Click for next page ( 170


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 169
APPENDIX A Data Sources and Coverage Issues This report considers the use of the Current Population Survey (CPS) or the Survey of Income and Program Participation (SIPP) to make na- tional estimates of the number of people eligible and likely to participate in WIC. In this appendix, we present details on each of these surveys. The appendix also addresses two issues of survey coverage that were raised in response to the panel's interim report (National Research Council, 20011: the coverage of undocumented immigrants and the coverage of military populations. DATA SOURCES The March CPS is the data set currently used to estimate WIC eligibil- ity and participation. While the CPS has many advantages, it does have limitations. Still, there are few viable options to the CPS. The most promis- ing alternative, SIPP, has several advantages over the CPS, but it too has limitations. This section provides background information on both of these surveys and points out their advantages and limitations for the specific purpose of estimating WIC eligibility and participation. Neither SIPP nor the CPS collect data on nutritional risk or breastEeeding status. Thus, information on the prevalence of these two con- ditions in the income-eligible population must be collected from other data sources and a multiplier used to adjust for them. Alternative data sources 169

OCR for page 169
170 APPENDIXA for making estimates of the prevalence of nutritional risk and the preva- lence and duration of breastSeeding among income-eligible populations are discussed in Chapter 6 (breastSeeding status) and Chapter 7 (nutritional risk). March CPS The March CPS is an annual supplemental survey to the monthly CPS. In addition to the information collected in the monthly survey, the March CPS collects data on household income, participation in federal programs, and demographic characteristics of the household. The March CPS is a cross-sectional household survey of the civilian noninstitution- alized population of the United States. Over 60,000 housing units are sampled. Since the CPS is a cross-sectional survey, individuals and their charac- teristics are observed at only one point in time. This has implications for estimating the number of individuals in each of the WIC eligibility catego- ries. Data are collected on the ages of individuals in the household, which means that it is possible to directly infer the number of infants up to age 1 and the number of children age 1 through 4. It is also possible to observe the number of women less than 1 year postpartum by observing the num- ber of infants less than a year old and identifying their mothers in the householder Because the survey does not, however, ask women about their pregnancy status, it is not possible to directly observe the number of preg- nant women. The survey does not collect data on breastSeeding status ei- ther, so it is not possible to separate postpartum women into breastSeeding and nonbreastSeeding status. The March CPS collects detailed information about annual income for the previous year, covering all the sources needed to determine income eligibility for WIC. The CPS does not collect monthly income data. The CPS also collects information on household participation in programs used to confer adjunctive eligibility for WIC, including Temporary Assistance for Needy Families (TANF), food stamps, Medicaid, and the State Children's Health Insurance Program (SCHIP). The CPS recently added questions on WIC receipt in the household in lit is not possible to observe postpartum women who do not live with their infants. These women are categorically eligible for WIC for up to 6 months as nonbreastfeeding postpartum women.

OCR for page 169
DATA SO URCES AND COVERAGE ISSUES 171 the past year, which is relevant to the estimation of WIC participation rates. However, the questionnaire does not ask who in the household re- ceives WIC benefits, or when or how often WIC benefits are received. An income screen is used to determine which respondents are asked about WIC receipt. This screen is sufficiently low that some people above it may be eligible for WICparticularly those who are enrolled in Medicaid but who have income above the WIC threshold and yet would not be asked the questions about the receipt of WIC benefits. CPS coverage of the U.S. population in general is lower than that of the decennial census but still quite high. However, coverage of some mi- nority groups is not as complete. For example, the overall coverage rate for the March 2001 CPS (the ratio of March CPS counts to 2000 census counts) is about 92 percent, but the coverage rate for black females ages 20-29 in 2001 was 81 percent (U.S. Census Bureau, 2002~. Coverage of illegal immigrants in the CPS is a special concern for some states with large immigrant populations because illegal immigrants are eligible for WIC. When compared with independent estimates of the number of undocu- mented immigrants, the CPS counts match closely (Passel, 2001~. How- ever, the CPS population weights are based on updated 1990 population projections that appear to severely underestimate the number of undocu- mented immigrants (Passel, 2001~. As a result, until CPS weights are based on updated population projections, the CPS probably will underestimate the number of undocumented immigrants. Nonresponse rates for the March CPS are generally low. In 2001, the nonresponse rate for the March supplement was 8.5 percent, which, when added to the nonresponse rate for the monthly survey, results in a total nonresponse rate of 15.9 percent. A key advantage of the March CPS for estimating WIC eligibility and participation is that it is conducted and released on a regular and timely basis. Every year the March supplement is conducted in March and data are released in the fall of the same year. Survey of Income and Program Participation SIPP is a longitudinal household survey representative of the U.S. ci- vilian noninstitutionalized population. The first SIPP panel was interviewed in 1984, and a new panel was introduced each year thereafter until 1993. Since 1993, a new panel was introduced in 1996 and again in 2001. Panel sample size ranges from 14,000-37,000 households. Panels are followed

