4
Other Federal Data Sources

A number of other federal datasets contain information that is relevant to food consumption and nutrition monitoring and could be used to address policy issues. Most of these datasets have purposes that are not directly related to food and diet monitoring, but they do contain some useful information. In this chapter, we briefly review several of these data sources and discuss possible ways they could be used to address food policy questions. To add value for this purpose, consideration could be given to enhancing one or more of these datasets in ways similar to those suggested in Chapter 2 for the NHANES and CE surveys—for example, by matching survey records with administrative records for food assistance programs to add to or improve the quality of program participation data, by appending neighborhood characteristics of various kinds from the 2000 census, the American Community Survey (when small-area data become available), and various geographic databases, or by adding supplemental modules with additional food and nutrition-related questions.

Our discussion is organized around five types of data. They are: data on monetary resources for food consumption, including food insecurity (the Current Population Survey); time use data (the American Time Use Survey); data sources for longitudinal analysis of food consumption and related behavior over a span of years (the Early Childhood Longitudinal Study, the Health and Retirement Study, and the Panel Study of Income Dynamics); data sources for relatively quick-turnaround studies of emerging issues, which can also provide key trends for the nation, regions, and



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies 4 Other Federal Data Sources A number of other federal datasets contain information that is relevant to food consumption and nutrition monitoring and could be used to address policy issues. Most of these datasets have purposes that are not directly related to food and diet monitoring, but they do contain some useful information. In this chapter, we briefly review several of these data sources and discuss possible ways they could be used to address food policy questions. To add value for this purpose, consideration could be given to enhancing one or more of these datasets in ways similar to those suggested in Chapter 2 for the NHANES and CE surveys—for example, by matching survey records with administrative records for food assistance programs to add to or improve the quality of program participation data, by appending neighborhood characteristics of various kinds from the 2000 census, the American Community Survey (when small-area data become available), and various geographic databases, or by adding supplemental modules with additional food and nutrition-related questions. Our discussion is organized around five types of data. They are: data on monetary resources for food consumption, including food insecurity (the Current Population Survey); time use data (the American Time Use Survey); data sources for longitudinal analysis of food consumption and related behavior over a span of years (the Early Childhood Longitudinal Study, the Health and Retirement Study, and the Panel Study of Income Dynamics); data sources for relatively quick-turnaround studies of emerging issues, which can also provide key trends for the nation, regions, and

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies states (the Behavioral Risk Factor Surveillance System and the State and Local Area Integrated Telephone Survey); and a possible data source for analyzing food consumption behavior of the low-income population (the Expanded Food and Nutrition Education Program). CURRENT POPULATION SURVEY The CPS is an ongoing monthly survey of about 56,000 households, which is fielded by the U.S. Census Bureau and supported by the Bureau of Labor Statistics. (The Census Bureau and other federal agencies pay for periodic supplements to the main labor force survey.) Its primary purpose is to provide estimates of employment, unemployment, and other characteristics of the labor force. Each year, beginning in 1995, USDA has supported a supplement (currently fielded in December) on food expenditures, food assistance program participation, food insecurity, and ways of coping with not having enough food—see Box 4-1. The CPS sample design is state representative, and its large size will support state-level estimates when the data are averaged over 3 years. Response rates are high, averaging 92 percent for the main labor force survey, although 12 percent of households do not complete the food insecurity module.1 The food expenditure, program participation, and food insecurity data from the December supplement can be analyzed with information from the main CPS questionnaire, which includes demographic characteristics for all household members, detailed information on labor force participation and usual hours worked and earnings for household members aged 15 and older, and total household income. Because of the rotating design of the CPS and the recent expansion of the March income supplement to include households in February and April, about one-half of the households in the December sample can be matched with the same households in February or March that have detailed income, program participation, and health insurance coverage information for the preceding calendar year from the renamed Annual Social and Economic Supplement.2 The combi- 1   Personal communication from Mark Nord, Economic Research Service, USDA, May 31, 2005. 2   More precisely, the same addresses can be matched. If a household has moved, the match will not represent the same people (see www.bls.census.gov/cps/asec/2003/sdataqua.htm [June 2005]).

