Survey Vehicles to Measure Food Insecurity and Hunger
The U.S. Department of Agriculture (USDA) bases its annual report and estimates of the prevalence of food insecurity on data collected from the Food Security Supplement (FSS) to the Current Population Survey (CPS). The Household Food Security Survey Module (HFSSM) within the FSS, or a modification of it, is or has been used in several surveys. One of the main objectives of the annual food insecurity measure is to monitor the estimated prevalence of food insecurity and changes in the prevalence over time at the national and state levels for purposes of program policies, as well as the need for program development.
Attaching the FSS to the CPS for the official estimates has several advantages. The CPS is the largest of the major national sample surveys and comes closest to fulfilling one of the key goals for the project—national and state-level monitoring of food security. Data from the FSS in the CPS are released on a timely basis. There are, however, reasons to consider other surveys as a possible primary vehicle or to complement the FSS food insecurity data for research purposes. This chapter briefly reviews the key features of selected national surveys and assesses the relative advantages and disadvantages of each survey reviewed.
KEY FEATURES OF SELECTED SURVEYS1
Current Population Survey
The design of the CPS is described in Chapter 4 of this report. To summarize, the CPS, sponsored jointly by the Bureau of Labor Statistics and the Census Bureau, is a national household sample survey. Like most major national surveys, the CPS covers the civilian noninstitutionalized household population, including noninstitutionalized groups not living in conventional housing units and groups living in housing containing nine or more persons unrelated to the person in charge. Its design is a nationally and state-level representative sample, but estimates for small states have large confidence intervals. It has a very large sample size of 60,000 households (about 119,000 individual respondents in households, with about 31,000 under age 18) interviewed every month. In addition to its core content, a different supplement is fielded each month. The CPS has high response rates, and adding the FSS to it is relatively inexpensive and allows for timely production of annual prevalence data on food insecurity. Data are available from the Census Bureau within 3–4 months and released by USDA in published form to the public about a year after collection. The cost of the supplement to USDA is about $450,000 each year. This includes data collection, initial editing for confidentiality and weighting, incorporating household-level food security variables calculated from the initial data by USDA, preparation of documentation for the public use file, and purveying the public use data on CD-ROM and on the Census Bureau’s Data FERRET system.2
As noted earlier, CPS uses a rotating sample design and technically is a panel survey because a household is in the survey for four months, then out for eight months, and then back in for four months. The sample unit is the household address and not the household, and for the reinterview period the original occupants of the address may have moved since the earlier interviews. One person is interviewed to obtain information for the entire household. The CPS is in the field every month, but only the Annual Social and Economic Supplement to the CPS in March and some rotation groups in February and April include detailed information on household income. The FSS is administered in December, at which time very little income information is collected. The CPS does not collect information on general health of the population or nutritional intake.
National Health Interview Survey
The National Health Interview Survey (NHIS), conducted by the National Center for Health Statistics (NCHS), is a multipurpose health survey and the principal source of basic information on the health of the civilian noninstitutionalized household population of the United States. It is a cross-sectional, nationally representative sample with reliable estimates for the four defined geographic regions of the United States.
The NHIS consists of a core or basic module as well as variable supplements. The basic module consists of three components: the family core, the sample adult core, and the sample child core. The family core component collects information on everyone in the family, including household composition and sociodemographic characteristics; tracking information; information that matches administrative databases and basic indicators of health status; activity limitations, injuries, health insurance coverage, and access to, and use of, health care services. For each family in the NHIS, one sample adult and one sample child (if any children under age 18 are present) are randomly selected and information is collected on their health status, health care services, and behavior with the specific core questionnaires.
For the family core questionnaire of the basic module, all members of the household age 18 and over who are at home at the time of the interview are invited to participate and to respond for themselves. For children and adults not at home during the interview, information is provided by a knowledgeable adult residing in the household. For the sample adult questionnaire, one adult per family is randomly selected; this individual responds for himself or herself, unless physically or mentally unable to do so. Information for the sample child questionnaire is obtained from a knowledgeable adult residing in the household. Topical modules are fielded less frequently.
