National Academies Press: OpenBook

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

Chapter: 3 Data and Knowledge Gaps

« Previous: 2 ERS's Current Consumer Food and Nutrition Data Infrastructure
Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

3

Data and Knowledge Gaps

The Consumer Food Data System (CFDS), produced by the Food and Economics Division of the Economic Research Service (ERS), can provide data, fill information holes, and facilitate science in three areas. First, as a statistical agency, ERS can use the CFDS to carry out a series of monitoring tasks that allow the public, policy makers, and researchers to track outcomes across time. Second, CFDS can be used to assess the quality and coverage properties of various types of data, including data collected by other agencies. Assessing the quality of data used within ERS’s Food Economics Division, in the statistical agencies more broadly, and by outside researchers in turn improves the quality of research.

The third area where the CFDS can contribute is in the creation of data and the carrying out of research by ERS and other USDA staff as well as outside researchers. Specifically, the CFDS can enable statistical approaches in the study of both descriptive and causal questions, particularly as they pertain to programs designed to improve the well-being of persons across the United States. Descriptive evidence establishes facts and terms of debates and provides hypotheses for further research. Causal designs provide evidence of the outcomes made possible by USDA programs. Of particular value is the way the CFDS can strengthen researchers’ ability to conduct evaluations of both current food assistance programs and potential future interventions.

In this chapter, each of these three functional areas is described. While, Chapter 4 explicitly presents strategies for improving the CFDS, this chapter describes in more general terms data areas warranting further attention by ERS and the statistical system more broadly.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

3.1. MONITORING NEEDS

Producing data that allow the monitoring of food and nutrition outcomes, together with related health outcomes, and using these and other data to create snapshots of the population’s nutritional and other food-related health, are key components of the CFDS. These snapshots also help stakeholders to understand trends in these areas over time. In addition, when monitoring data are available at more granular levels, state and local information—such as that complied for the Food Access Atlas, discussed in Chapter 2—can be created.

Food Security

The first subject area for which the CFDS provides crucial data related to monitoring population outcomes, particularly through the December Current Population Survey (DCPS), is the measurement of food security. The DCPS data collection instrument enables annual snapshots of food security by state and by various demographic and economic measures over a long time period, but there are holes in its existing measurement of food security. One drawback is that the DCPS is only conducted during 1 month of the year, December. That choice does have the advantage of providing consistent estimates across years, helping to mitigate the wide oscillations that occur in the Current Population Survey (CPS) depending on which month is used to measure food insecurity. But in light of these oscillations, in order to monitor seasonal patterns, it would be extremely useful to collect the Core Food Security Module during other months of the DCPS, at least occasionally. A particularly important gap is our understanding of the experiences of households with children during the summer months, when school meal programs are not available.

Additional important data are provided by including various measures of food security on a range of other surveys beyond the DCPS. One of the most useful is the National Health and Nutrition Examination Survey (NHANES), which also generates objective measures of health (participating individuals have their health measured objectively by professionals). This allows researchers to see how food security varies with objective and subjective measures of health. This feature of the survey is important, given that some individuals might not know about health conditions that they have if they do not see a medical professional frequently. However, NHANES uses only small samples, limiting the ability of researchers to correlate these health measures with food security to help understand its ramifications. Moreover, in part due to these small sample sizes, NHANES is only nationally representative in demographic terms and lacks the detailed geographic information needed to measure at the state and substate level. In practice, data are collected for a limited number of counties each year.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

This lack of state-level representativeness means that different years of NHANES may include people from entirely different state and local policy settings, complicating standard two-way fixed-effects model designs,1 which require repeated measures from many of the same locations over time. NHANES surveys persons from a small number of counties each year, so it would not allow for such fine granularity unless its sample sizes were increased. These small samples across locations limit the array of analyses. National Center for Health Statistics (NCHS) is responsible for NHANES and deciding sample sizes and questions; however, in the past, ERS has sponsored modules that expanded samples.

Similarly, there are disadvantages, especially for child-focus analyses and replicability with other datasets, to having the National Health Interview Survey (NHIS) not include all 18 questions but only the 10 food security items asked of adult respondents in the NHIS Food Security Module at the annual level (rather than 30 day). Were the NHIS instead to have 18 rather than 10 items, it would greatly enhance the extensive health information collected in the NHIS. It would also allow for research on the effects of long-term health problems and disability on food insecurity or the effects of food insecurity on more short-term health outcomes be advanced.

The Panel Survey of Income Dynamics (PSID) also monitors food security. The only nationally representative longitudinal dataset that uses the full 18-question food security instrument, it includes detailed measures of individuals’ income, employment, health, wealth, consumption, food expenditures, and family structure in a panel fashion. While the CPS data allow for longitudinal analysis of repeated cross-sections, the PSID measures the same people over time, allowing for more precise measure of changes in outcomes than repeated cross-sections (e.g., Duncan and Kalton, 1987). If the PSID discontinued questions about food insecurity, it would leave gaps in the data sources used for tracking it.