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies BOX 4-1 Food and Nutrition-Related Data in the December Current Population Survey Where Bought Food Last Week Supermarket or grocery store Other places where people buy food (meat markets, produce stands, bakeries, warehouse clubs, convenience stores) Restaurant, fast-food place, cafeteria, vending machine Any other place How Much Spent on Food Last Week Supermarkets and grocery stores (how much of total for nonfood items) Stores such as meat markets, produce stands, etc. (how much of total for non-food items) Restaurants, fast-food places, cafeterias, vending machines Other places Whether Would Need to Spend More or Less for Just Enough Food and How Much More or Less Food Assistance Program Participation Receive food stamps in last 12 months and which months Amount of most recent food stamp benefit Children aged 5-18 receive free or reduced-price lunches at school in last month Children receive free or reduced-price breakfasts at school in last month Children receive free or reduced-price food at day care or Head Start center in last month How many women and children receive WIC foods in last month Food Insecurity Scale Questions for Households with and without Children Ways of Coping with Not Having Enough Food Receive meals from “Meals on Wheels” or other community programs in last month Eat prepared meals at a community program or senior center in last month Get emergency food from a food pantry, church, or food bank in last 12 months and how often Have a source of emergency food nearby Eat meals at a soup kitchen in last 12 months and how often SOURCE: National Research Council (2005b: Appendix A).

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies nation of detailed income and program participation information and the food expenditure and insecurity data for matched households could support in-depth analysis of the effects of income constraints on food purchasing, the role of food assistance programs in alleviating food insecurity, whether people who lack health insurance coverage are more or less food insecure than households with public or private coverage, and similar topics. AMERICAN TIME USE SURVEY The American Time Use Survey (ATUS) is a relatively new, ongoing survey conducted by the U.S. Census Bureau for the Bureau of Labor Statistics (BLS) to measure how Americans spend their time. Data collection for ATUS began in January 2003 (see Abraham, 2004). Reports are produced quarterly and annually. The sample population is drawn each month from households in the outgoing rotation group of the CPS (see above), except that the sample is somewhat smaller, especially in less-populous states (those oversampled in the CPS). The ATUS household sample is stratified by race and ethnicity of the householder, presence and age of children, and number of adults in households without children. Households that have a Hispanic or black householder and households with children are oversampled. One respondent is selected randomly from those aged 15 and over in each household in the sample. The current sample size is about 26,000 households annually (see Herz, 2004; see also www.bls.gov/tus/home.htm [June 2005]). Respondents are sent instructions and time diary materials and are assigned a day for which to report their activities. Interviews are conducted by telephone on the day following the assigned day. ATUS provides an incentive for households without telephones to call on the scheduled interview date, but the response rate for households without telephones is about one-half the rate for households with telephones, and the overall response rate is low: it was 57 percent in 2003 (Bureau of Labor Statistics, 2004a). Survey questions include review of the time diary; height and weight of the respondents, to permit calculation of the body mass index (BMI) and determination of being overweight or obese; identification of persons who were with the respondent during various times of the day; work summary questions; summary questions about secondary child care (that is, taking care of a child while doing something else); volunteering summary questions; and questions about trips away from home for 2 or more nights in a row. If respondents engaged in two or more activities simultaneously, they