OCR for page 169
172 APPENDIXA for two and a half to four years. Households in the panel are interviewed every four months. Because SIPP is longitudinal, it is possible, by construction, to directly estimate the number of pregnant and postpartum women by observing infants in the sample The survey does not collect information on the breastSeeding status of postpartum women. The SIPP collects detailed information on each household's income, including all the sources of income used to determine WIC eligibility. A big advantage of SIPP is that it collects monthly income. ~ , . Collecting information on participation in federal programs is a major component of SIPP's mission. Data about AFDC/TANF, food stamps, and Medicaid are collected in SIPP on a monthly basis. As in the CPS, underreporting of program participation is a problem in SIPP. SIPP inter- views respondents every four months and uses a four-month recall period. Under this format, for some items in the survey, respondents tend to report the same thing for all four months of a recall period. For example, a respon- dent may report four months of food stamps receipt and then report four months of nonreceipt for the next interview period. This so-called seam bias may be an indication of response error (see National Research Council, 1993 for a more thorough documentation of this problem). SIPP collects data on WIC participation. Unlike the CPS, it does not use an income screen in asking these questions, and it collects information about which household members receive WIC. Coverage of the U.S. population in general in SIPP is comparable to that of the CPS. There are no estimates of how well SIPP covers undocu- mented immigrants. However, SIPP weights are also based on 1990-based population projections that appear to underestimate the number of un- documented immigrants. Response rates for each interview wave are generally high in SIPP. How- ever, cumulative attrition over the waves of the interviews can be severe. By the eighth wave of the 1996 panel, 31 percent of the original panel were nonrespondents (U.S. Census Bureau, 19981. Evidence from previous pan- els suggests that attrition is more likely to occur among young adults, males, minority groups, never-married people, people with incomes below the poverty level, and people with low educational attainment (Lames et al., 2Like the CPS, it iS not possible to observe postpartum women who do not live with their infants.

OCR for page 169
DATA SO URCES AND COVERAGE ISSUES 173 19941. To observe 9 months of pregnancy and 1 year of the postpartum period, several waves of information on a household are needed. Differen- tial attrition of WIC-eligible populations could result in an underestimate of eligibility. The longitudinal nature of SIPP means that data production is a more complicated process than that required for the CPS. As a result, it can be one, two, or more years after a panel is interviewed before the data are released. Moreover, SIPP data files are not released on a regular schedule as the CPS estimates are. SPECIAL POPULATIONS COVERAGE ISSUES The accuracy of the estimates of WIC eligibility depend on the quality of survey data used to make the estimates and how well those surveys cover the WIC-eligible population. Some populations served by WIC (e.g., low- income populations, immigrants) may be underrepresented because of nonresponse, meaning that estimates based on these data, if not adjusted for characteristics that make one not respond, may be biased. Other popu- lations (e.g., families of military personnel who live on base) may not be part of the sampling universe of the survey and not represented in estimates of WIC eligibility. This section briefly examines such coverage issues in SIPP and the CPS. Responding to concerns raised about underestimates of WIC-eligible people in certain states, we specifically focus on coverage of undocumented immigrants and of military personnel and their families. Immigrants are eligible for WIC regardless of whether they are documented or not. WIC has no special rules for military personnel and their family members living in the United States. We also briefly discuss estimates of the number of eligible people in the U.S. territories of Puerto Rico, American Samoa, Guam, and the Virgin Islands. Both the CPS and SIPP are designed to represent the U.S. civilian noninstitutionalized population. In theory, this means that both surveys should cover immigrants regardless of their documented status. However, reasons possibly leading to undercoverage of undocumented immigrants include language barriers, migration, and avoidance of interviewers from an agency of the federal government because of perceived fears of being turned in to the Immigration and Naturalization Service (INS).3 SINS is now the Bureau of Citizenship and Immigration Services in the Department of Homeland Security.

