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 26
Improving Data to Analyze Food and Nutrition Policies 2 Federal Datasets on Food and Nutrition This chapter reviews the primary sources of data from federal surveys on food consumption, food expenditures, and dietary attitudes and knowledge. Those surveys are the National Health and Nutrition Examination Survey, the Continuing Survey of Food Intakes by Individuals, which was incorporated into NHANES beginning in 2002, and the Consumer Expenditure Survey. Table 2-1 summarizes the design and lists the content relevant to food and nutrition policy and research of these three surveys and the Diet and Health Knowledge Survey that was included in two rounds of the Continuing Survey of Food Intakes by Individuals. The periodic Health and Diet Survey discussed in Chapter 1 also provides information on adults’ health-related knowledge and attitudes. NATIONAL HEALTH AND NUTRITION EXAMINATION SURVEY The continuing National Health and Nutrition Examination Survey (NHANES) is conducted by the National Center for Health Statistics (NCHS) of the Centers for Disease Control and Prevention in the U.S. Department of Health and Human Services. NHANES monitors the health and dietary practices and outcomes of Americans and is used in developing public health policy. The dietary component of NHANES, a cooperative effort of the Departments of Agriculture and Health and Human Services, is called “What We Eat in America” (www.barc.usda.gov/bhnrc/foodsurvey/wweia.html [June 2005]). NCHS is responsible for the sample design and
OCR for page 27
Improving Data to Analyze Food and Nutrition Policies data collection for the dietary component; USDA is responsible for the design of the dietary data collection procedure, maintenance of the databases that are used to code and process the dietary intake information, and review and processing of the dietary information (see “Integrated NHANES and CSFII” below). The NHANES has evolved from a periodic survey that originated in 1971, when the National Health Examination Survey was combined with the National Nutrition Surveillance System, to a continuously fielded survey. NHANES I data were collected for 1971-1975, NHANES II for 1976-1980, and NHANES III for 1988-1994. There was also an Hispanic Health and Nutrition Examination Survey (HHANES), which was conducted from 1982 through 1984. In 1999, NHANES became an ongoing survey, with detailed health, nutrition, and medical information collected from about 5,000 participants annually. Beginning with data for 1999-2000, NHANES findings have been released at 2-year intervals. To enable users to produce estimates from the NHANES public-use microdata files with sufficient reliability, sample weights are provided on a pooled basis for each 2 years’ worth of information. To protect confidentiality of respondents, not all variables are included in the public-use microdata files. The sampling process is stratified and multistage: counties or groups of contiguous small counties are designated as primary sampling units (PSUs), and each year 15 PSUs are selected into the sample for household visits. Within PSUs, blocks or groups of blocks are selected, then households, and, finally, one or more individuals within households. The sample does not include people living in institutions or members of the Armed Forces. At sampled households, interviewers obtain the demographic characteristics of all household members, and one or more household members are selected for interview and examination by using fixed sampling fractions that distribute the sample into specific age-sex-race-ethnicity-income categories. If a child under the age of 6 is selected into the sample, then a proxy interview is conducted with the child’s primary caretaker. Interviews with children aged 6-11 years old are conducted through an assisted interview with a caretaker present. Blacks, Mexican-Americans, 12- to 19-year-olds, people aged 60 and over, pregnant women, and (beginning in 2000) people in low-income households are oversampled (see www.cdc.gov/nchs/data/nhanes/guidelines1.pdf [June 2005]). In addition to demographic and economic background information and self-reported health status for each individual in the sample, physiological information, including precise measurements of height and weight
OCR for page 28