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies must identify which was the primary activity, and that activity is coded as the only activity (Herz, 2004). Because the sample comprises members of households that completed their eighth and final CPS main labor force survey, the employment and earnings data from that survey are available for use with the ATUS information. ATUS allows supplemental modules to be added. Modules are developed in cooperation with the BLS ATUS staff, and all modules’ questions must be pretested to ensure that they elicit the desired information. Modules must run for a minimum of six months (Herz, 2004). ATUS has many possible uses for food and nutrition-related research and policy analysis, and USDA’s Economic Research Service (ERS) is currently working with BLS to plan a food and eating module to include in ATUS. The module is intended to provide data that can be used to analyze the relationship between patterns of time use and eating patterns, nutrition, and obesity, as well as food assistance program participation and grocery shopping and meal preparation. One potential use for the proposed ATUS food and eating module is to estimate how much time it takes to prepare meals. Although time is not part of the “costs” calculated in USDA’s four food plans (thrifty, low-cost, moderate, and liberal), data from the ATUS could be used to give a more robust estimate of the costs of meal preparation at various expenditure levels, including time costs. Another potential use of ATUS is to understand and improve access to food assistance programs by indicating how much time it takes to acquire, use, and retain program benefits, such as food stamps and WIC (Frost, 2004). Such data could also be used to understand how food consumption and preparation practices vary across households with different work schedules and arrangements. The time tradeoffs for meal preparation and the effects of recent policy initiatives for increased work effort among low-income populations could also be studied. One feature of the food and eating module is that it will code eating as a secondary activity if, for example, an individual’s primary activity is watching television. The module will also gather information on whether the individual was snacking on food (Hamrick, 2004). The ATUS food and eating module could be used to determine if snacking and the number of eating episodes throughout the day are increasing, whether it takes longer on average to prepare a meal at home than to eat away from home, what technologies and practices provide time savings for home meal preparation, and whether there are true time “shortages” that are leading to less cooking. There could also be questions that would provide information about the

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies time it takes consumers to read and understand nutrition labels and ingredient lists. Other possible uses are easy to imagine. They include how the number of eating episodes is related to obesity; how physical activity (or lack of it) affects eating patterns and obesity;3 how income is related to the number of eating episodes, eating at home or away from home, and consuming previously prepared foods; how marital, social, and economic status relate to food shopping time; how food-related activities function as child care; demographic factors that relate to eating times; and which groups of people are able to eat during work and what effect (if any) this has on being overweight or obese (Hamermesh, 2004). PANEL SURVEYS For many kinds of analysis, particularly to inform policy planning, it is desirable to have measures on the same individuals over time. Longitudinal information from a panel survey that repeatedly interviews the same respondents would facilitate research on changes in food consumption behavior, diet, and health at the household or individual level and how they might relate to such factors as changes in income and program participation, initiatives for food education and safety, or changes in other contextual factors. Longitudinal data are usually expensive to collect, so that fielding a comprehensive new panel survey specifically for food and nutrition-related behavioral analysis does not seem feasible with the resources currently available to the Economic Research Service. However, several existing longitudinal surveys have the potential for use by ERS and other relevant agencies and perhaps could be enhanced to better support food and nutrition-related analyses.4 Early Childhood Longitudinal Study The Early Childhood Longitudinal Study (ECLS) is sponsored by the National Center for Education Statistics in collaboration with several other 3   See Dong, Block, and Mandel (2004) for an analysis of energy expenditure, using data from EPA’s 1992-1994 National Human Activity Pattern Survey, which collected 24-hour time diaries by telephone. 4   In addition to the three panel surveys described in the text, a number of other panel surveys could potentially be useful for nutrition-related analysis, such as the National Longitudinal Surveys of BLS (see Logan, Fox, and Lin, 2002).

