strictly about privacy and confidentiality.

As described throughout this report, analytic capacity can be greatly enhanced when data are combined across a wide range of sources to enable both monitoring and causal research—including scanner data on people and store sales and prices, UPC-level nutrient and product characteristics, food environment data, population-representative survey data, and administrative data on program participation. Taking advantage of multiple data sources requires that the ERS FED partner with other agencies to leverage strengths. For example, ERS may decide it is cost-effective to leverage Census survey methodology expertise for some data projects. In other cases, the agency should take advantage of interagency work on developing standards to assess survey, administrative, and proprietary data.

RECOMMENDATION 4.15: The Economic Research Service’s (ERS’s) Food Economics Division should create a data council to prioritize which data should be created and specify access rules while ensuring that the Consumer Food Data System addresses ongoing U.S. Department of Agriculture research data needs. This council should also help create and update a longer-term data-infrastructure plan. This plan should balance two goals. Access should be as wide as possible to facilitate policy making, scientific advances, education and training, and public understanding about society. Yet, at the same time, data stewards are ethically and legally obligated to protect privacy and sensitive attributes. ERS should seek input from the American Statistical Association, the federal statistical system, and the broader data and research community on how to prevent re-identification, protect sensitive attributes, and increase access. This data council could also be tasked with setting and reviewing the rules for access to ERS and/or Federal Statistical Research Data Centers, described above. This approach could follow the model of the Department of Health and Human Services’ data council, and it should include nongovernment stakeholders.

Finally, the CFDS will need to remain open to future changes in how people in the United States acquire and prepare food, what they eat,

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×

and how it affects health. This will require paying attention to changing demographics—for example, an aging society may require more congregate meals. Moreover, climate change may lead to changes in how food is produced and what it costs, and changes in technology are sure to affect what food is available.

We have made a series of recommendations that span the current and past CFDS and also make suggestions for the future. Listed here are those that we regard as the highest priority relative to the rest (but not in priority order): (1) recommendations related to checking data and linkage quality, (2) recommendations to enhance more access to existing data and future data by outside researchers as well as through existing relationships with more geography, (3) recommendations laying out strategies to include more administrative data into the CFDS, (4) recommendations that the CFDS systematically focus on serving monitoring needs (e.g., measuring food security consistently) and causal research needs through longitudinal designs, and (5) recommendations to create policy data bases to enhance causal research.

Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 105
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 106
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 107
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 108
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 109
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 110
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 111
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 112
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 113
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 114
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 115
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 116
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 117
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 118
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 119
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 120
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 121
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 122
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 123
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 124
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 125
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 126
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 127
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 128
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 129
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 130
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 131
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 132
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 133
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 134
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 135
Suggested Citation:"4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System." National Academies of Sciences, Engineering, and Medicine. 2020. A Consumer Food Data System for 2030 and Beyond. Washington, DC: The National Academies Press. doi: 10.17226/25657.
×
Page 136
Next: References »
A Consumer Food Data System for 2030 and Beyond Get This Book
×
Buy Paperback | $70.00 Buy Ebook | $54.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Patterns of food consumption and nutritional intake strongly affect the population's health and well-being. The Food Economics Division of USDA's Economic Research Service (ERS) engages in research and data collection to inform policy making related to the leading federal nutrition assistance programs managed by USDA's Food and Nutrition Service. The ERS uses the Consumer Food Data System to understand why people choose foods, how food assistance programs affect these choices, and the health impacts of those choices.

At the request of ERS, A Consumer Food Data System for 2030 and Beyond provides a blueprint for ERS's Food Economics Division for its data strategy over the next decade. This report explores the quality of data collected, the data collection process, and the kinds of data that may be most valuable to researchers, policy makers, and program administrators going forward. The recommendations of A Consumer Food Data System for 2030 and Beyond will guide ERS to provide and sustain a multisource, interconnected, reliable data system.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!