Improving Data to Analyze Food and Nutrition Policies TABLE 2-1 Design Features and Relevant Content for Food and Nutrition Research of Four Surveys: the National Health and Nutrition Examination Survey (NHANES), the Continuing Survey of Food Intakes by Individuals (CSFII), the Diet and Health Knowledge Survey (DHKS), and the Consumer Expenditure Survey (CE) Feature NHANES CSFII Time Period Continuing since 1999; released at approximately 2-year intervals, with data pooled for 2 years for reliability Conducted most recently in 1989-1991, 1994-1996, and 1998 (children aged 0-9 only); folded into NHANES in 2002 Universe Civilian noninstitutionalized population 1994-1996: Same as NHANES Design Each year about 11,000 households screened in 15 primary sampling units (counties); about 3,000 households identified with 6,000 eligible people, of whom 5,000-5,500 interviewed and 4,600-5,200 examined in mobile exam centers (MEC) Oversampling of blacks, Mexican-Americans, low-income people not black or Mexican (beginning in 2000), people aged 12-19 or aged 60 and over, pregnant women 1994-1996:15,000 people in households (5,000 people per year) provided dietary intake information Oversampling of low-income people Major Questionnaire Components Household screeners; Family questionnaire; Sample person questionnaire; MEC audio computer-assisted self interview; MEC computer-assisted personal interview; MEC dietary recall; MEC examination; MEC laboratory analysis Household interview; First-day in-person dietary recall; Second-day in-person dietary recall
OCR for page 29
Improving Data to Analyze Food and Nutrition Policies DHKS CE Conducted as a supplement to CSFII in 1989-1991 and 1994-1996; not now carried over to NHANES Continuing since 1980; released yearly; two components, Diary Survey and Household Survey Same as NHANES Same as NHANES Random sample of adults aged 20 and over with a completed dietary intake for day one, who were interviewed by telephone follow-up 2-3 weeks after the second day of dietary intake Household Survey: 7,500 consumer units per year; each month, one-fifth of sample is new; households in sample for five quarterly in-person interviews; respondent is anyone aged 16 or older who knows household finances Diary Survey: 7,500 consumer units per year, each of which fills out two consecutive weekly diaries No oversampling Knowledge questions; Attitude questions; Factors related to grocery shopping; Food label questions; Behavior questions; Food safety questions Household Survey: Demographic characteristics; Work experience; Expenditures by month (65% of items); Usual expenditures per quarter (35% of items); Real assets; Financial assets; Last 12-months’ income; Taxes Diary Survey: Demographic characteristics; Work experience; Income; Taxes; First week diary; Second week diary
OCR for page 30
Improving Data to Analyze Food and Nutrition Policies Feature NHANES CSFII Household Background Characteristics Family size, ages, relationships (data collected for sampling but not publicly released); health insurance; housing characteristics; pesticide use; smoking Family size; tenancy Person Background Characteristics Sampled persons: Age, race, sex; country of birth, marital status, early childhood; education; occupation; social support All household members: Age, race, sex Members aged 15 and over: Education, employment status, occupation, hours worked last week, usual hours worked, reason not working Income Total household income in last 12 months; who received different income types in last 12 months (number of months for welfare assistance) Total household income last year; last month’s income by source; savings or cash assets under $5,000 Food Assistance Program Participation Household and sampled person food stamp and WIC participation in last 12 months, number months receiving food stamps and WIC; amount food stamps received by household last month; sampled person participation in school breakfast and lunch (usual times per week), Meals on Wheels in past 12 months, and summer program meals Household food stamp participation in last 12 months, value of food stamps, when last received, members eligible now; each member’s WIC participation and how long; school-age children’s school breakfast and lunch participation; younger children’s child care feeding participation
OCR for page 31
Improving Data to Analyze Food and Nutrition Policies DHKS CE Obtained from main CFSII Household Survey: Family size; health insurance; inventory of housing and financial assets, durable goods (first quarter) Diary Survey: Family size; housing, vehicle ownership Obtained from main CFSII Household Survey: All household members: Age, race, sex, marital status, education; Members aged 14 and over: Work experience, job characteristics Diary Survey: Same as Household Obtained from main CSFII Household Survey: Members aged 14 and over: Earnings and retirement benefits (each quarter); Income by source, prior 12 months (2nd and 5th interviews) Diary Survey: Members aged 14 and over: Earnings and retirement benefits Household: Other income by source, prior 12 months Obtained from main CSFII Household Survey: Benefits from food stamps (and months received); value of other free meals (each quarter) Diary Survey: Same as Household
OCR for page 32