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies agencies. This study follows two cohorts of children to collect information on young children and their family, school, and community environments. About 22,000 children at about 1,000 public and private, part-day and full-day kindergartens from the kindergarten class of 1998-1999 make up the kindergarten cohort (see West, Denton, and Reaney, 2000). Data were collected in the fall and spring of their kindergarten and first-grade school years and in the spring of their third- and fifth-grade school years, and it is planned to follow these children through twelfth grade. Questionnaires were administered to parents, teachers, and the children themselves (for details, see nces.ed.gov/ecls/kindergarten.asp [June 2005]). The birth cohort survey is following more than 10,600 children born in 2001. Data were collected through parent and caregiver questionnaires when the children were 9 months and 2 years old and will be collected when they are 4 years old and in kindergarten (for details, see nces.ed.gov/ecls/birth.asp [June 2005]). The parent questionnaire for both cohorts includes the USDA’s food insecurity scale questions, which provide information about the food security status of the children’s families. Both cohort questionnaire sets include questions on the height, weight, and physical activity of the children and about participation in WIC and the Food Stamp Program. The parent questionnaire for the birth cohort at 2 years old asked about breastfeeding, formula use, and other beverage consumption by the child of interest. The parent questionnaire for the kindergarten cohort asks about school lunch and breakfast program participation, and the fifth-grade questionnaire asked children about their purchase of sweet and salty snacks and soda at school, and their consumption in the past 7 days of milk, juice, soda, carrots, green salad, potatoes, other vegetables, fruit, and meals at fast-food outlets. Health and Retirement Study The Health and Retirement Study is an ongoing panel survey of about 22,000 people who were aged 51 and over when they were first interviewed and their spouses. Blacks, Hispanics, and residents of the state of Florida are oversampled. The initial cohorts consisted of about 12,500 people aged 51-61 and about 8,000 people aged 70 and older in 1992-1993. New cohorts of people aged 56-61 and 69-75 and their spouses were introduced in 1998, filling in the entire older age span. The survey attained a steady state in 2004 with the introduction of another new cohort of people aged

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies 51-56. New cohorts will be introduced in a similar manner every 6 years (sample sizes are smaller for newer cohorts than for the original HRS cohort). Interviews are conducted in person the first year a household is included in the sample and by telephone every 2 years thereafter. Response rates for each cohort have been 80 percent or less at the original interview and 90 percent or more at subsequent interviews. The survey is conducted by the University of Michigan with funding from the National Institute on Aging (see hrsonline.isr.umich.edu [June 2005]). The HRS collects comprehensive information on many characteristics: demographic background; disability; employment status and job history; family structure and transfers; self-reported health status and medical conditions; self-reported smoking, drinking, and exercise; cognitive status; health insurance and pension plans; housing; income and net worth; retirement plans and perspectives; and attitudes, preferences, expectations, and subjective probabilities related to retirement. The HRS also collects information on housing costs, out-of-pocket medical care expenditures, food expenditures per week or month, in stores and delivered, and expenditures for meals eaten out. Versions of the HRS are available under special access arrangements with links to Medicare and Social Security earnings and benefits. In addition to questions that are asked at every interview, the HRS typically includes a large number of modules with supplemental or experimental questions, which could be used to ask questions related to food consumption and related topics. A special supplemental survey, the 2001 Consumption and Activities Mail Survey (CAMS), obtained relevant information on time use and spending. It was mailed to a random sample of 5,000 HRS respondents and there were 3,800 usable responses. CAMS covered time use (36 categories), spending (32 categories, including food-related items), and anticipated and actual changes in spending pre- and post-retirement (Hurd and Rohwedder, 2003). The HRS is a source of extensive information about the financial well-being and health situation of older Americans and, more important, how their situations change as they age and experience such life events as retirement or loss of a spouse. Consideration could be given to enhancing the food and nutrition-related content of the core questionnaire beyond the limited information obtained on food expenditures, as well as to adding modules about food and nutrition. Such enhancements would support