OCR for page 169
174 APPENDIXA Coverage of Undocumented Immigrants Estimates of the number of undocumented immigrants in the United States have received considerable attention recently because of significant discrepancies in estimates of the 2000 U.S. population among three data sources. The 2000 census provided one estimate. The Accuracy and Cover- age Evaluation Survey (ACE), which is a survey conducted shortly after the 2000 census based on a sample of the population, provided the second estimate. The final estimate used population projections based on demo- graphic techniques, called demographic analysis (DA), takes the base popu- lation projections from the census year 1990, adds births and subtracts deaths from birth and death records, and adds net immigration from INS administrative records. The estimated population of the United States in 2000 based on DA was 279.6 million. Estimates based on the 2000 census were 281.4 million, and estimates based on the ACE were 284.7 million. Undercounts of the number of undocumented immigrants in DA were considered as a possible explanation for the relatively low estimates from this data source. To examine estimates of the number of undocumented immigrants, Passel (2001) compared March 2000 CPS survey data with data from the Census 2000 Supplementary Survey (C2SS) an independent sample sur- vey of 700,000 households conducted at the same time as the 2000 census. C2SS includes questions on country of origin, citizenship, and year of im- migration to the United States if foreign-born (the census short-form does not include these questions, so it is not possible to estimate the number of undocumented immigrants with census data). Passel's work with the CPS provides an opportunity to assess how estimates of the number of undocu- mented immigrants from the CPS compare with other estimates from re- cent sources of data. Two estimates of the number of undocumented immigrants were made based on the CPS data. One used a weighting scheme controlled to the CPS population control totals, and the other used a weighting scheme con- trolled to 2000 census data. The second weighting scheme was used to better match the CPS population (civilian noninstitutionalized population) to the 2000 census population (all persons in the United States), which is the population represented in the C2SS survey. The CPS estimate of undocumented immigrants based on CPS weights was 6.6 million close to DA estimates of 6 million, but still 11 percent higher. The CPS population controls are based on DA-like projections.

OCR for page 169
DATA SO URCES AND COVERAGE ISSUES 175 The estimates of undocumented immigrants in the CPS weighted to the 2000 census population control totals (8.3 million) were very close to the estimate of undocumented immigrants in the C2SS survey. The weighting scheme used tried to match the CPS population (civilian noninstitution- alized population) to the 2000 census population (all persons in the United States), which is the population represented in the C2SS survey (Passer 2001~. CPS estimates of undocumented immigrants weighted to the 2000 census population controls compare closely to estimates from the larger C2SS data source. Current and recent years' CPS weights are still based on 1990 census-based control totals, which are constructed in a manner that is quite similar to the DA estimates. Thus, as long as the CPS weights are still based on 1990 census estimates, the CPS will undercount undocumented immigrants. After the CPS is redesigned based on 2000 census results (which will probably happen in 2003), it appears that CPS estimates of undocumented immigrants will at least be comparable to 2000 level esti- mates. Estimates of undocumented immigrants from the SIPP are not avail- able, although it would be possible to make them. The latest waves of SIPP data are also controlled to 1990 census-based populations, so any underes- timate of undocumented immigrants due to recent immigration through- out the 1990s is likely to be manifest in SIPP as well. The longitudinal nature of SIPP requires reinterviewing survey respondents. If immigrant populations are harder to interview initially, they are also likely to be harder to reinterview. It is thus possible that disproportionate attrition could oc- cur for immigrant populations, which could add an additional complica- tion to immigrant coverage in the SIPP. Furthermore, immigrants who enter the country after a SIPP panel begins would not be covered by the data. Coverage of Military Personnel Members of the military who live on a military base are not included in the sampling universe of the CPS or SIPP; however, their family mem- bers who live off base are included. Military members who live off base are also included in the sampling universe of these two surveys. According to data from the 2000 Survey of Active Duty Personnel, over half of military personnel with children live off base and are therefore part of the sampling