Improving Data to Analyze Food and Nutrition Policies Feature NHANES CSFII Food Expenditures None Usual weekly or monthly household expenditures at grocery stores (total and on food items), at specialty stores, at fast food or carryout places (for food brought into the home and food bought and eaten away from home) Food Eating and Shopping Practices Sampled persons: Where food obtained; number of times eat at restaurant in a week Sampled persons: Where food obtained (store, restaurant, fast food place, etc.); whether eaten at home or away; sources of water for drinking, cooking, preparing beverages; time of eating; name of eating occasion (e.g., breakfast); who in household does planning, shopping, and preparing meals; how often shop and type of store Food Security 18-item household food security scale module; individual-level questions for children, adolescents, and adults Food sufficiency indicator
OCR for page 33
Improving Data to Analyze Food and Nutrition Policies DHKS CE Obtained from main CSFII Household Survey, Usual weekly expenditures at supermarkets, specialty food stores, school meals; usual monthly expenditures for liquor at home, liquor away from home, food away from home (every quarter) Diary Survey: Usual weekly expenditures at supermarkets, specialty food stores (last month); From the diary: Price of each purchase of food away from home, categorized by fast food-type outlets, full-service meals, vending machines, employer and school cafeterias, catered affairs; price of each item purchased for consumption at home, categorized by grain products, bakery products, beef, pork, poultry, other meats, fish and seafood, fats and dressings, eggs and dairy products, fruits and juices, sugars, vegetables, other food items, nonalcoholic drinks, alcoholic drinks Obtained from main CSFII Where food obtained (see “Food Expenditures” above) Obtained from main CSFII None
OCR for page 34
Improving Data to Analyze Food and Nutrition Policies Feature NHANES CSFII Person Health and Nutrition Characteristics Sampled persons: Alcohol use; balance; biochemistries and hematologies; blood lipids; blood pressure; blood and urine analysis; body composition; bone markers; dermatology; diabetes; heart disease; illegal and prescription drug use; immunization status; kidney disease; medical conditions; oral health; osteoporosis; pain; physical activity; pregnancy status; respiratory health; self-assessed health status; sexual behavior; smoking; vision; weight and height measures; weight history Sampled persons: Alcohol use; food allergies; physician-diagnosed diseases; physical activity; hours of TV or videos watched yesterday; pregnancy or lactation status; self-assessed height and weight; self-assessed health status; smoking Person Dietary Intake Sampled persons: 2-day 24-hour dietary recall of foods and amounts eaten (first day at the MEC; second, nonconsecutive day by telephone); all dietary supplements (e.g., vitamins, herbals—name, strength, days and amount in past 30 days, dosage, length of use); medications; water intake; salt use; special diet; milk consumption history; frequency of consumption of milk, green leafy vegetables, legumes, fish, shellfish; comprehensive food frequency questionnaire Sampled persons: 2-day 24-hour dietary recall of foods and amounts eaten (first day in-person; second day 3-10 days later in-person); dietary supplements (categories); water intake; salt use (specific for certain foods); special diet
OCR for page 35
Improving Data to Analyze Food and Nutrition Policies DHKS CE Obtained from main CSFII None Obtained from main CSFII None
OCR for page 36
Improving Data to Analyze Food and Nutrition Policies Feature NHANES CSFII Person Diet and Health Knowledge None (except for food frequency questions above) Obtained from DHKS NOTE: Information pertains to most recent version of each survey: Integrated NHANES/CSFII (beginning in 2002); 1994-1996 CSFII; 1994-1996 DHKS; current CE. A Flexible Consumer Behavior Survey Module will be fully implemented in NHANES in 2007-2008, with questions drawn from the DHKS and CSFII together with new questions (see text); some parts of the DHKS were asked in NHANES in 2005-2006. and blood chemistry, is obtained. This information is collected when respondents visit mobile examination centers (MECs), which travel around the country to administer the survey. Analysis of blood chemistry includes measures of total cholesterol and HDL cholesterol. Respondents are also asked about physical activities and, beginning in 2003-2004, those aged 6 and over are asked to wear a physical activity monitor for 7 days and return it by mail. With regard to food consumption, respondents are asked to complete a 24-hour dietary recall for two nonsuccessive days; the first recall is conducted in person at the MEC and the second by telephone. Prior to 2002, the dietary recall covered only one 24-hour period, except that NHANES III (1988-1994) collected a second dietary recall in person from a 5-10