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies behavioral analyses of food consumption behavior and health effects for the older population. Panel Study of Income Dynamics The Panel Study of Income Dynamics (PSID) is a continuing panel survey of a cohort of families that began in 1968. The survey is sponsored and conducted by the University of Michigan Survey Research Center. Since 1983 the National Science Foundation has been the principal funder, with substantial continuing support from the Office of the Assistant Secretary for Planning and Evaluation in the U.S. Department of Health and Human Services and some support from other agencies, including USDA (see psidonline.isr.umich.edu [June 2005]). The original PSID sample comprised two components: (1) 2,900 families drawn from the Survey Research Center national sampling frame, representative of the civilian, noninstitutionalized population; (2) 1,900 low-income families with heads under age 60 drawn from the 1966-1967 Survey of Economic Opportunity conducted by the U.S. Census Bureau. In 1990, 2,000 Hispanic families were added, but these families were subsequently dropped in 1996, and 441 immigrant families (including Asians) were added in 1997. Currently, more than 7,000 families (including original sample families and the subsequent families of their members) are interviewed once every other year, mostly by telephone (prior to 1997, interviews were conducted annually). The PSID experienced a large sample loss—24 percent—at the initial interview in 1968, but additional sample loss dropped to 8 percent of the eligible families at the second interview, and it was only 1-2 percent at each interview thereafter. The extent to which attrition introduces bias into estimates from the PSID is not clear; some studies have reported little effect; others have found some biases in estimates of poverty rates before the new Hispanic sample was added in 1990 (see National Research Council, 1995:App. B). The core content of the PSID includes many elements: family members’ demographic characteristics; detailed employment histories and income by source for the household head and spouse; less detailed income information for other family members; program participation, including amounts and months received for food stamps; estimates of federal taxes paid; housing costs; average weekly food expenditures for home consump-

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies tion and away from home; housework time; socioeconomic background; religion; and health status. Versions of the PSID are available under special access arrangements that contain geographic match codes for locations of PSID households down to the census tract level. With these match codes, researchers can append neighborhood data from the decennial census long-form sample or other files to the PSID records. The PSID has included supplements on many topics, including eligibility for food stamps and Supplemental Security Income, smoking and exercise, time use, and wealth, among others. The PSID is the longest running nationwide panel survey in the United States with detailed socioeconomic information, and it has some information that is relevant for food and nutrition policy analysis. It could be worthwhile for USDA to explore ways to add questions to this rich data source. QUICK-TURNAROUND SURVEYS There may be some cases for which a few questions about dieting practices and attitudes or concerns about food safety will provide useful information for monitoring food market trends or food consumption behavior. For example, USDA may want to know how many Americans are practicing the Atkins diet or how people are reacting to stories about mad cow disease. Such information may be used to begin to understand a trend, particularly when policy makers do not want to wait for the results of other surveys that collect this information. Two data collection programs that are geared to the addition of questions and modules to track emerging trends—the Behavioral Risk Factor Surveillance System (BRFSS) and the State and Local Area Integrated Telephone Survey (SLAITS)—have the potential to serve this purpose. These programs also have the advantage that they can provide state-level estimates. A disadvantage is that they are limited to households with land-line telephones. Behavioral Risk Factor Surveillance System The BRFSS is an ongoing cross-sectional survey designed by the Centers for Disease Control and Prevention (CDC) and conducted by health departments of the states and territories with technical advice and oversight by the CDC (see www.cdc.gov/brfss/index.htm [June 2005]). The purpose of the survey is to track the health habits of the population. Information is collected by telephone interviews each month from a random

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies sample of adults aged 18 and over in each state. CDC aggregates the results for each year and provides them to states to use to direct health promotion and disease prevention programs. The sample size is about 4,000 interviews per state. The CDC initiated the BRFSS in 1984, at which time 15 states participated in monthly data collection. CDC developed standard core questionnaire for states to use to provide data that could be compared across states. By 1994, all states, the District of Columbia, and three territories were participating in BRFSS. The BRFSS was designed to collect state-level data, and a number of states from the outset stratified their samples to allow them to develop estimates for intrastate areas. The core survey, which includes fixed questions, rotating questions asked every other year, and several questions on “emerging issues,” is reviewed every year and changes are made by CDC and the states working together. All states administer the core survey and may also choose among several optional modules; they may also add their own questions. The core survey contains no food or nutrition-related questions except those regarding consumption of alcohol, a question on physical exercise, and self-reported height and weight. Several modules do have such questions. Optional Module 12, on cardiovascular disease, specifically asks if respondents are eating fewer high-fat or high-cholesterol foods and if they are eating more fruits and vegetables to reduce the risk of developing heart disease. Module 13, on folic acid, includes questions about supplement use and specifically whether supplements used contain folic acid. Module 19, on binge drinking, asks a number of follow-up questions to determine risky health behavior related to binge drinking. The information currently available from Modules 12 and 13 can be used to understand reasons for changes in fat and cholesterol consumption, on one hand, and fruit and vegetable consumption on the other, as well as the use of supplements, multivitamins, and, specifically, folic acid. It could be possible to add questions to Module 13 if further information about supplement use is desired. Other modules could be developed to address specific food and nutrition behaviors. Such modules could include a food frequency questionnaire designed to track food consumption behaviors with confirmed or suspected health risks or benefits, such as excessive consumption of seafood with high levels of mercury. The food frequency questionnaire could also provide information on overall eating practices that could be linked to the self-reported heights and weights in the core BRFSS that are used to calculate