___________________

1 Such regression models are used for estimating causal effects from panel data. The simplest case of such a two-way fixed-effects model compares outcomes for two populations and two time periods. In this simplest model (also known as differences in differences), the goal is to understand the effects of a policy change or treatment that takes please in the second time period for one of the groups (the treated group). The simplest approach would be to take the difference in the treated group before and after the policy change (a difference). But the concern with doing only that is that other shocks occurring at the same time as the policy change would confound this simple difference. If the control group also faces the same shocks, however, then the change in the control group serves as a counterfactual for what would have happened in the treated group in the absence of the policy change, and the resulting difference in difference provides a causal estimate of the effects of the policy. This can be generalized to a setting with many periods and groups, where the group-fixed effects control for time-invariant factors in each group and the time-fixed effects control for period-specific shocks that all the groups face, yielding the terminology “two-way fixed effects” models.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Program Rules about Food Assistance and Other USDA Activities

Another missing link in the monitoring area is the absence of a comprehensive set of measures of rules affecting participants, firms, and nonprofit providers, such as schools and clinics, in the food assistance network in the United States, including those that participate in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). ERS has collected data on both eligibility policy and disbursement policy for the Supplemental Nutrition Assistance Program (SNAP), which have led to a host of papers studying the effects of these policies on health, education, food security, and economic outcomes. But these rules are not always updated annually, and the data only cover SNAP to date and leave out the other important child nutrition programs. An important missing link is an accounting of the state- and county-level choices about rules for the participation of individuals in these other programs, such as WIC, the school meal programs, the Child and Adult Care Food Program (CACFP), summer feeding programs, and other USDA programs at the relevant geographic and temporal level for modelling eligibility. This would enable researchers within and outside of USDA to measure how changes in rules affect use of programs; it would also enable them to model disaggregated measures of program eligibility. ERS is also limited in its ability to go to the state, county, and local entities administering these programs due to limits on burden, although the USDA Food and Nutrition Service (FNS) knows more about these programs in many cases. It would also help research to be able to document where local, state, or federal eligibility decisions are made, for example by listing the locations of SNAP offices.

In addition, some of these food assistance programs affect a set of nonprofit organizations that administer the programs, such as schools (school meals programs), CACFP facilities, and WIC clinics. It is also important to understand how these rules affect the choices made by these nonprofits. One example would be the Community Eligibility Program (CEP), which enables schools with sufficiently high levels of students eligible for free and reduced-price meals to choose to offer free meals to all students with a federal subsidy. As part of this program, states are required to make the lists of schools that qualify for this new option publicly available. Currently, these lists of CEP-eligible schools are available from the nonprofit Food Research and Action Center,2 having them collected by and made available through ERS would be a more natural choice. Another example of the kind of data that it would seem natural for FED to collect is the timing and choices for meal pattern requirements for CACFP or school meals to be reimbursed for meals/snacks provided, disaggregated by detailed geography and time.

___________________

2 See https://www.frac.org/cep-map/cep-map.html.

Suggested Citation:"3 Data and Knowledge Gaps." 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 know how these nutrition policies affect food choices, one must have information about where and when they apply, including when they are enacted and actually implemented.

Food assistance programs, mainly SNAP and WIC, also affect businesses such as convenience stores, grocery stores, and larger chains. Firms choose to participate in these programs by providing food in return for payments from customers using Electronic Benefits Transfer (EBT) card payments or, in the case of WIC, paper vouchers, for which the firms are then reimbursed. Here, there is a wide set of questions that would be useful to address. These questions range from knowing whether the state requires WIC-authorized stores to participate in SNAP, to what options the state has selected for the WIC cash-value vouchers for fruits and vegetables, to when the state implemented other rules such as for EBT use and what form EBT takes for SNAP and WIC. Information about these rules, and how states interpret the rules, helps researchers understand varying eligibility status across states as well as the administrative burden of program participation (for firms, nonprofits, and potential participants). Having an accurate accounting of the firm rules is also important for studying how stores choose to participate in or leave specific programs. In addition, knowing which firms are potentially eligible to participate allows researchers to see where there are gaps in access to consumers and how these gaps may affect the prices faced by participating individuals. Finally, as EBT is implemented for WIC, states have more information about what individuals purchase using WIC vouchers, and it should already be possible to see what individuals purchase with SNAP. This could fill gaps in our knowledge of what food assistance programs facilitate.

Program Participation and Eligibility, and Locations Where Eligibility Is Determined

To understand program use, it is important to have geographic and time measures with fine detail on program participation and eligibility. USDA provides researchers, policy makers, and the public with snapshots of outcomes such as participation among individuals eligible for SNAP, available at the state level and nationally by group. Data on participation in WIC (program characteristics data) and demographic and state measures of take-up for WIC (reports) are also available. Yet more fine-grained counts of participation refined by demographic and geographic detail and, if possible, by eligibility would add value by pinpointing where targeting has been useful and where it is failing. Of course, measuring eligibility among nonparticipants is challenging. Some of the components that would enable this to be done for participation already exist but are not easily accessible or are only available by restricted geographic aggregations (e.g., the WIC

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Participant and Program Characteristics). Other measures of eligibility are produced at the state level by contractors for FNS.

It is important to know where the various programs are available. This would include having access to a list of stores that historically have participated in WIC or SNAP without requiring a Freedom of Information Act (FOIA) request. It would also include having access to information about stores that have been sanctioned from the program as well as lists of schools participating in the School Breakfast and National School Lunch programs, of child and adult care facilities that use CACFP, and of locations participating in the summer feeding program. Some of these lists are currently available but only as snapshots, whereas causal research requires having historical data too. These measures of access to free and reduced-price food allow researchers and stakeholders to see where benefits are available and how that geography compares to where eligible participants and nonparticipants live. Access to data on stores on and off the programs allows researchers to study how firms participate and also how sanctions affect firms and participants.

