Click for next page ( 43


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 42
4 Categorical Eligibility of Infants and Children The first step both in eligibility determination and in estimating eligi- bility is to determine if an individual meets the categorical eligibility re- quirements, that is, whether the applicant is an infant, child, pregnant woman, breastSeeding postpartum woman, or nonbreastSeeding postpar- tum woman. As mentioned earlier, the current method for estimating eligi- bility for WIC uses the March Demographic Supplement of the Current Population Survey (CPS) to obtain a count of infants ages O through 12 months and the number of children ages 1 through 4 years. Since the num- bers of pregnant and postpartum women cannot be directly observed in the CPS, estimates of the number of women in these categories are based on estimates of the number of infants in the CPS. Thus, the accuracy of the estimate of the number of income-eligible infants is especially important in . . the estimation process. The panel's Phase I report found an undercount of infants in the CPS (National Research Council, 2001: Table 3-11. This undercount ranged between 1.0 and 4.1 percent in 1992-2000, and averaged over 2 percent a year. The CPS underestimated the number of children in 1992 and 1993 but overestimated that number in the years 1994-2000. In this chapter, we consider why the CPS estimates undercount infants and outline a procedure that USDA can use to adjust the CPS estimates to more accurately estimate the numbers of infants and children. The chapter concludes with a general discussion of how the Survey of Income and Pro- 42

OCR for page 42
CATEGORICAL ELIGIBILITY OFINFANTS AND CHILDREN 43 gram Participation (SIPP), an alternative data set for estimating WIC eligi- bility, compares to the CPS in estimating the number of infants and chil- dren. EXPLANATION FOR THE UNDERCOUNT OF INFANTS In its Phase I report, the panel hypothesized that the underestimation of infants is at least partially due to the age groupings for black and other race individuals used in the development of posts/ratification weights (Na- tional Research Council, 20011. CPS-based estimates of the numbers of persons less than 1 year old and 1 through 4 years are determined largely by the sampling weights. The last step in the development of CPS weights, sometimes referred to as a posts/ratification or population control adjust- ment, compares CPS estimates with the available Census Bureau popula- tion projections by age, gender, and race. These projections are derived through a month-by-month adjustment of decennial census counts that adds births, subtracts deaths (both births and deaths statistics come from vital records), accounts for net migration, and corrects for the decennial census undercount. The adjustment procedure is completed using total population projections available at the time of final CPS weight construc- tion. ~ CPS posts/ratification adjustments are done in age, gender, and race subgroups of sufficient sample size to yield adjustment factors that are stable. For infants and children, adjustments are made for each gender by single year of age for white respondents, two-year age groups for black respondents (0-1, 2-3, and 4-5 years) and a five-year age grouping (0-5 years) for other race respondents. These adjustment groups provide the larger sample sizes needed to obtain more stable values of the population control adjustments by gender. Because the numbers of black and other race infants are not required The projection counts may, at the time of the CPS adjustment, not be based on the estimated number of births or deaths, since vital statistics used to provide final projections are not available until two to three years after the projection year. As vital statistics data become available, the Census Bureau releases new projections on a continuing basis for each year and month throughout the course of a decade. Thus, estimates of population counts derived from CPS weights do not agree with more recent Census Bureau population projec- tions for March of a given year.

OCR for page 42
44 ESTI~TING ELIGIBILI>~D PAR TICIPATION FOR THE ~CPROG~ to match control totals for single-year age intervals, the estimated number of infants overall from the CPS does not match the number of infants in the control totals. The underestimation of infants may also be partly due to reporting error in the CPS respondents may tend to push up the age of infants to 1 year when they are really only 10 months old. Or it may be the result of other unknown factors. The panel examined the nature of the differences between CPS esti- mates and Census Bureau projections for single years of age for each of the three racial groups used in the adjustment process. Table 4-1 presents a comparison ofthe March CPS estimates and census projections ofthe num- ber of children in single-year age intervals for each racial group for the year 2000. The pattern of the ratios, repeated across years, clearly shows that during the decade, CPS estimates substantially underestimate infants for black and other races and modestly underestimate the number of infants TABLE 4-1 Census Projections and CPS Estimates by Single-Year Age Groups, March 2000 White Black Other Total Age 0 Census projection 3,103,504 623,345 235,485 3,962,334 CPS estimate 3,102,955 560,460 204,477 3,867,892 Ratio of projection to estimate 1.00 1.11 1.15 1.02 Age 1 Census projection 3,092,302 604,020 228,588 3,924,910 CPS estimate 3,091,779 596,446 272,324 3,960,549 Ratio of projection to estimate 1.00 1.01 0.84 0.99 Age 2 Census projection 3,059,341 593,856 231,388 3,884,585 CPS estimate 3,058,792 611,779 213,566 3,884,137 Ratio of projection to estimate 1.00 0.97 1.08 1.00 Age 3 Census projection 3,065,235 587,374 230,898 3,883,507 CPS estimate 3,064,706 584,498 228,661 3,877,865 Ratio of projection to estimate 1.00 1.01 1.01 1.00 Age 4 Census projection 3,121,016 608,019 230,192 3,959,227 CPS estimate 3,120,429 664,301 232,269 4,016,999 Ratio of projection to estimate 1.00 0.92 0.99 0.99