About 40,000 households are interviewed annually, including approximately 112,000 people of whom 29,000 are under age 18. The interviewed sample for 2004 consisted of 36,579 households, which yielded 94,000 persons in 37,466 families. The interviewed sample for the sample child component, by proxy response from a knowledgeable adult in the family, was 12,424 children under age 18. Interviewing is conducted continually, so that seasonality can be studied from any systematic month-to-month variation over the year and the effects of seasonality can be averaged out.
NHIS is not currently designed to provide state-level estimates. However, black and Hispanic populations are oversampled now to allow more precise estimation of health in these growing minority populations.
A possible problem is that the NHIS is faced with periodic budgetary shortfalls, as occurred in 2002, 2003, and 2004. As a result, NCHS reduced the size of the NHIS sample in these years. Since grouping the sample cuts
into consecutive weeks yields the greatest monetary savings, the NHIS sample was cut in three consecutive weeks in May 2004. For 2003 and 2002, the NHIS sample was cut in three consecutive weeks in April 2003 and two consecutive weeks in December 2002. Thus no new interviews were conducted during 3 of the normal 50 weeks of interviewing.
The data collected in the NHIS contain extensive measures of health and well-being of the respondents. Supplementary modules on different health-related topics are added by sponsoring agencies, and they vary from year to year. Some recent and planned examples are:
2000: Cancer risk factors (National Cancer Institute); also 1987 and 1992
2001: Healthy People 2010; child mental health (National Institute of Mental Health)
2002: Healthy People 2010; complementary and alternative medicine (National Center for Complementary and Alternative Medicine); child mental health (National Institute of Mental Health)
2003: Healthy People 2010; child mental health (National Institute of Mental Health)
2004: Child mental health (National Institute of Mental Health); cell phone questions
2005: Cancer risk factors (National Cancer Institute); child mental health (National Institute of Mental Health); cell phone questions
2006: Diabetes (National Institute of Diabetes and Digestive and Kidney Diseases); child mental health (National Institute of Mental Health)
2007: Complementary and alternative medicine (National Center for Complementary and Alternative Medicine)
NHIS also includes questions about participation in the most common assistance programs. Although it asks about household income, the income module is brief and does not ask about food expenditures.
Data from the supplements are released in the same files as the core data and are released at the same time as data from the basic module. The timing of data release has improved noticeably in the past few years, and NCHS is working to improve it further. Quarterly estimates for 15 key health indicators are released on its website about 6 months after data collection. In general, other data are released in published form 14–18 months after the end of the data collection year.
National Health and Nutrition Examination Survey
The National Health and Nutrition Examination Survey (NHANES), conducted by NCHS, collects detailed anthropometrical, medical, and nutritional information on all sample persons. These data are valuable for understanding the links between food insecurity and health and between food insecurity and diet. Food security and sufficiency have been measured in NHANES since NHANES III. From 1999 to 2001 the food sufficiency question was expanded to include food adequacy. The question was dropped beginning in 2002. NHANES has included the 18-item household food security survey module in the family interview part of the household interview since 1999. Individually referenced food security questions (7 for participants age 16 and older, 6 for participants younger than age 16) were added to the postdietary recall component of the examination section and were released with the 2001–2002 food security data release. NHANES continues to collect both the household and individual-level food security data in collaboration with USDA.
Beginning in 2000, NHANES included questions about the individual-level hunger of participants age 16 and over and of a proxy regarding children under age 12. Beginning in 2005, NHANES will ask 12–16-year-olds these questions. These questions ask about the individual’s experience, in contrast to the household-level questions, which just ask about anyone in the household. These individual-level responses can then be assessed in relation to individual measures from other examination components.
NHANES collects an appreciable amount of dietary information. It uses a dietary recall method to collect information on the complete dietary intake for two nonconsecutive 24-hour periods, and supplemental questions are asked about the intake of infrequently consumed food for the past 30 days. In addition, it collects self-reports about health conditions and includes a detailed medical examination including blood analysis.
