Methodological Issues for Future Consideration
The first phase of the panel’s work identified parts of the estimation methodology for which improvements could be made. The panel has not had the time to fully consider what improvements should be made but plans to in Phase II. Possible improvements to four of the methodological issues reviewed in this report will be considered in Phase II. These include consideration of: a weighting scheme for the Current Population Survey (CPS) data that is appropriate for WIC age groupings; methods for estimating the number of people adjunctively eligible for WIC; data and methods for estimating the prevalence of nutritional risk; and methods for estimating the number of people who are eligible for WIC who will participate in WIC.
The number of infants estimated in the CPS is understated compared with yearly Census Bureau population estimates, as Chapter 4 details, because the CPS weights for nonwhite infants are not controlled to population totals for that age group (0 to 1 year). As a result, the number of income eligible infants is probably understated using the CPS, and hence the number of pregnant and postpartum women is also likely to be understated since that number is based on the estimate of the number of income eligible infants. In Phase II, the panel will explore the use of revised weights that are more appropriate for WIC age groupings and other alternative weighting schemes.
The panel concluded in Chapter 4 that accounting for WIC eligibility through TANF, the Food Stamp Program, and Medicaid has a large effect
on estimates of the number of WIC eligible infants and children. Current methods make only a minor adjustment for those who are adjunctively eligible for WIC that does not fully account for all who are eligible for WIC. Thus, the estimates of WIC eligibility are substantially understated. A priority for the panel’s Phase II is to explore alternative ways to estimate the number of people who are adjunctively eligible for WIC.
Current methods for adjusting the estimates of the number of income eligible persons for the prevalence of nutritional risk are based on old data about nutritional risk prevalence. More recent estimates have been made, but there are reasons to believe they may be flawed. The panel recommends that estimates of nutritional risk should be reexamined. In Phase II, the panel will consider alternative data and methods for estimating the prevalence of nutritional risk.
The panel reviewed current methods for estimating the number of income eligible persons who would participate in WIC. We conclude that the current method of using food stamp participation rates as a proxy for WIC participation is problematic and that new methods should be considered. A priority for Phase II is to more fully consider different methods to estimate participation in the WIC program.
The panel plans to explore a number of new topics in Phase II: use of alternative datasets for the core estimates, estimating eligibility in the U.S. territories, methods for estimating eligible pregnant women, and methods for estimating breastfeeding rates in order to estimate the number of eligible postpartum women.
ALTERNATIVE DATASETS FOR ESTIMATING INCOME ELIGIBILITY
The March Income Supplement of the CPS is currently used to estimate the core number of persons who are income eligible for the WIC program. However, other data sets have certain features that may make them better-suited for estimating WIC eligibility. The Survey of Income and Program Participation (SIPP) is the primary example, since it collects monthly income data and monthly data on demographic and household composition. The panel will extend the work of the Food and Nutrition Service (FNS) publication (U.S. Department of Agriculture, 1999a) regarding the strengths and limitations of the CPS and the SIPP for WIC eligibility estimation purposes.
An alternative strategy for obtaining WIC eligibility estimates may be
to add supplements or single questions (for example, a question on WIC participation) to the March CPS that would provide information needed to estimate the various components of eligibility for WIC. Another alternative strategy may be to conduct a small special-purpose survey to collect data needed to assess WIC eligibility. Phase II of the panel will consider such alternative sources of data.
ESTIMATES OF THE NUMBER OF ELIGIBLE PEOPLE IN THE TERRITORIES
To estimate the number of income eligible infants and children residing in the U.S. territories of American Samoa, Guam, Puerto Rico, and the American Virgin Islands who are eligible to receive WIC, FNS employs a constant multiplier of 1.0388 to adjust the estimates derived from the CPS, since the CPS universe does not include the territories. This proportional adjustment was estimated from the 1990 census. During the second phase of the panel’s work, we intend to examine the validity of this assumption by examining the historical trends in enrollment in WIC in the United States versus the territories over the 1990s.