OCR for page 42
CATEGORICAL ELIGIBILITY OFINFANTS AND CHILDREN TABLE 4-2 Five-Year Accumulations of Census Projections and CPS Estimates by Single-Year Age Groups, March 1996-2000 45 White Black Other Total Age 0 Census projection 15,508,050 3,060,394 1,141,043 19,709,487 CPS estimate 15,416,027 2,851,665 1,019,426 19,287,118 Ratio of projection to estimate 1.01 1.07 1.12 1.02 Age 1 Census projection 15,394,415 3,008,550 1,110,768 19,513,733 CPS estimate 15,407,985 3,182,528 1,198,862 19,789,375 Ratio of projection to estimate 1.00 0.95 0.93 0.99 Age 2 Census projection 15,434,140 3,045,150 1,117,163 19,596,453 CPS estimate 15,463,718 2,979,156 1,198,652 19,562,526 Ratio of projection to estimate 1.00 0.95 1.00 1.00 Age 3 Census projection 15,533,367 3,098,387 1,119,274 19,751,028 CPS estimate 15,533,536 3,221,080 1,127,924 19,882,540 Ratio of projection to estimate 1.00 0.96 0.99 0.99 Age 4 Census projection 15,873,802 3,260,672 1,123,277 20,257,751 CPS estimate 15,875,210 3,201,070 1,147,431 20,223,711 Ratio of projection to estimate 1.00 1.02 0.98 1.00 across all races. For example, the CPS estimate of black infants is 11 per- cent below the census projection. The CPS estimates of white children of different age groups match the control totals very closely. For children ages 1 through 4 years who are black or of other races, the CPS estimates are sometimes too large and sometimes too small relative to control totals, but they closely match control totals for some age groups. Table 4-2 presents the ratios for a five-year accumulation of census projections and CPS estimates.2 The ratios are more stable estimates of the relationship between census projections and CPS estimates for the period 1996-2000 because they are based on five years of data. These ratios indi- 2The five-year interval was chosen to allow accumulation of sufficient CPS sample in each survey year of the "other race" groups.

OCR for page 42
46 ESTI~TING ELIGIBILI>~D PAR TICIPATION FOR THE ~CPROG~ care what appears to be a consistent underestimate of the number of non- white infants in the CPS. The patterns of the ratios across single years of age and races, as well as the total across races, are similar to the ratios ob- served for the year 2000 comparison in Table 4-2. ADJUSTMENT PROCEDURE TO IMPROVE THE ACCURACY OF THE COUNTS OF INFANTS AND CHILDREN The panel considered the size of the underestimation of infants sub- stantial enough to suggest a procedure to adjust CPS estimates as a correc- tion. The five-year accumulated ratios of the census projections to the CPS estimates like those presented in Table 4-2 could be used to adjust the weight value for an individual in a given age and racial group in the corre- sponding cell in the table. For example, to estimate the number of infants in 2001, the cumulative ratios from 1996-2000 would be used to change the weights for the 2001 estimates. Use of these ratios would slightly in- crease the weight given to a white infant (multiply the CPS individual weight by 1.01), increase the weight given to a black infant (multiply the CPS individual weight by 1.07), and increase the weight given to an infant whose race falls into the "other" category (multiply the CPS individual weight by 1.12). A similar adjustment would be applied to each age and race group. Furthermore, for the sake of the panel's exercise, males and females were combined. The CPS weights are developed separately by gen- der, so for a complete adjustment, separate adjustments should be made for males and females. Table 4-3 shows the results of this adjustment procedure using the 1996-2000 ratios for 2001 CPS estimates by age. A shift in estimates from older ages to infants is indicated by the percentage relative change for each age. For example, the adjustment results in a 2.1 percent increase in the total number of infants in 2001 and a 1.3 percent decrease in the total number of 1-year-olds. There are several ways to use such accumulated ratios to adjust the CPS weights. In the example given in Table 4-3, we used accumulated ratios from five past years to estimate the next year's population (e.g.,1996- 2000 accumulated ratios were used to adjust 2001 estimates). On one hand, to the extent that the ratios of accumulated data reflect stable trends over time (that is, relatively unchanging from year to year), the accumulated ratios from the 1996-2000 data could be made for several subsequent years, such as 2002, 2003, 2004, and 2005. On the other hand, if five-year accu-