Similar to NHIS, NHANES also includes questions about participation in the most common assistance programs. Although it asks about household income, the income module is brief. The sampling for NHANES is complicated. Although it selects individuals in the same household at a higher probability than individuals in different households, individuals in a household are still subject to selection in order for NHANES to achieve its complex oversampling scheme. It uses a household respondent for various household characteristics but then interviews, and gives a medical examination to, each sampled person.
NHANES was initially conducted on a periodic basis, but since 1999 it is conducted on a continuing basis. The NHANES samples about 5,000 households annually, so it is not large enough for annual estimates or for subgroup or state analysis. It groups two years of data to develop national
estimates, and these two-year blocks of data are available for research purposes.
Survey of Income and Program Participation
The Survey of Income and Program Participation (SIPP), conducted by the Census Bureau, is a nationally representative household sample survey of the noninstitutional population. The sample is drawn to be representative at the state level, but estimates for most states have large confidence intervals. The survey design is a continuous series of national panels, with sample size ranging from about 14,000 to 37,000 interviewed households. Although the sampling unit is the household, SIPP attempts to interview all household members age 15 and older by self-response. SIPP is designed as a panel survey of all individuals in a household, and therefore movers are followed. A new SIPP panel was fielded every three to four years over the past decade, and each panel is reinterviewed every four months. The last complete SIPP panel began in 2001 and interviewed nearly 37,000 households in a panel design of 9 waves spaced 4 months apart. The current SIPP panel, started in 2004, has approximately 43,000 households interviewed at Wave 1 (110,700 individuals of whom 29,700 are under age 18). It is scheduled for 12 waves.
SIPP is divided into core content, which is collected during each wave, and topical content, which is collected only during certain waves. SIPP collects detailed social and economic information, including program participation. It also collects some self-reported health measures, such as limitations in activities of daily living and self-assessed health status. Besides some health information, it queries income sources so that household income can be constructed. In principle, then, SIPP can be used to construct a household-level variable, such as income or wealth. However, SIPP does not obtain information on food expenditures. Low-income households are oversampled, resulting in about an 11 percent increase in the number of low-income households compared with what would be without oversampling.3 SIPP has included a subset of six food insecurity questions (but not the standard six-item set) in the adult well-being module once during each panel beginning in 1998.
Because of funding restrictions, the sample size and design of SIPP has changed often in the past. Whether it will retain a stable design and sample size is not known, but the changing design is a feature of SIPP that makes it less attractive than, say, the CPS. Moreover, the content of SIPP is decided
by an interagency committee, which has to review an agency’s request before questions can be added to the survey. The survey has a lengthy processing cycle. The period between data collection and data availability can be one to two years.
RELATIVE ADVANTAGES AND DISADVANTAGES
Because food insecurity and hunger as now measured are relatively rare events, fairly large sample sizes are required to estimate accurately their prevalence in subpopulations of interest.
The CPS December interview has been used for the past few years as the vehicle for the Food Security Supplement. An important advantage of CPS is its widespread acceptance as an authoritative source of statistical information. A second advantage is its sample size of about 50–60,000 households and its state-level representative sample. Thus one can get reliable estimates of prevalence in subpopulations of interest. A third advantage is the timeliness of the reporting of data. CPS does, however, have a number of important disadvantages. It is a survey of households; only the respondent is surveyed, so that any reports about the hunger of the individuals in the household are by proxy. A second disadvantage is that detailed income information is collected in the CPS only in the Annual Social and Economic Supplement. Household income may have changed, and that very change may have given rise to food insecurity. Household composition may have changed, resulting in the loss of income or the gain of income. As with most cross-sectional household surveys, CPS is based on dwelling units, so that subpopulations that move often would be less likely to be interviewed in both December and the Social and Economic Supplement. No health information is obtained in the CPS.