OCR for page 42
Improving Data to Analyze Food and Nutrition Policies demands will likely compete with demands for other kinds of health-related information in NHANES. We were not charged to assess those competing demands, but we offer a range of approaches to meet the needs for improved food and nutrition data from NHANES, some of which do not require additional data collection from NHANES respondents directly (for our recommendations, see Chapter 5). We begin by describing efforts that the Economic Research Service (ERS) already has under way to enhance the information on diet and nutrition in NHANES and then discuss five other possible enhancements: special supplements for subsamples, links to food assistance program records, links to geographic information on food outlets, links to neighborhood characteristics, and links to price information. Some of these enhancements pose concerns of confidentiality protection and data access, which we briefly address at the end of the section. ERS Supplement Initiative Questions on household food expenditures and shopping patterns from the CSFII were not part of the original integration of the CSFII with NHANES, nor did that integration include the questions on diet and health knowledge and behavior that were part of the Diet and Health Knowledge Survey supplement to the CSFII. Yet the DHKS questions clearly provided useful information for policy making. Moreover, although the food expenditure and shopping data collected in the CSFII were not very detailed, they provided useful information on where people shopped, where they ate away from home, and how much they generally spent. Such data could be used to understand food stamp (or WIC) purchases and, more broadly, purchases by people who receive public assistance and those who do not. CSFII-type food expenditure data, together with the integrated NHANES dietary intake data, could also be used to understand the relationship between total household spending on food and what each person in the household consumes. Moreover, the health data in NHANES would make it possible to conduct analyses that trace through shopping patterns to food intakes to health outcomes. Participants in a workshop on the integrated NHANES-CSFII called for including the DHKS questions and other improvements in the data related to food and nutrition (Dwyer et al., 2003). In response to the needs that were previously met by content that was part of the CSFII and the DHKS but not integrated with NHANES, the ERS is working with the National Center for Health Statistics to develop a
OCR for page 43
Improving Data to Analyze Food and Nutrition Policies Flexible Consumer Behavior Survey Module (FCBSM). The full module will be included in NHANES in 2007 and 2008 and a scaled-down version in 2005 and 2006. The FCBSM will include questions on food shopping, food expenditures, self-assessment of diet quality, frequency of eating food away from home, attitudes toward and knowledge about diet and food safety, use of food labels, and safety-related preparation practices (e.g., frequency of washing raw foods). The module will take questions from the DHKS and the CSFII; it is also likely to include new or revised questions and questions drawn from other surveys. The intent is for the FCBSM to be a continuing supplement with a set of core questions and some questions that might change from year to year to meet emerging data needs.3 The work on the FCBSM is a positive development toward addressing gaps in data coverage for food and nutrition policy analysis and planning. A careful, thorough research and development program for developing content for the core FCBSM will be important so that the module provides information of most value at least burden on respondents (see Chapter 5). Special Supplements In addition to, or as part of, the Flexible Consumer Behavior Survey Module, supplementary survey questions could be given to subsamples of the full NHANES sample. This approach would be a means to obtain more detailed information on selected topics while not increasing the respondent burden. For example, a subsample could be asked to provide more detailed information on income, assets, food expenditures, food purchasing practices, and participation in food assistance programs. A problem with a subsampling approach is that NHANES is already small for some subgroup analyses (such as subgroups of adults, adolescents, and children) even when pooling 2 or more years of data. One possible approach would be to rotate supplementary topics across years for the entire sample so that, for example, detailed questions on assets might be included in one 2-year cycle and detailed questions on income and program benefits in a subsequent 2-year cycle. 3 Personal communication from James Blaylock, Economic Research Service, USDA, June 22, 2005.