OCR for page 42
CATEGORICAL ELIGIBILITY OFINFANTS AND CHILDREN TABLE 4-3 Adjustment of 2001 CPS Estimated Population by Race Using 1996-2000 Accumulated Census Projection to CPS Estimate Ratios 47 White Black Other Total Age 0 Ratio of projection to estimate 1.006 1.073 1.119 1.022 Revised2001 CPS estimate 3,137,907 581,358 223,417 3,942,682 % relative change 2.13 Age 1 Ratio of projection to estimate 0.999 0.945 0.927 0.986 Revised2001 CPS estimate 3,094,949 657,197 223,060 3,975,206 % relative change -1.31 Age 2 Ratio of projection to estimate 0.998 0.945 0.998 1.002 Revised 2001 CPS estimate 3,093,102 615,377 262,134 3,970,613 % relative change 0.17 Age 3 Ratio of projection to estimate 1.000 0.962 0.992 0.993 Revised2001 CPS estimate 3,058,363 588,310 237,622 3,884,295 % relative change -0.64 Age 4 Ratio of projection to estimate 1.000 1.019 0.979 1.002 Revised 2001 CPS estimate 3,103,575 679,129 235,810 4,018,514 % relative change 0.17 The revised 2001 CPS estimates are made by multiplying the ratio of the 2001 popula- tion estimates by the five-year accumulated adjustment ratios by age and race. mutated ratios can be computed for each year (for example, 1997-2001, 1998-2002, and so on) and used instead of the 1996-2000 ratios, changes in the trend of the ratios could be partially accounted for in the estimates. For example, to adjust the 2003 CPS estimates of infants and children, ratios accumulated over 1998,1999,2000,2001, and 2002 could be used. The panel did not explore how stable the adjustments are over time to see if one of these two methods is preferable. But such an activity should be conducted before a specific procedure is chosen. The panel does not give specific advice about how the adjustments are created (e.g., whether the five-year accumulated ratios in Table 4-2 could continue to be used in future years or if new five-year accumulated ratios

OCR for page 42
48 ESTI~TING ELIGIBILI>~D PAR TICIPATION FOR THE ~CPROG~ should be estimated every year). The five-year adjustments derived and presented in Table 4-2 should serve as a model for the general derivation of the adjustment factors. The multiple years of data used will help stabilize adjustments but still reflect changes in trends over time. RECOMMENDATION: To accurately estimate the number of in- fants and children using the CPS, USDA should apply five-year accu- mulated ratio adjustment factors to individual CPS weights using a procedure similar to the one outlined above. The adjustment factors should be calculated separately by single-year age intervals for each of the CPS control total race and gender groups. SIPP-BASED ESTIMATES OF THE NUMBERS OF INFANTS AND CHILDREN SIPP could be used to produce estimates ofthe numbers of infants and children who are eligible for WIC. SIPP, like the CPS, uses census popula- tion projections by age, race, and gender to construct posts/ratification ad- justments. The SIPP sample sizes are not large enough to support stable estimates of population controls for the "other race" category. Thus, SIPP population control adjustments are developed for each gender by single- year age intervals for nonblacks and for two-year age intervals for blacks (0-1, 2-3, and 4-51. The panel compared SIPP estimates of the population from December 1997 with census projections for December 1997. The comparison shows that the SIPP slightly overestimates the number of infants compared with the census by 0.6 percent and overestimates the number of children by 4.1 percent. SIPP estimates of women of childbearing ages (15-45) for De- cember 1997 overestimated the number of women relative to the census by 1.3 percent. This comparison included only one month's estimates from SIPP. It is difficult to draw conclusions about the accuracy of the counts of infants in SIPP. Because the number of black infants is not required to match totals for single-year age groups, it is likely that the count of black infants would be underestimated relative to infants in the white category. In SIPP, people of other races are included in the white race category. In the CPS, infants reported as "other race" were underestimated more than white and black infants, relative to control totals. Not having separate controls for this group and including them in the nonblack category would reduce the problem of

OCR for page 42
CATEGORICAL ELIGIBILITY OFINFANTS AND CHILDREN 49 underestimation of infants. However, the smaller SIPP sample sizes would result, all else equal, in less stability of estimates of single-year age groups for nonblacks and less stability in estimates of two-year age intervals for blacks. The SIPP estimates of the number of income-eligible infants are not used to estimate the number of pregnant and postpartum women as those from the CPS are. However, the presence of an infant in the household is used to infer that a woman was pregnant or that a woman is currently postpartum. If infants are undercounted, then estimates of the number of postpartum women will be undercounted as well. Whether or not this trans- lates into an undercount of pregnant women depends on the reason why the infant was not counted. If the household was missed entirely, then information on the mother will not be available. If the mother of the infant rounds the infant's age up to 1 year, the mother's pregnancy status can still be inferred, but the timing to which pregnancy is attributed will be off by a number of months. SUMMARY This chapter examines the undercount of infants and overcount of children in the CPS relative to population control totals. The undercount is especially problematic for blacks and for members of other nonwhite racial groups. It appears to be at least partially due to the age groupings for black and other race individuals when the CPS posts/ratification weights are constructed. The panel recommends a procedure to adjust the CPS estimates of the numbers of infants and children to correct for the undercount. We did not examine whether such a procedure is needed to adjust SIPP-based estimates of these groups. If SIPP is used to estimate eligibility, further exploration of the accuracy of estimates of infants and the implications of any accuracies on the estimates of pregnant and post- partum women may be warranted.