For reasons of sample size, and not necessarily the content, the panel concludes that NHANES is not an appropriate primary vehicle for monitoring the prevalence of food insecurity in the population at the national and state levels. It could be an important vehicle for a research program for understanding the relationship between food insecurity and hunger and indicators of inadequate nutrition and other health characteristics. It could also be an appropriate vehicle for the new research recommended in Chapter 3 on aspects of deprivation, alienation, and family and social interactions. As indicated in the previous section, USDA has been collaborating with NHANES in collecting data on the HFSSM and has started analyzing those data.
Recommendation 6-1: USDA should continue to collaborate with the National Center for Health Statistics to use the National Health and Nutrition Examination Survey to conduct research on methods
of measuring household food insecurity and individual hunger and the consequences for nutritional intake and other relevant health measures.
In Chapter 3 the panel concluded that food insecurity is a household-level concept. It is rooted in the lack of economic resources in a household. Hunger is distinct from food insecurity and is an individual-level concept. Both measures are important, and to measure both concepts, the panel recommends that USDA should undertake a research program on how best to measure individual hunger and other important consequences of food insecurity. The CPS cannot now be used to measure hunger in the population because it interviews only the household respondent. While the CPS can ask the household respondent about his or her experience with hunger, no one respondent is likely to be representative of all adults in the household.
The NHIS interviews a randomly chosen adult who can be asked about his or her experience with hunger. NHIS is based on a probability sample; one can use the responses to estimate the prevalence of hunger in the adult population. With the randomly chosen child, one can estimate the prevalence of hunger among children. Similarly, the consequences to individuals of food insecurity or hunger, such as health problems and social exclusion or alienation, can be ascertained from the interviews. Another major positive aspect of the NHIS is its extensive measures of health, which could be linked to food insecurity or hunger.
Recall bias is also a concern about the current reliance on the December CPS. Asking a respondent to recall instances of food insecurity over a 12-month period is likely to produce recall bias, in which the respondent overweights the current situation. This bias is probably exacerbated by the measurement of food insecurity in the December CPS because December is not representative of the experience over the whole year. Although the FSS has probe questions and questions about a 30-day recall, the NHIS collects data across the entire year (which increases the cost of the supplement to USDA) and also has the capability to estimate the frequency and duration of food insecurity. Ultimately, USDA would need to weigh comparative costs and benefits, timeliness, and the ability to include the supplement on a regular basis in the NHIS, among other issues.
Recommendation 6-2: USDA should carefully review the strengths and weaknesses of the National Health Interview Survey in relation to the Current Population Survey in order to determine the best possible survey vehicle for the Food Security Supplement at a future date. In the meantime, the Food Security Supplement should continue to be conducted in the Current Population Survey.
The four-month intervals between SIPP interviews make it well suited to ask about intervals of food insecurity and/or hunger over both short intervals such as last week or last month, or over the last four months. Because SIPP measures contemporaneous income, analysis can find the relationships between income and food insecurity and/or hunger. This could be particularly important to find the direct effects of income loss via, say, unemployment or a health shock and to separate out short-term from long-term food insecurity.
CPS is limited in its ability to obtain information that would permit the scientific study of the antecedents and consequences of food insecurity and hunger. Such studies require panel data, such as in SIPP. Panel data would provide an additional benefit to USDA. They would permit a better assessment of the success of such programs as food stamps, because analysts could study the dynamics of economic need, food insecurity, and subsequent relief due to food stamps.
The panel recognizes that in an era of budgetary constraints it would be difficult for USDA to accomplish all the goals of reliably measuring food insecurity prevalence, hunger and their dynamics. But we note that not all measures need to be collected at all times. In particular, USDA should explore the feasibility of a one-time multiwave study using SIPP to examine the dynamic relationships among income, food insecurity, and hunger. Especially useful would be a study of the persistence of food insecurity and hunger, including its prevalence and frequency. The four-month interviewing cycle of SIPP should provide much more accurate information than the CPS. In addition, data from the shorter recall period could be compared with those from a 12-month recall period to study any bias in the current 12-month recall design.
Recommendation 6-3: USDA should explore the feasibility of funding a one-time panel study, preferably using the Survey of Income and Program Participation, to establish the relationship between household food insecurity and individual hunger and how they co-evolve with income and health.