OCR for page 44
Improving Data to Analyze Food and Nutrition Policies Links to Food Assistance Program Records An important set of policy concerns for USDA involves the costs, coverage, and effects on poverty, health, and nutrition of its food assistance and education programs. In recognition of the extensive data needs for analyzing programs, the Food and Nutrition Service and Economic Research Service of USDA have long supported targeted surveys and experiments involving program participants and other low-income households. These data have been used to study many aspects of the Food Stamp Program and other major food assistance programs, such as WIC and school feeding programs (see Hamilton and Rossi, 2002; Logan, Fox, and Lin, 2002; Fox, Hamilton, and Lin, 2004a, 2004b). NHANES, for which the primary focus is monitoring health conditions for the general population, will never be that good a vehicle for analyses that compare participants with eligible nonparticipants in food assistance programs. Moreover, NHANES does not collect sufficient information on income, assets, and expenditures with which to estimate precisely who is and is not eligible for food assistance among the low-income population. Yet NHANES does include data on participation in major food assistance programs, which can provide independent variables to include in econometric analyses to understand factors that relate to better and worse conditions of health and nutrition. However, except for last month’s food stamp benefit, there are no NHANES data on benefit amounts or patterns of receipt, such as whether households participate in more than one program at the same time or different times, spells of participation, or who in the household is covered. Moreover, surveys tend to underestimate program participation by substantial amounts. For example, Cody and Tuttle (2000:21) found underestimates of participation in the Food Stamp Program of 26-37 percent in the March Current Population Survey Income Supplement for 1989-1999. Taeuber et al. (2004) found large differences between reporting of food stamp receipt in the 2001 Supplementary Survey (a predecessor to the new American Community Survey) and in matched records for the state of Maryland, due largely to underreporting by household respondents. A low-cost means to obtain more detailed, accurate information on food assistance program participation in NHANES could be to match administrative records to the NHANES sample and append relevant program variables to the NHANES household and person records. An ERS-
OCR for page 45
Improving Data to Analyze Food and Nutrition Policies sponsored Survey of Food Assistance Information Systems that was conducted by mail in 2002 of program directors in 26 states found that states maintain Food Stamp Program and WIC data systems that are generally updated in real time, but they do not maintain such systems for school feeding programs (Cole, 2003; see also Cole and Lee, 2004). There is considerable experience in linking Food Stamp Program data with other record systems and surveys; linkages of WIC data have rarely been conducted. Although NHANES collects Social Security numbers from sample members, consent has never been sought to use the numbers for record linkage. Until such consent is sought, linking Food Stamp Program records to NHANES records would require probabilistic rather than exact matching, by using such variables as name, date of birth, race, ethnicity, and, possibly, household income. Probabilistic matching would be required for WIC in any case, because, unlike the Food Stamp Program, most WIC programs do not require Social Security numbers from participants. Such matching, using software from the U.S. Census Bureau, was successfully performed in three states for a study of joint participation in food stamps and WIC (Cole and Lee, 2004). Many analyses of food assistance program effects, whether with linked NHANES data or other sources, make use of empirical strategies that rely on instrumental variables to deal with possible bias—variables that are correlated to a potentially endogenous variable (for example, program intake), but are not themselves associated with the outcome of interest (for example, the nutritional health of food assistance program recipients). For example, if unusually effective mothers participate in WIC, then estimates of WIC program effects are likely biased upward; the opposite bias would exist if unusually ineffective mothers participate in WIC. One way to determine the extent and direction of this bias would be to try to use variation in state- and local-level administrative practices in WIC offices. If these practices occur roughly randomly, then variation in administrative practices could be used as instrumental variables for WIC participation. Some information is collected on these practices and has been used in some applications. With greater attention to collecting these types of administrative data, systematic research to enhance understanding of the behavioral effects of WIC might be feasible. More generally, it would be worthwhile to devote systematic effort to obtaining geographic detail on administrative practices that vary across locations to use for analysis of food assistance program effects on diet and health.