In addition, a list of charitable food agencies would be useful for the purpose of understanding the larger food environment. The food banks under the Feeding America network (and the agencies under those food banks) constitute the overwhelming majority of charitable food assistance in the United States. Feeding America has information on these food banks and their locations, as well as about their hours of service, number of people served, and so on. A data-sharing partnership with this network would therefore help achieve this data collection objective.

It is also important to know where the agents who assess eligibility for food assistance programs are located. In the case of schools, this is included in one of the above lists. But SNAP eligibility is assessed at county SNAP offices, and WIC eligibility is assessed at WIC clinics. These data should also be aggregated to allow research monitoring access and also on how access affects use of programs and ultimately, participant outcomes. Combined with this information, knowing where eligible and participating people are located would also be useful. Ideally, information would be detailed enough to ascertain who is eligible for SNAP, WIC, and other programs at a fine geographic level, such as by county, as well as who is participating in these programs and how this is affected by the food and program eligibility determination environment.

Finally, how these programs interact with one another is often not well understood. For example, joint participation by individuals and households in SNAP and some other non-USDA program can be assessed using quality control data, and similar joint participation in WIC and some other programs can be viewed in the WIC PC data. However, outside of survey data, which struggle to count each program accurately, and these eligibility determination administrative data, joint participation is hard to measure.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Measuring it better would allow improved research on a number of policy issues ranging from eligibility determinations to program effects on child outcomes and the inability to do so represents a current gap.

A gap in the data infrastructure on program participation is the lack of guidance for compiling a list of and then updating policies over time. Ideally, there would be a standardized scheme for including policies, a method for geocoding all locations, and a method for managing timestamps and providing version control. Sound data management would improve efficient processing and dissemination for researchers and policy makers.

The Food Environment

Another valuable innovation from USDA is the Food Access Research Atlas, one of the agency’s most widely used collections of data provided to the public.3 This atlas reports at two points in time where the prevalence of retail food outlets is available at the census tract level. Having repeated cross-sectional measures of the types of food available by store is important, as it allows researchers to study access.

Currently, data from IRI and Nielsen do a good job of obtaining this information. The challenge in accessing this data, however, is three-fold. First, purchasing the data from IRI or Nielsen can be prohibitively expensive for many researchers, especially those at less wealthy institutions, limiting the number of researchers with access and, in turn, the set of questions that can be posed. Second, while the sampling frame for the IRI and Nielsen data does include some smaller stores, these data are not released to outside researchers because they are highly proprietary.4 Even though overall coverage may be good, especially for non-low-income households, some stores, especially those that may disproportionately be located in low-income neighborhoods, may be overlooked. Perhaps methods could be used to incentivize IRI and Nielsen to include these stores in their sampling frames. The most valuable way to do that might be to lower the sales volume needed to be included in the sample, an approach that would not necessarily require including specific stores. Another part of the problem is that stores enter and exit the sample, and smaller stores are most likely to experience this. Greater retention of small stores in the sample would therefore be helpful. A third issue is that IRI requires researchers to sign indemnity clauses to use its data, when linked to FoodAPS, and Nielsen has

___________________

3 For a detailed description of the Atlas, see Chapter 2 and a summary of the presentation to the panel by Michele Ver Ploeg in Appendix A.

4 In general, coverage in rural areas is limited. For example, for counties with 20,000 or fewer residents, Nielsen uses an average of the surrounding counties. This is a “rural issue” insofar as these counties have fewer stores, but it is not, strictly speaking, due to the sampling methods used by Nielsen.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

similar requirements for TDLinx. Most state universities will not sign such clauses as they violate state law, restricting who can work on these topics.

Measures such as those indicating the presence of certain types of stores by detailed geography also might serve as indicators of the set of firms “at risk” for participating in SNAP or WIC.5 It is also important that researchers, policy makers, and other stakeholders can identify where stores and other entities that participate in the program are geographically located so that the information can be combined with data about the location and stores at risk of participating (see above point). Combined data would allow researchers to investigate whether there are places where food programs are not present but where people and stores are. It would also let researchers study how the food environment is affected when stores are temporarily or permanently kicked off WIC or SNAP.

Similarly, data such as which schools are participating in universal meals or school breakfasts also should be available and could conceivably be added to such data dissemination vehicles as the National Center for Education Statistics Common Core database or the Food Security Supplement of the Current Population Survey. Going forward, the food environment is sure to change as families increasingly use delivery services such as Amazon, Grub Hub, or Blue Apron. It is also useful to know where various providers of free food (pantries) are located as well as where restaurants (by level of healthfulness) are available at fine geography in order to study consumer choices.

Another component of the food environment is the broader set of expenditures facing vulnerable households. Consider the problem of food insecurity. Over one-half of poor households are food secure, while approximately 15 percent of nonpoor households are food insecure. This is due to many factors that are often observed in datasets, but the influence of other expenditures is often not observed on the same datasets that record food insecurity. Of particular note might be the expenditures households make on housing, which vary dramatically across the United States. Another factor that varies is transportation costs. Information on both of these could be included in the geographic landscape dataset mentioned above.