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies BMI and the prevalence of obesity. Since the BRFSS is a sample survey within each state, however, modules can be added only through negotiations with each state. State and Local Area Integrated Telephone Survey Another possibility for obtaining food and nutrition-related information to track emerging trends would be to make use of SLAITS. SLAITS is a mechanism for government agencies and other sponsors to obtain customized state-level information by using the sampling frame from the National Immunization Survey (NIS). The NIS, in turn, is an ongoing telephone survey conducted by the National Center for Health Statistics (NCHS) that screens almost 1 million households per year to produce estimates of vaccination coverage levels among children aged 19-35 months. NCHS will work with sponsors to design a specific questionnaire and sampling scheme that can be piggybacked on the NIS at any time. Typically, it takes 3-6 months of design and testing before data collection can begin. Sponsors can use previously developed SLAITS modules (which include health, child well-being and welfare, early childhood health, and asthma) or specify new modules. A SLAITS sample can be designed to target population groups, such as low-income households or those with specific characteristics, and NCHS will adjust the results for noncoverage of households without land-line telephones (see www.cdc.gov/nchs/slaits.htm [June 2005]). EXPANDED FOOD AND NUTRITION EDUCATION PROGRAM A dataset that could be used to understand food consumption in populations served by food assistance programs comes from the Expanded Food and Nutrition Education Program (EFNEP), which is part of USDA’s Cooperative State Research, Education and Extension Service. The EFNEP program provides advice and counseling to low-income families with children through an experiential education program to improve nutrition, food resource management, and food safety behaviors. The goal is to enable participants to provide nutritionally adequate meals for themselves and their families. As part of the program, EFNEP participants fill out 24-hour food recall forms and a food practices checklist at initiation and at the completion of locally administered courses in which they learn such skills as food budgeting, selection, preparation, storage, and safety.

OCR for page 63
Improving Data to Analyze Food and Nutrition Policies About 150,000 low-income families participate in EFNEP each year, and food records have been collected by EFNEP for about 100,000 individuals for each of the past 10 years. The food records are entered into an EFNEP Evaluation/Reporting System, which is used to determine overall diet quality, based on key indicators: total fat, protein, carbohydrate, fiber, calories, iron, calcium, and vitamins A, C, and B6, as well as the number of servings of each of the food guide pyramid food groups. Also administered with the food recall is a 10-item food practice checklist covering other behaviors of interest to EFNEP, including food safety, meal planning, use of nutrition labeling, comparing prices, and having children eat breakfast. Aggregate data are available as national summaries, as well as by state and by race—white, black, Hispanic, Native American, and Asian/Pacific Islander. Individual-level data are not currently available for research use, but future plans may include a release of individual-level record data. The EFNEP data have yet to be used for policy analysis purposes outside of the EFNEP program. Thus, the utility of these data is not fully known. The large samples of low-income individuals and survey questions are potentially valuable for enhancing understanding of policy issues for the low-income population. But since participation in this program is voluntary, participants are not a random sample of the low-income population and are likely to be different from those low-income individuals who did not choose to participate.