OCR for page 46
Improving Data to Analyze Food and Nutrition Policies Links to Geographic Databases on Food Outlets A question of policy interest has been the availability of food shopping outlets that provide a range of reasonably priced, healthy food choices in comparison with the availability and concentration of fast-food and convenience store outlets that may not provide healthy alternatives. The 1996 National Food Stamp Program Survey obtained responses from program participants and other low-income households on the location of stores where they usually shopped and the supermarkets nearest their homes. These addresses were geocoded to latitude and longitudinal coordinates and the resulting information used to calculate distances to the nearest supermarkets and food outlets actually used. The results indicated that most low-income households use supermarkets as their main type of food store and do not typically face barriers to shopping at supermarkets (Ohls et al., 1999:xiii-xiv). ERS proposes to include questions about food shopping habits in the new Flexible Consumer Behavior Survey Module to be added to NHANES. It would be burdensome for the FCBSM to obtain the level of detail in the 1996 survey on locations of specific food outlets used by households. However, with the advances in content availability on the Internet, a possible approach would be to link NHANES records to geographically based information on eating establishments and food retail outlets. This approach would require that the NHANES household addresses be geocoded by using the U.S. Census Bureau’s TIGER (Topologically Integrated Geographic Encoding and Referencing) System or a commercial system based on TIGER. The same geocoding would then need to be done for the addresses of retail food outlets of various types, as well as fast-food establishments, sit-down restaurants, and other away-from-home options in cities or counties in the NHANES sample, using on-line directories and maps. Information might also be added from directories on the price range for eating establishments. The addition and regular updating of geographically based information on food outlets and eating establishments to NHANES records could provide valuable—if admittedly crudely measured—input for analyses of the environmental context of food decision making and changes in that context over time. Links to Neighborhood Characteristics In addition to adding information about nearby food shopping and eating establishments for households in the NHANES sample, it would be
OCR for page 47
Improving Data to Analyze Food and Nutrition Policies useful to add other characteristics of sample households’ neighborhoods to permit contextual analyses of various kinds. Historically, the decennial census long-form sample has provided information on demographic, social, and economic characteristics for counties, cities, and neighborhoods (census tracts and groups of blocks). The 2000 census included a long-form sample of about one-sixth of all households (16 million records), but planning for the 2010 and future censuses assumes that the census will use only a short form with basic demographic information (age, sex, race, ethnicity, household relationship, and housing tenure). The new American Community Survey (ACS), which was fully implemented in early 2005, includes the content of the former census long form. The ACS questionnaire is sent monthly to a sample of 250,000 households, for an annual sample of about 3 million households. To provide estimates of sufficient reliability for areas with fewer than 20,000 people, 5 years’ worth of data will be accumulated and averaged. It would be useful to begin planning now on how to incorporate ACS estimates of neighborhood characteristics into the NHANES database. Such neighborhood characteristics as poverty ratios, educational levels, distributions of housing and utility costs, and ethnic composition would be useful contextual variables to include in analyses of food expenditure and consumption patterns. Because neighborhood-level data from the ACS will not be available until 2011 (for the period 2006-2010), it would be useful to consider incorporating some 2000 census neighborhood characteristics into NHANES. Links to Price Information Ideally, NHANES would provide information not only on food expenditures, but also on prices for specific types of foods for households included in the NHANES sample. Yet it would be very difficult and burdensome to collect information in NHANES about the prices that households pay for food. The aggregate data on food prices that are available through the Bureau of Labor Statistics (BLS) price index program could be used to augment household records in NHANES. BLS publishes Consumer Price Indexes (CPIs) for food at home and food away from home by region (Northeast, Midwest, South, and West), population size classes of metropolitan statistical areas (MSAs), and the 26 largest MSAs. Data are collected, but not published separately, for 87 other geographic areas, not including rural areas; separate indexes for types
OCR for page 48