OCR for page 169
176 APPENDIXA frame for the CPS and SIPP.4 In 2000, 57 percent of infants and 56 percent of children with at least one parent in the military stationed in the United States lived off base.5 Thus, more than half of those in military families are covered in the CPS-based WIC eligibility estimates. WIC eligible family members of military personnel who live on base are not fully accounted for in the CPS and SIPP sampling frames and therefore are not included in estimates of eligibility for WIC. The resulting number of potentially missed income-eligible infants and children is small since the total number of infants and children living on base is small (at most 31,855 infants and 144,876 children ages 1 to 5 in 20001. Another complication in estimating the number of military family members eligible for WIC concerns how sources of income are counted toward eligibility determination and what income data are collected on the CPS. Housing allowances received because a family member is enlisted in the military may be counted as income when the income eligibility of a WIC applicant is being determined. The CPS March Income Supplement does not specifically ask for income from housing allowances or other ben- efits specific to military personnel. Therefore, CPS or SIPP estimates of incomes of military personnel would tend to overestimate eligibility for WIC. It is not clear how large such an overestimate may be, but it would affect estimates of off-base military personnel, since the CPS and SIPP cover only off-base military families. Since WIC income eligibility guide- lines are flexible in terms of reporting income, it is not clear that the hous- ing allowance is consistently counted toward an applicant's income in WIC offices anyway. If this is the case, then the overestimation of income-eli- gible women, infants, and children with family members in the military due to income accounting differences may be small. General Coverage of Low Income and WIC-Eligible Populations A general question for using the CPS or SIPP to estimate WIC eligibil- ity is how well WIC-eligible populations are covered. There are no direct 4The survey, which is conducted by the Defense Manpower Data Center, samples about 60,000 active-duty members of the Army, Navy, Air Force, Marine Corps, and Coast Guard. It is a mail survey that is conducted approximately every two years. 5There is a new WIC program for families of military personnel who are stationed overseas, but funds for this program are provided through the Department of Defense, so eligible families do not need to be accented for in the USDA eligibility estimates.

OCR for page 169
DATA SO URCES AND COVERAGE ISSUES 177 measures of how well low-income populations are covered in these surveys, but both surveys undercover minority populations, which tend to have lower incomes, to a similar degree (Kalton, 1998~. Surveys with control totals based on 1990 census data do adjust for the undercount of minorities in the 1990 census. However, these adjustments do not necessarily fully account for an undercount of low-income populations. When the CPS and SIPP are redesigned based on 2000 census population totals, no adjust- ments for an undercount of the U.S. population will be made. The net undercount of the 2000 census population by race was not as great as that of the 1990 census. ESTIMATES OF THE NUMBER OF ELIGIBLE PEOPLE IN THE TERRITORIES Since neither the CPS or SIPP universes include the U.S. territories, to estimate the number of income-eligible infants and children residing in American Samoa, Guam, Puerto Rico, and the U.S. Virgin Islands who are eligible to receive WIC, the Food and Nutrition Service employs a constant multiplier of 1.0388 to adjust the estimates derived from the CPS. This proportional adjustment was estimated from the 1990 census. As a crude check on this method, Table As L uses WIC administrative data on the number of participants in the territories and in the United States and shows the ratio of participants in the territories to participants in the United States. From 1996 to 2000, the ratio was consistently 3.2 per- cent. These calculations indicate that the 3.88 percent adjustment for eligi- bility in the territories is in the right ballpark, at least in terms of partici- pants. However, recent data from the 2000 census long form include the U.S. territories and should be used to update this adjustment factor. Throughout the decade between censuses, a crude multiplication factor TABLE A-1 Ratio of WIC Participants from the U.S. Territories to WIC Participants from United States: 1996-2000 Number of Participant 1996 1997 1998 1999 2000 Total territories 222,596 230,421 225,806 224,829 233,458 United States 6,965,235 7,176,445 7,141,591 7,086,377 6,964,801 Ratio 0.032 0.032 0.032 0.032 0.034

OCR for page 169
178 APPENDIXA could be derived annually from data on pregnant women participating in WIC by using similar ratios as those produced in Table A-1. The ratio of WIC participants in the territories to the total number of WIC participants (United States plus the territories) could be added to the total number of WIC income-eligible persons by category. This adjustment factor could be based on any difference in WIC participation rates between the United States and its territories. These participation rates could be computed for decennial census years, when a measure of the income-eligible women in the territories is available.