ESTIMATION OF ELIGIBLE PREGNANT WOMEN
FNS estimates the number of income eligible pregnant women based on the number of income eligible infants. The only adjustment that is made to the number of infants is to multiply the count by 0.75. This assumes that the number of income eligible infants in 9 months of a year is equal to three-quarters of the number of income eligible infants in one year and that the number of income eligible pregnant mothers is exactly the number of income eligible infants in a 9-month period. Although a pregnant woman is eligible as soon as she is pregnant, there is usually a delay between the time a mother conceives and the time she realizes she is pregnant, and also a lag between the time a woman finds out she is pregnant and the time she decides to apply for WIC. The current methodology does not take either of these lags into account. Use of this assumption may result in an overstatement of the number of pregnant women who participate in WIC, although technically not the number who are eligible for WIC.
The estimation methodology also assumes that the number of infant deaths and the number of multiple births cancel each other out (although
the estimates of postpartum women do adjust for multiple births and infant deaths). Using the counts of infants to count pregnant women without accounting for infant deaths would understate the number of pregnant women. The presence of multiple births would overstate the number of women. Estimates from the late 1980s in the second WIC Evaluation Study indicate that multiple births may be more common than infant deaths (U.S. Department of Agriculture, 1987). The panel has not, however, reviewed this study nor data about infant deaths and multiple births to assess the appropriateness of this assumption.
The method for estimating income eligible pregnant women also assumes that income during pregnancy is similar to income after the birth of a child, as we discussed above. It also assumes that birth rates do not change over the time frame between when the estimates of infants are made and the 9 months prior to that. Finally, the census definition of families used by the CPS does not count cohabitating partners of pregnant women as part of the family unit until the baby is born. If, in assessing the eligibility of families, WIC staff workers do count cohabitating partners as part of the family unit, a bias in the CPS estimates of eligible pregnant women could be created. Further explorations into these methodological assumptions will be conducted in Phase II.
ESTIMATION OF ELIGIBLE POSTPARTUM WOMEN
Estimates of the number of income eligible postpartum women are also based on the estimates of the number of income eligible infants, plus an adjustment to account for the percentage of postpartum women who breastfeed their infants and the duration of breastfeeding. Although the panel has not thoroughly examined methods used in these estimations, some methodological and data issues deserve further attention.
The accuracy of adjustments to account for the rate and duration of breastfeeding among low-income mothers is one such issue. The National Maternal and Infant Health Survey (NMIHS) data, which are used to estimate breastfeeding rates and duration, are 13 years old, and some evidence indicates that breastfeeding rates have increased since then. Data from the Ross Laboratories Mothers Survey indicates that breastfeeding rates in the U.S. population have increased from 54.2 percent in 1988 to 68.4 percent in 2000 for mothers in the hospital after delivery of their child and from 19.5 percent in 1988 to 31.4 percent in 2000 for women 6 months after
the birth of their child (Smith, 2001).1 Changes in breastfeeding rates for low-income groups have historically lagged behind those of higher income groups (see U.S. Department of Agriculture, 1992, for a summary of historical breastfeeding practices), and we do not know whether breastfeeding rates among low-income groups have increased over recent years as much as rates in the total population have. Results from the Ross Laboratories Mothers Survey should be corroborated with data from other sources, which do not provide trends in breastfeeding but do give point-in-time estimates that are fairly recent, such as the National Health and Nutrition Examination Survey (NHANES III), the National Survey of Family Growth, the Continuing Survey of Food Intakes by Individuals, the Early Childhood Longitudinal Study Birth Cohort 2000, and the Infant Feeding Practice Survey. These datasets also include measures of family income, which the Ross Laboratories Mothers Survey data do not include. The panel has not reviewed the methods and assumptions used in the FNS life table estimates of the probability of breastfeeding over time. In our further review, we will consider these.
CONFIDENCE INTERVALS FOR ELIGIBILITY ESTIMATES
Sampling variability and random errors in reporting of income can create uncertainty in the estimates apart from the systematic biases that have been examined in this report. In the second phase of the study, the panel will investigate the appropriate level of confidence to place in the estimates by examining standard errors of the estimates of eligibility.