Improving Data to Analyze Food and Nutrition Policies of foods are provided only at the national level (see www.bls.gov/cpi/cpifaq.htm [June 2005]). These indexes can properly be used only to compare rates of change in prices across areas—not price levels—because the data come from a probability sample of prices that is designed to produce the national CPI and so there is no particular consistency across areas in items that are priced. BLS and the Bureau of Economic Analysis (BEA) have a research program to reanalyze the price data for geographic areas to develop fixed-weight interarea price indexes for major commodities that can be used to compare relative costs across areas (not just relative rates of change in prices). The approach uses hedonic regression methods to determine the contributions of geographic locations to the prices of various items (see Aten, 2005; Kokoski, Moulton, and Zieschang, 1999). ERS could investigate the geographic area pricing work by BLS and BEA to determine if appropriate relative price indexes for food at home and away from home for specific metropolitan areas are available that could be added to household records in the NHANES database. If so, the indexes could be used to study such issues as whether consumption of fruit and vegetables is related to areas with generally higher food prices. The analytical uses of area food price indexes would be limited for several reasons. One is that they lack specificity by type of food, although it is possible that the BLS/BEA work could be disaggregated for some types of foods. Another is that their use would necessarily involve an assumption that food prices differ more among metropolitan areas than within them, which may not hold true in some or most areas. The results in Aten (2005) for food and beverages show a considerable range in price index values among areas—from 0.85 (Cincinnati) to 1.29 (New York City), with the mean value set at 1.00. This range (0.44) is higher than the price index value range for transportation (0.29), about the same as the price index value ranges for recreation and apparel, and lower than the price index value ranges for housing (0.79) and out-of-pocket medical care (1.38). While not addressing the issue of intra-area variability in food prices, for which data are not readily available, the BLS/BEA work suggests that the addition of area price indexes for food to NHANES records could be useful for some kinds of research. It would be a very low-cost enhancement to the data. Another possible avenue to explore for adding food price information to NHANES involves sales outlet and household scanner data collected by private market research firms. If such data could be obtained at the city or
OCR for page 49
Improving Data to Analyze Food and Nutrition Policies neighborhood level and appended to NHANES records, the potential for understanding food consumption behavior would be greatly expanded. Scanner datasets have serious limitations, their quality is largely unknown, and making them usable for analysis purposes could be difficult. Nonetheless, scanner data are unparalleled in the detail they provide on prices for specific foods (see Chapter 3). Confidentiality and Data Access Many of the enhancements to NHANES that we outline in this section represent low-cost improvements for research on understanding household food expenditures and consumption by linking NHANES records with other data sources. Such data linkages, however, may increase the risks that individual households could be reidentified in public-use microdata files. Because of the small sample size and clustered design for the continuing NHANES, in which only 15 primary sampling units are in each year’s sample, special precautions are already taken to safeguard confidentiality: data are not released for single years but instead are pooled over 2 years, data masking steps are used on specific variables, and some variables (for example, age of nonsampled household members) are not included in any form on the public-use microdata files. Because of the 2-year pooling procedure, food and nutrition-related data that began to be collected in 2002 when the CSFII was integrated with NHANES are not yet available. Such items include the second day of dietary intake information and some items included for the first time in the first day of dietary intake, such as where each food was obtained; these items will first be publicly available in the 2003-2004 NHANES release. The National Center for Health Statistics has two mechanisms for providing user access to data (such as geographic detail and day two dietary intake in 2002) that are not provided on public-use microdata files or that are altered in some way. Researchers may be approved to access those data at the NCHS research data center (RDC) at its headquarters in Hyattsville, Maryland, or remotely by e-mail, submitting SAS code to produce tables or regression coefficients, for example. NCHS staff review output from either the RDC or remote access submissions for confidentiality protection. There is a fee for using the RDC or the monitored remote access service; the fee for monitored remote access is reduced for users who plan repeated analyses of selected datasets that have been developed for frequent, multiple use (for the details of the NCHS policy, see www.cdc.gov/nchs/R&D/rdcfr.htm
OCR for page 50
Improving Data to Analyze Food and Nutrition Policies [June 2005]). USDA could consider working with NCHS to develop low-cost access to NHANES records linked to other data at the NCHS RDC or by monitored remote access, perhaps developing special extracts oriented to food consumption analysis. CONSUMER EXPENDITURE SURVEY The Consumer Expenditure Survey (CE), which has been conducted on a continuing basis by BLS since 1980 to monitor the purchasing activities and habits of American consumers, is the most comprehensive source of expenditure data collected by the federal statistical system. (Predecessor surveys were conducted in 1972-1973, 1960-1961, and at intervals of about 10-15 years back to 1901; see Jacobs and Shipp, 1990.) CE data are used to estimate market basket weights of goods purchased for the Consumer Price Index. The CE is comprised of two surveys: (1) a diary survey, in which households keep two weekly expenditure diaries; and (2) a quarterly household interview survey, which obtains information for households on major purchases on a quarterly basis