Consumption

It is also important to understand how restaurants, stores, and the rest of the food environment affect consumer demand. This includes understanding demand for product attributes, such as whether foods are locally

___________________

5 Given that nearly all stores receive SNAP, this is more of a WIC issue than a SNAP issue. Although, it is true that a small share of stores are removed from each program when they violate rules.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

produced, natural, or organic. Other examples of relevant questions include these: How do consumers value and use side-of-the-package information, such as nutrition facts? How do they use front-of-the-package health claims (e.g., “heart healthy”)? How do they value production characteristics, such as whether foods are organic or hormone-free? And, How do they value other characteristics (e.g., “natural”)? Understanding these questions requires data about the characteristics or attributes of food—such as that which has been collected in IRI data and, prior to that, in the Gladstone data—as well as the ingredients. This is discussed further in Chapter 4. This would also encompass questions such as understanding demand for novel products, such as plant-based meat alternatives, and how this affects the food system.

Prices

Food choices are affected by availability, income, and the characteristics of the food environment—each discussed above. Another element affecting food choice is, of course, pricing. The Bureau of Labor Statistics (BLS) collects excellent data on prices for urban areas but has limited coverage of rural areas. Knowing the prices for foods at refined geographic levels is important for understanding the value of food assistance benefits and for monitoring the performance of programs that administer therm. For example, breaking down purchased foods into “healthful” and “unhealthful” categories can be useful.

For 34 geographic regions, price indices (levels and changes) were provided by USDA’s Quarterly Food at Home Data. However, these data undoubtedly miss within-market variation in prices; moreover, this information has not been updated since 2012 and lacks detail about variety concerning what food types appear within each category. Datasets derived from scanners report weekly prices, but some level of detail is suppressed for many stores or only averages are provided. So, while these scanner data are far more disaggregated than most of the data obtainable from existing government sources, and while they represent the average shopping experience, they do not necessarily contain individual variation in prices paid. Data on individual prices paid at specific stores, net of taxes and other features like coupons, would let researchers know where food prices are high, allow them to measure the pass-through of factors affecting prices to consumers, and let them measure the incidence of the food assistance programs. Additionally, when collecting prices and building price indices, it is important that particular attention be paid to items that will improve the reference diets.

There is one example of price data that are already available that may be useful for program monitoring. As part of Feeding America’s Map the

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Meal Gap,6 which maps food insecurity rates by county and congressional district for each state, Nielsen data are used by USDA to compile the price of the Thrifty Food Plan for all counties in the United States.7 The Thrifty Food Plan is designed by USDA to specify foods and amounts of foods that provide adequate nutrition in a way that provides the basis for determining the monetary value of SNAP benefits.8 While data on the Thrifty Food Plan do not provide the level of disaggregation needed for some analyses, they can be employed in other cases, such as SNAP analyses. The value of these data would be even greater if they could be linked to the December CPS (or NHANES or NHIS) within the Research Data System. One additional benefit of working with Nielsen to ensure that more stores are covered (in the manner described above) is the potential to improve measurements of the Thrifty Food Plan.

Time Costs

Food choices are also affected by time costs (both distance and travel time to food acquisition venues) and the time it takes to prepare food. Time cost is particularly important given that food assistance program benefits are in-kind and, in the case of SNAP, they can only include food items that are not already prepared. BLS’s American Time Use Survey (ATUS) provides a useful snapshot of time use, and the eating and health module to ATUS also captures eating when it is a secondary activity (e.g., while watching TV) but includes no other useful features. Secondary eating (eating while doing other activities) is understudied in time use data.

FoodAPS documented that the usual store for a family’s food purchases is only 3.8 miles away from their home and that the vast bulk of individuals use a car to shop even when they do not own one (Ver Ploeg et al., 2015). Yet little is known about the time costs of travel to stores, which includes both driving time and waiting time. The tradeoffs between time and money in food acquisition would be useful to know. In the same way, it is also hard to model the administrative burden of programs on recipients, firms, or lower levels of government or nonprofits administering programs without knowing the administrative time costs of their participation.

___________________

6 See https://map.feedingamerica.org.

7 Among small counties, some do not have any stores, and even among those that do none of the stores is included in the Nielsen dataset, so all counties with fewer than 20,000 residents are represented by a weighted average of the county itself combined with its surrounding counties.

8 See https://www.fns.usda.gov/cnpp/usda-food-plans-cost-food-reports.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Food-Related Health Measurement

Finally, there are various other food-related health measures beyond food insecurity that are crucial for modeling the effects of programs, the food environment, and prices on food choice, consumption, and even outcomes like obesity. These include data on food consumption and the consumption of micro and macro nutrients, which can be obtained from food recall studies. Objective measures of some of the same details can be drawn from biospecimens, including blood and urine, which can be obtained in person along with other health measures including the presence of chronic and acute conditions. NHANES conducts in-person physical examinations, and thus it captures many of these objective biomarkers (using blood and urine draws and a measure of BMI), which offer important insights over and above what one can learn from self-reported measures such as food security and dietary intake.

NHANES’s method for obtaining these data also has the advantage over self-reported surveys of providing objective measures of the presence of some conditions that the underinsured and those who rarely visit medical facilities might not know they have. Yet as discussed above, and as is the case with many other datasets, NHANES samples are too small for accurately measuring outcomes by demographic categories (e.g., pregnant women, infants and children, prime-age adults, and the elderly) or by socioeconomic status, and the lack of accurate and well-measured program participation rates limits the ability to compare these health outcomes across programs. There are also limitations due to the lack of state representativeness.

It would be useful if more health-related measures were available that specified program participation and income. This shortcoming stems from the fact that, in most datasets about program participation, all of the characteristics of the respondents are self-reported, so that data on program participation are not accurate. It would be extremely helpful if these health data were linked to administrative program data, as is done in the Census Bureau’s Next Generation Data Program and by NCHS, for which Medicaid, Medicare, and Department of Housing and Urban Development (HUD) data are linked to NHANES and NHIS. Broader geographic coverage of food acquisition, nutrition, and food patterns is also important for monitoring.

3.2. ASSESSING THE QUALITY AND COVERAGE OF DATA

In addition to monitoring and surveillance, there are important aspects of the statistical agency role that ERS fills and that crucially need to be continued or expanded. USDA has done an excellent job of examining the quality of proprietary commercial data and some administrative data (Muth, 2018).

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

This needs to be continued and expanded. It is also crucial to understand who is left out of each form of data, whether that be survey data, proprietary commercial data, or administrative data. Those left out may be firms, such as small ethnic stores or A50 stores. Often outside the scope of surveys or else poorly measured are certain categories of persons, particularly the homeless, certain military personnel, and those living in group quarters. For a different reason, unauthorized immigrants are often left out, because they are less likely to respond to surveys (Capps, Gelatt, and Fix, 2018).

Once data are combined, it is even more important to evaluate the quality of the data integration.9 This is a major focus of the Next Generation Data Platform (described in Chapter 2, Box 2.2), a cooperative effort between ERS, FNS, and the Census Bureau that has enabled the linking of administrative and survey data to improve models of SNAP eligibility and participation. Whenever time data are merged across sources, it opens the possibility of mismatches or missing matches. Given the reliance on data produced by state and local governments and commercial entities, it is essential to have a process for continually assessing and improving data quality.

In particular, the proprietary sources that FED uses are collected for a variety of clients, and FED is unlikely to be able to dictate items or terms. State and local data, meanwhile, are created for the purposes of running programs and not for ease of integration. For example, a state may have one system for determining WIC eligibility and another unrelated system for reimbursing stores for WIC vouchers. FoodAPS combined several administrative sources of SNAP data, including data from redemptions (ALERT) and from caseload data, but these measures were almost as discordant as the self-reported data from SNAP receipts (Courtemanche, Denteh, and Tchernis, 2019). Even administrative data can have weaknesses stemming from being linked to surveys when it is hard to create individual-level de-identified and linkable Protected Identification Key (PIK) data, such as for new infants who are not yet in the tax data, for highly mobile populations, or when it is hard to assign the administrative records to geographies, as was documented in a 2010 Census planning memo (Rastogi et al, 2010).

3.3. A DATA INFRASTRUCTURE FOR ADDRESSING DESCRIPTIVE AND CAUSAL QUESTIONS

To fulfill its mission, ERS’s CFDS must make it possible to answer both descriptive and causal questions. To answer descriptive questions, researchers need access to data to measure the geography of deprivation and nutrition across time, for example, as well as other data to identify correlations

___________________

9 See the discussion of this issue by Robert Moffitt in Appendix B, under the summary of the second meeting.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

among demographics, program use, and income. To answer causal questions with observational data, researchers need access to data tracking changes in diet preferences and nutritional knowledge, as well as data on households’ ability to use information to make healthful dietary choices.10 Below, we identify each of the priority areas laid out above and present important questions of each type that remain unanswered but would, ideally, be answerable within the next decade or so with the aid of CFDS.

Food Security

There are multiple research questions regarding food insecurity in the United States. However, to fully address these questions, gaps in currently available information must be filled through upgrades to data sources. Additional questions could be added to surveys, sample sizes could be expanded, and/or often overlooked groups could be better incorporated into data collections with food insecurity components being used by ERS and other researchers and agencies. For example, as covered in Gundersen and Ziliak (2018), there are a host of descriptive research questions, and some causal ones, that deserve investigation. Descriptive questions include these: How is food insecurity distributed within the household? What types of coping mechanisms do low-income but food-secure families use, and what are the effects of those mechanisms? Questions of a more causal nature include

___________________

10 The use of observational data to support causal inferences in the social sciences is not universally accepted as being consistent with best current statistical principles and practices. Some argue that causal inferences are best supported through the use of randomized controlled trials and that observational studies can be misleading and are generally not reliable for this purpose. There is no question that the use of randomized controlled trials has greatly improved our understanding of the efficacy of ‘treatments’ versus ‘controls’ in many areas of science and that the causal inferences produced through use of randomized controlled trials are often well-supported. However, it is also the case that randomized controlled trials themselves are sometimes flawed for making causal inferences. For example, Deaton and Cartwright, (2017) point out that “randomization does not equalize everything but the treatment across treatments and controls,” and it does not relieve the need by researchers to think about observed or unobserved confounders. Furthermore, perhaps particularly in the social sciences, there are situations where, due to various constraints, randomization of treatment to subject is not feasible (and, in some cases, not ethnical). But the presence of these constraints is not reason to rule out research where causal analyses are nonetheless needed to support policy. Fortunately, as is discussed by Hernán et al. (2008) and others, a number of techniques exist, including various uses of propensity scores, instrumental variables, panel data, differences in differences, and reweighting, that can be used to create treatment effect estimates from observational data that match the results from clinical trials. The techniques used for supporting causal inference in observational studies do require assumptions regarding the adequacy of the set of measured baseline confounders and usefulness of any control groups, but many (although not all) of these assumptions are testable in most observational settings. However, in large studies, the assumptions are often quite reasonable and, in addition, tools for assessing the sensitivity of inferences to the assumptions are available.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

these: What are the effects of charitable food assistance programs on food insecurity? What is the causal relationship between food insecurity and health outcomes? How does disability status influence food insecurity? Why is there a declining age gradient in the probability of food insecurity among seniors? How does labor force participation affect food insecurity? What is the impact of changes in the minimum wage on food insecurity? And, what is the impact of the Affordable Care Act on food insecurity?

Regarding the more causal research questions, knowing more than we presently do about patterns of food security over time (within the year) for stable geographies would allow for causal estimates of the effects of school meals. Similarly, being able to make comparisons between food insecurity during the school year and during the summer, a season when program availability is more limited, would be indicative of the role played by policies that remediate food insecurity.

Program Rules about Food Assistance and Other USDA Activities

USDA and the federal government in general have important roles in running food assistance programs. At the same time, state and local governments, which share responsibility with USDA for administering food assistance programs, have many options for deciding what rules individuals, firms, and nonprofits must follow. These options range from deciding which forms of fruits and vegetables—fresh, frozen, or canned—stores must stock in their WIC programs, to choosing the date on which SNAP and WIC benefits are disbursed, to deciding the extent to which applications for SNAP can be made online and where applicants must go to determine their eligibility, to requiring that all schools with a free and reduced lunch certification amount above a certain level participate in the School Breakfast Program. Despite this wide authority and range of options, these state- and locally decided rules are rarely tracked.

As recommended in Chapter 4, creating a database for each of the other programs analogous to the databases currently tracking SNAP options for eligibility and disbursal would advance research estimating associations and causal effects, on a number of topics that now happen only slowly and haphazardly. Research would also be empowered by the tracking of rules for firms and the geography of those assessing eligibility and providing benefits. Constructing these databases would enable the study of program use and eligibility by persons, of program participation by firms, and the extent to which the locations of government entities determining eligibility affect take-up by individuals.

Tracking the rules described in the previous section is the first step in calculating whether individuals are potentially eligible to participate in the food assistance programs. Eligibility will vary for specific members

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

of the population depending on income and assets and, at times, other factors. High-quality panel data on people and firms are also needed to be able to use the eligibility rules described above to predict whether households, individuals, or firms are potentially eligible to participate in the programs.

Moreover, it is also necessary to maintain high-quality administrative data on where program participants live and where they spend their benefits to understand how programs are used and who among those eligible takes up programs. Incomplete take-up is thought to be a product of stigma, transactions costs, lack of information, and the cognitive burden of poverty. Eligibility and participation data would allow for comprehensive examinations of the targeting of these programs and help promote better understanding of issues pertaining to take-up. All of this is important for understanding whether programs are effectively on the margin targeting the most needy, as behavioral science would suggest be done, or whether they are on the margin reaching the least needy among participants. Recent research suggests this is not uniform across programs and eligibility and outreach efforts (e.g., Deshpande and Li, 2017; Finkelstein and Notowidigdo, 2019).

Ideally, assistance programs should reach those most in need; in other words, they should be “well-targeted.” When this is the case, households that are less-in-need (in terms of the goals of the program) would be less likely to participate, including households with incomes near the eligibility threshold and/or incomes that are likely to exceed the threshold in the near future. In addition, for households with characteristics reflecting higher asset levels (including human capital), the benefits they receive from food assistance may not exceed the total costs when one considers stigma and transaction costs.

A concern emerges, though, when households with higher levels of need do not enter a program. In many cases, this may happen because the costs to enter the program may be perceived to be higher than the benefits or involve some other dimension than can be addressed by policy changes. For example, many vulnerable households may not have information about the program or, even if they are aware of it, may find that the administrative hurdles to entering the program are too high. By reducing these costs—such as the cost of obtaining information or the costs of the application and recertification process—programs may be able to increase participation rates.11

In the case of SNAP, this difference between types of nonreceipt is evident when one looks at the over-age-60 group in comparison with the

___________________

11 For analyses of SNAP churn, see articles by Ribar and Edelhoch (2008) and Mills et al. (2014).

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

age-40-to-60 group. For the former, the low participation rates can largely be explained by observed characteristics while, for the latter, it is not clear why the participation rates are low (Gundersen and Ziliak, 2008). The geographic and time detail discussed above can provide insights into the reasons for nonparticipation and, in particular, whether there are ways to increase participation among the more vulnerable.

While some of these administrative data, made available with limited geographic detail (such as the state level for the WIC PC data by FNS) are already used by individual researchers or by the Census Bureau and others, there is no comprehensive set of such data spanning all states and programs. Further, simply tracking the location of WIC clinics, schools participating in school meals, SNAP offices, and other program delivery and eligibility determination sites would facilitate important research on program take-up, especially for the smaller programs that are not well studied, such as CACFP. Knowing where potentially eligible stores are located and which stores participate would allow researchers to model store choices regarding program participation and track where programs are hard to access. Knowing the geography of nonprofit school, clinic, and CACFP participation choices will also be useful.

Maintaining accessible administrative data on the programs from all 50 states and the District of Columbia would also improve statistics on take-up and poverty, as long as there is also a suitable source of data on the full population to determine who is eligible for the programs. These data could be used to augment Census or other measures of self-reported program participation if linkages from administrative sources to possible sources of the full population are robust. While a large number of states currently share at least one source of state-run program data with the Census, a much smaller group shares WIC, TANF, and SNAP information. To the extent that these states are not randomly sharing their data but may have different populations, evidence generated from analysis using these states’ data may not generalize to the whole nation. Expansion of the ERS/Census Bureau’s Next Generation Data Platform would enable researchers to study the interactions between USDA programs and other programs, if the data were made widely available to outside researchers.

The Food Environment

It would be useful to have descriptive information to track changes in the retail food environment—for example, the rise of dollar stores, the locations of ethnic food markets, the growth of delivery services through venues such as Amazon, Costco, Walmart, and meal-kits. Little is known, even descriptively, about the many small nonchain stores lacking point-of-service technology where low-income individuals and families often shop and

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

redeem food assistance vouchers.12Cuffey and Beatty (2019) suggest that about 11 percent of SNAP redemptions in Minneapolis occur at such stores. Tying these data to data covering other sources of food, such as restaurants (with associated information about the dietary quality of offerings), schools, day care facilities, and providers of free or subsidized meals would help complete the picture of the food landscape. Combining existing scanner data with data on stores participating in SNAP (using the Store Tracking and Redemption System or STARS) and WIC (TIPS13) would help researchers to track smaller stores and would augment existing sources such as TD LINX.14

Next, we turn to some causal questions, which place even greater demands on the underlying data needed to answer them. Repeated cross-sections of food environment data are needed to evaluate geographically targeted policies, such as those addressing poor food environments by proposing taxes on sugar-sweetened beverages, or the Healthy Food Financing Initiative, which sought to improve access to healthy foods by helping to cover some of the costs of setting up grocery stores. It is not easy for many researchers to access to data with detailed geography on food intake or food acquisition in order to study how such local policies might affect outcomes. Other data sources, such as scanner data may be expensive, lack coverage, or are unable to link to fine geographies. Government survey alternatives, such as NHANES, are constrained by limited geographic coverage.

In addition to the food environment, information on other components of the geography facing low-income households would be relevant. As an example, high housing prices are often a constraint on the ability of households to be food secure. By overlaying housing prices onto the information noted above, this could be investigated. As another example, in some parts of the country there have been increases (or proposed increases) in the minimum wage. The impact of these changes on food insecurity and other food outcomes are ambiguous and, therefore, including this in a comprehensive overview of the food environment could be useful.

___________________

12 FNS recently published results of a survey of small SNAP retail stores about their ability to adopt scanning technology if needed in the future. See https://fns-prod.azureedge.net/sites/default/files/resource-files/SNAPScanner-Capability.pdf.

13 TIP Data Collection is intended to provide FNS and WIC state agencies with “an annual dataset that can be used to assess State agencies’ compliance with WIC vendor management requirements and estimate State agencies’ progress in eliminating fraud, waste, and abuse.” See https://www.federalregister.gov/documents/2018/03/21/2018-05704/agency-information-collection-activities-proposed-collection-comment-request-special-supplemental.

14 As described in Chapter 2, STARS from FNS and TDLinx from Nielsen have been used to assess characteristics of the food retail environment, such as the locations and characteristics of food retailers and restaurants.

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

Consumption

Descriptive data enable researchers to track the evolution of new product attributes and novel food characteristics. For causal research, however, more is needed. For example, it would be helpful to be able to track changes in diet preferences and in people’s knowledge of and ability to use information to make healthful dietary choices. Yet there are holes in our knowledge of food consumption preferences and the choices of some urban residents and very-low-income households, because these populations are insufficiently represented in Nielsen and IRI data sources and because errors are made in recording in general (e.g., Einav, Leibtag, and Nevo, 2010).

Other questions that could be addressed with more data include whether new products arise because of changes in preferences or, alternatively, because of technological change. Data tracking of net product characteristics would also enable more focused analyses of the role of information provision from the federal government, private sources (advertising), and other public sources in changing food demand. This is particularly important for understanding possible market failures caused by imperfect information that is linked to food choices. These market failures arise due to changes in nutrition science and our understanding of what a healthful diet is, because food is a “credence good,” that is, a good whose quality consumers cannot assess until after they have consumed it, and also because the provision of information provision can shift the salience of attributes of food, such as “healthful” or “organic.”

Prices

Descriptively, detailed price data, as discussed above, would permit analysis of the relative affordability of different kinds of foods, healthful and otherwise, across time and over space. Detailed price data would also allow researchers to study causal questions, such as how differences in the real value of SNAP and WIC benefits affect food acquisition and consumption and subsequent health and other outcomes. While there has been some research on this topic (e.g., Gregory and Coleman-Jensen, 2013; Courtemanche, Denteh, and Tchernis, 2019; Bronchetti, Christensen, and Hoynes, 2019), it has used data aggregated at perhaps too high a level, such as the county or regional level, and it has depended on too limited a set of price indexes. By incorporating the relative prices of non-store sources of food helps researchers to better understand the impacts of food prices.

Time Use

Insofar as people often eat while engaged in other tasks, it can be difficult to track eating as a “primary task.” Monitoring eating as a secondary activity is important and has been enabled by the Eating and Health Module of the ATUS. Continuing to track these important time-use patterns

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

requires an ongoing commitment to this sort of data collection. It also would be useful to know more about how people trade off their time with other resources across various income flows coming into people’s homes. This could be done, for example, when considering where and how people shop and the extent to which they purchase near-ready foods, such as frozen meals, versus raw ingredients. Lastly, the various food plans make assumptions about peoples’ ability and willingness to prepare food from scratch. It would be useful to consider time concerns when creating the food plans.

Food-Related Health Measurement

Larger samples of these health outcomes that span more detailed geographic areas would permit researchers to better study the effects of policy and other determinants of food choice on health and nutrition outcomes, beyond the most basic outcome of food security. They would also permit better monitoring of health and nutrition and program use. Combining health data with nationally representative data that accurately measure use of programs and income flows would allow the study of questions such as how policy affects programs, food choice, and health.

We also note that many useful projects have been conducted with data collected by other agencies. Better coordination with other agencies might help avoid such problems as the failure of important surveys to collect or merge administrative data on programs, as happened with the Early Childhood Longitudinal Study, Kindergarten Class of 2010–2011, where receipt of school meals was not reported at the individual level. One example of such a useful link is provided with NHANES’s and NHIS’s links to Medicaid, Medicare, and other administrative data.

Although extensive work has been done on the impact of food insecurity on current health status (for a review, see Gundersen and Ziliak, 2018), the longer-term impacts are still an open question. This is primarily due to not having a consistent set of food insecurity questions on any panel dataset. PSID included questions from 1997 to 2003, but due to lack of funding those questions were removed until 2015. The continued inclusion of these questions on PSID would be welcome, as it would enable an understanding of issues such as how food insecurity in childhood is transmitted into long-term health and human capital outcomes as adults and whether food insecurity is transmitted across generations.

3.4. CONCLUSION

In this chapter, strengths and some gaps in the statistical system’s coverage of consumer food and nutrition choices and associated outcomes have been laid out. In some cases, identified in Chapter 4, these are domains

Suggested Citation:"3 Data and Knowledge Gaps." 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.
×

where the CFDS does and should provide data to enable the study of descriptive and causal questions. We have also discussed the important role of the FED in serving ERS’s role as a statistical agency. We have included the specific questions we think are required to provide evidence for policy makers and the public alike so that they have the information necessary to make decisions that will make the country a better place in 2050.

But new issues are sure to arise. We urge the Food Economics Division to keep the following issues in mind going forward, toward a time when, should current demographic trends continue, the country is sure to be more racially and ethnically diverse. The country is also likely to have more mixed family structures, including more cohabitation, single-parent households, and multi-generational households, varying concentrations of poverty, and mixed immigration status. For example, shared custody may have implications for defining an economic unit of people who eat together. Separately from demographic changes, changing food technology and preferences may alter the shelf stability of many foods, with implications for the types of storage needed to store foods. Tastes are set by early exposure to specific kinds of foods, and programs can affect this.

All of these evolving issues require a forward-looking mindset and a cohesive agenda in data collection. In the next chapter, we lay out a series of recommendations to advance the CFDS in a way that would fill a number of the data and knowledge gaps identified here.

Suggested Citation:"3 Data and Knowledge Gaps." 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 85
Suggested Citation:"3 Data and Knowledge Gaps." 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 86
Suggested Citation:"3 Data and Knowledge Gaps." 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 87
Suggested Citation:"3 Data and Knowledge Gaps." 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 88
Suggested Citation:"3 Data and Knowledge Gaps." 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 89
Suggested Citation:"3 Data and Knowledge Gaps." 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 90
Suggested Citation:"3 Data and Knowledge Gaps." 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 91
Suggested Citation:"3 Data and Knowledge Gaps." 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 92
Suggested Citation:"3 Data and Knowledge Gaps." 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 93
Suggested Citation:"3 Data and Knowledge Gaps." 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 94
Suggested Citation:"3 Data and Knowledge Gaps." 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 95
Suggested Citation:"3 Data and Knowledge Gaps." 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 96
Suggested Citation:"3 Data and Knowledge Gaps." 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 97
Suggested Citation:"3 Data and Knowledge Gaps." 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 98
Suggested Citation:"3 Data and Knowledge Gaps." 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 99
Suggested Citation:"3 Data and Knowledge Gaps." 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 100
Suggested Citation:"3 Data and Knowledge Gaps." 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 101
Suggested Citation:"3 Data and Knowledge Gaps." 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 102
Suggested Citation:"3 Data and Knowledge Gaps." 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 103
Suggested Citation:"3 Data and Knowledge Gaps." 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 104
Next: 4 Strategies to Strengthen the Infrastructure of a Consumer Food Data System »
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!