over the course of a year together with demographic, employment, and income information for the household members (Bureau of Labor Statistics, 2004b). The diary survey is a record of daily expenses for a consumer unit (members of a household that share living expenses) that is kept by a respondent from each consumer unit for two consecutive 7-day periods. About 7,500 consumer units are surveyed each year. All daily expenses, except business expenses and expenses incurred while out of the home overnight, are included in the diary. The diary survey also collects demographic, work experience, and earnings data on household members aged 14 and over together with household income by source in the last 12 months. Food-related items in the diary survey include usual weekly expenditures at supermarkets and specialty food stores together with detailed information on food purchases in each 1-week diary. The diary provides space for food purchases away from home by type of outlet and meal (for example, breakfast or lunch from a fast food outlet, vending machine, or full-service restaurant) and food purchases for consumption at home by type of food (for example, grain products, beef, sugars, vegetables—see Table 2-1). For each purchase, the respondent is asked to record the item, its cost, and some information about the form of the item (for example, canned, frozen, fresh). The quarterly household interview survey collects information from
OCR for page 51
Improving Data to Analyze Food and Nutrition Policies about 7,500 consumer units every quarter for five consecutive quarters on expenses over the 3 months prior to the survey. It is designed to capture information about major purchases—including vehicles, major appliances, housing costs, and vacation costs—that are not generally available through the diary survey. This survey also collects information on housing characteristics, household appliances, ownership of real estate, work experience of household members, sources and amounts of household income, and information on financial assets, such as savings accounts, stocks, bonds, and mutual funds. The household interview survey includes questions to double-check food purchases reported in the diary portion of the survey, including questions about food purchased away from home. Data from the two surveys are integrated to provide information about detailed day-to-day purchases, as well as long-term, major purchases. Uses and Limitations CE data have been used for analyses related to food and nutrition: for example, expenditures on fruits and vegetables by low-income households (Blisard, Stewart, and Jolliffe, 2004); household expenditures on vitamins and minerals (Lino et al., 1999); trends in food purchases away from home (Paulin, 1995); and factors related to food expenditures for use in projections (Blisard, Variyam, and Cromartie, 2003). The primary advantage of the CE is its rich data on all types of expenditures; in addition, it obtains relatively rich data on household employment and income. However, the CE also has disadvantages for food and nutrition-related analysis. With regard to food consumption, the CE does not obtain information on actual dietary intake. Moreover, the food expenditure information in the CE does not always well describe the foods purchased or their quantity, so that the CE cannot be used as a source for prices paid for specific amounts of individual food items. For example, a survey respondent may report purchasing milk, but may not report what kind of milk—whole, low fat, or skim. The respondent may at the same time report purchasing milk for $2.19, but information on how much milk was purchased is not systematically collected. The CE also does not obtain information on who in the household consumed the food or how it was prepared. The CE’s relatively small sample size further limits its potential for analysis. The CE data on expenditures and income also exhibit underreporting problems: in comparison with BEA’s personal consumption expenditures series (PCE),
OCR for page 52
Improving Data to Analyze Food and Nutrition Policies the CE underreports spending in many expenditure categories. In particular, the aggregate CE amount for food for 1992-2000 averaged 72 percent of the corresponding PCE aggregate (see Garner et al., 2003). Income is also underreported, particularly for low-income families for whom reported expenditures often exceed reported income (Meyer and Sullivan, 2004). Possible Improvements Enhancements to the Consumer Expenditure Survey similar to those we suggest above for the integrated NHANES could extend the usefulness of the CE data, particularly the diary survey, for food and nutrition-related research and policy analysis. For example, reporting of Food Stamp Program participation could be validated and enhanced by matching CE records with the program’s administrative records. Neighborhood characteristics from the 2000 census and the American Community Survey could be added to the CE records, as could links to geographically based information on retail food outlets. Although the CE is one of the most burdensome federal surveys, it might be possible to occasionally include supplemental questions related to food purchasing and consumption behavior that would enhance the value of the data for food and nutrition-related research. To the extent that these initiatives increase the possible risks of disclosing confidential information, researchers would need to access the data at the BLS research data center at its headquarters in Washington, DC. It might also be possible to arrange for researchers to access enhanced CE microdata files on their own computers through a licensing agreement (see National Research Council, 2005a).
Representative terms from entire chapter: