Click for next page ( 189


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 188
APPENDIX C Reconciling Different Estimates of Income and Adjunctive Eligibility This report has attempted to characterize how estimates of eligibility and participation vary when new methods of estimation are used. The pri- mary focus of this analysis is on estimating income and adjunctive eligibil- ity. To conduct this analysis, the panel employed three different data sets to estimate the effects alternative methods have on eligibility and participa- tion estimates: the March Current Population Survey (CPS), the Survey of Income and Program Participation (SIPP), and the Transfer Income Microsimulation model data (TRIM), which is based on the March CPS. The CPS is the data base that USDA currently uses to estimate eligibility. The SIPP and TRIM data bases each have features that allow estimation of new methodologies, but each does so in different ways. For example, SIPP directly asks respondents to report their monthly income, while TRIM uses annual reports of income and benefit receipt combined with respondent accounts of employment periods throughout the year to simulate monthly income. These differing approaches yield different estimates of eligibility. In this appendix, we attempt to explain why estimates of eligibility differ across these data sets. We first examine differences in estimates of eligibility when monthly income and WIC certification periods are used to estimate eligibility. We then examine differences in estimating adjunctive eligibility. 188

OCR for page 188
RECONCILING DIFFERENT ESTIMATES ESTIMATES OF ELIGIBILITY USING MONTHLY INCOME MEASURES AND ACCOUNTING FORWIC CERTIFICATION PERIODS 189 While the March CPS does not provide monthly income data, the Urban Institute's Transfer Income Microsimulation model (TRIM 3) pro- vides routines that impute monthly income to CPS files. These imputation routines utilize the income and unemployment data from the CPS to re- flect both monthly unemployment flows and the degree of income variabil- ity found in SIPP. Estimates of the proportion of infants and children in- come eligible from this modified CPS data base and the SIPP 1996 Panel data are presented in Table C-1. Employing the TRIM imputed monthly income and using a monthly certification period (the row labeled "average monthly" in the table), we find roughly the same percentages of income-eligible infants and children that were found by Gordon et al. (19971. The number of eligible infants increases 3 percent over the estimates using annual income, while the num- ber of income-eligible children increases by 1 to 2 percent. If we impute 12 months of eligibility to infants and children if their worst month's income is less than 185 percent of federal poverty guidelines (the "eligible in any month" row), again we find roughly the same percentage increases (28 and 24 percent) in the number of income-eligible infants that were found in the Gordon et al. study and a slightly smaller increase in the number of chil- dren (19 and 20 percent). The monthly income data imputed with the TRIM model does not provide sufficient information to fully model the WIC certification pro- cess. Since the public use files of the CPS do not provide the birthdates of individuals (only ages in March), we cannot model which month in the year infants initially become eligible and hence the number of months dur- ing the year they are income eligible. A similar problem occurs for children ages 1 through 4. To provide an upper bound estimate, we assume that all infants would be eligible for 12 months if their worst month's income quali- fied them for WIC. For children, we imputed 6 months of eligibility if their worst month's income qualified them for WIC and an additional 6 months of eligibility if their seventh worst month also made them income eligible. An alternative would be the second worst month. This modeling of the WIC certification process does not alter the esti- mates of the percentage of infants who are income eligible. The number of income-eligible children is reduced by this modeling of WIC certification

OCR for page 188
190 APPENDIX C TABLE C-1 Estimates of the Monthly Versus Annual Income Eligibility (Percentage of All Individuals) 1997 1998 Infants CPS/TRIM3 Annual 39.7 39.2 Average monthly 40.8 (1.03) 40.2 (1.03) Eligible in any month 50.6 (1.28) 48.7 (1.24) Certification periods 50.6 (1.28) 48.7 (1.24) SIPP Annual 38.9 35.1 Average monthly 43.5 (1.12) 41.3 (1.17) Eligible in any month 58.9 (1 52) 57.5 (1.64) Certification periods 56.6 (1.46) 54.1 (1.54) Children CPS/TRIM3 Annual 41.1 40.4 Average monthly 41.5 (1.01) 41.1 (1.02) Eligible in any month 49.0 (1.19) 48.4 (1.20) Certification periods 45.9 (1 12) 45.5 (1.13) SIPP Annual 42.4 39.6 Average monthly 44.8 (1.06) 42.1 (1.06) Eligible in any month 62.0 (1 46) 59.4 (1.50) Certification periods 56.8 (1 34) 53.9 (1.36) NOTE: The bracketed numbers represent the ratio of respective estimate to the corre- sponding estimate using annual income eligibility. SOURCES: The CPS estimates are from panel calculations based on extracts from the Urban Institute TRIM files for the respective calendar years. The SIPP estimates are from calculations made by Bitler et al. (2002). periods. Instead of 19- and 20-percent increases in the number of income- eligible children when only the worst month was considered, monthly in- come with the certification process is estimated to increase the number of income-eligible children 12 and 13 percent compared with estimates em- ploying an annual measure of income. Comparing the estimates derived from the TRIM data with those em- ploying data from the 1996 SIPP panel, we conclude that the marginal effect of monthly income with WIC certification periods is significantly larger when employing the SIPP data as opposed to the TRIM data. The

OCR for page 188
RECONCILING DIFFERENT ESTIMATES 191 SIPP data produces a marginal effect that is roughly twice that found when using the TRIM data, even though monthly income was imputed to reflect the variability of income found in the SIPP data. In an attempt to under- stand these differences, we examined the SIPP- and TRIM-imputed distri- bution of children by the number of months their monthly income was less than 185 percent of federal poverty guidelines. These distributions, as well as the distribution of children who had at least one month in which their income was less than 185 percent of poverty are presented in Table C-2. The distribution of TRIM-imputed months of income eligibility is quite different from the distribution based on the reported incomes in SIPP. Children are more likely (roughly 20 percent more likely) to have at least one month of income eligibility in SIPP than in the TRIM-imputed data. TABLE C-2 Distribution of Children Ages 1 to 4 Years by the Number of Months in Which Monthly Income Is Less Than or Equal to 185 Percent of Poverty in 1998 (Percentage) SIPP CPS/TRIM Number All Months Greater All Months Greater of Months Children Than Zero Children Than Zero 0 40.8 5 1.6 1 4.5 7.5 1.6 3.3 2 3.4 5.7 1.0 2.0 3 3.1 5.2 1.0 2.0 4 4.1 6.8 .9 1.9 5 1.8 3.1 .5 1.1 6 2.2 3.7 .9 1.9 7 2.5 4.2 .7 1.4 8 3.5 5.9 6.1 12.5 9 2.4 4.1 1.4 2.8 10 3.4 5.7 1.3 2.6 1 1 4.5 7.6 1.0 2.1 12 23.9 40.4 32.2 66.4 Computed percentage of all children employing the above distributions: Average monthly 41.3 41.1 Eligible in any month 59.2 (1.43) 48.4 (1.18) Certification periods 49.7 (1.20) 45.5 (1.1 1) Note: Numbers in parentheses represent the percentage increase in eligibility compared with estimates using annual income.

OCR for page 188
192 APPENDIX C But the distribution of children who have at least one month of eligibility is also quite different. In the SIPP data, the modal value is 12 months. How- ever, 60 percent of children have at least 1 month but less than 12 months of eligibility. The TRIM-imputed data, however, show a large percentage of children with 8 months of eligibility. The spike at 8 months reflects a pecu- liarity of the TRIM imputation routines. Each year there are four months with one more weekly pay period than the other eight months. TRIM accounts for these monthly differences and constructs eight months with fewer numbers of pay periods and hence less income. TRIM also imputes a much higher proportion of children with 12 months of income eligibility (66 percent) than is found in the SIPP data (40 percent). This evidence suggests that the differences between the SIPP and TRIM estimates can be explained in the following manner. Since we can- not estimate the effect of which months the children are certified as income eligible in the TRIM data, we can try to make SIPP estimates closer to the TRIM data by using the distributions of income-eligible months in Table C-2. Using the distribution of all children in the SIPP data to weight the number of months of eligibility, we can compute an average monthly esti- mate of the proportion of children in SIPP that is identical to the estimate from the TRIM data. Making this calculation, we find that 41.3 percent of children in 1998 were income eligible if monthly income was employed. This is almost identical to the 41.1 percent estimated from the TRIM data.1 The upper bound estimate, "eligible in any month," increases in the SIPP data by 43 percent compared with the estimates that use monthly certification, "average monthly." The similar estimate from the TRIM in- creases only 18 percent. The difference is the result of the underlying differ- ence in the SIPP estimates of the number of eligible children with at least one month of eligibility, which is much greater than the TRIM estimate. To mirror the TRIM certification process in the SIPP data, we im- puted 6 months of income eligibility to those children with 1 to 6 months of income eligibility and 12 months of eligibility to those children with 7 to 12 months of eligibility. Compared with the situation when monthly certification periods (average monthly) are employed, these calculations in- dicate that the SIPP data would have estimated that 20 percent more chil- 1The computed SIPP estimates in Table C-3 will not match the estimates for the SIPP data found in Table 5-1 because the earlier estimates take account of the timing of certifica- tion during the year, while the calculations in Table C-3 do not. But these calculations are done to make the SIPP estimates mirror what is done in the TRIM-imputed data.

OCR for page 188
RECONCILING DIFFERENT ESTIMATES 193 dren (49.7 instead of 41.3 percent) would have been eligible. The identical calculations from the TRIM data indicate only 11 percent more children are estimated to be eligible (45.5 instead of 41.1 percent). To decompose this difference, we first divide the population of children with at least 1 month of income eligibility into two groups one in which the children have 1 to 6 months of eligibility and the other in which the children have 7 to 12 months of eligibility. In the SIPP data, the first group constitutes 32.1 percent of the total number of children with at least 1 month of eligi- bility. When this form of certification is used instead of monthly certifica- tion, this group has an average gain of 3.2 months of income eligibility. In the TRIM data, this group experiences the same average gain in eligibility, but constitutes only 12.2 percent of the total number of children with at least 1 month of eligibility. In the second group of children with 7 to 12 months of eligibility, the SIPP data indicate that 67.9 percent of children with at least 1 month of eligibility are in this group and would experience an average gain of 1.1 months when certification is accounted for in this manner. The TRIM data show a larger proportion of children with at least 1 month of eligibility falling in the 7 tol2 month group and experience an average gain of only 0.8 months.2 Thus, SIPP is indicating a much larger gain in income eligibility due to the WIC certification process and monthly income because of two factors: . SIPP reported monthly income indicates that among those children with at least one month of eligibility, the average increase in the months of eligibility is largely due to more children having only a few months (1 to 5) in which the effect of certification on the aver- age number of months is greater than when the child has 7 to 12 months of monthly eligibility. SIPP data indicate that a higher proportion of children have at least 1 month of eligibility. Table C-3 extends the TRIM-based analysis to calendar years 1994 through 1999. This longer time series of estimates shows the relative stabil- ity of the impact of imputed monthly income on the size of income-eligible populations. The proportion of all infants and children who are income 2The average increase in the number of months of eligibility is smaller in the second group because of the high proportion of children at 12 months of eligibility compared with six months, when certification has no effect.

OCR for page 188
194 an o o be ._ to - ._ be o he; o .; - o be o ~ o - 1 1 no _1 ~ _ ~ GN GN GN Do GN GN GN GN GN GN ON GN GN GN GN o ~ US US ~ o ~ ~ . . . . . . . . Do Do ~ ~ Do GN ~ ~ A_ A_ A_ A_ A_ A_ ~ ~ ~ ~ o o ~ ~ o . . . . . . . . . . GN O Do Do O ~ Do u N ~_ ~_ ~_ ~_ ~_ ~_ ~ oo oo ~ GN o ~ ~ o . . . . . . ~ oo ~ ~ ~ U N O GN . . . . . . . . GN O O O ~ ~ GN u~ ~ ~ u~ u~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ o . . . . . . . . ~ ~ ~ ~ ~ ~ ~ oo ~ ~ u~ u~ ~ ~ u~ ~ GN ~ oo oo ~ ~ ~ GN . . . . . . . . ~ ~ u~ u~ ~ ~ u~ ~ ~ GN GN ~ O o ~ ~ o ~ ~ . . . . . . ~ ~ ~ ~ ~ ~ u~ o . . . . . . . . ~ ~ ~ ~ ~ u~ ~ o ~ ~ u~ u~ ~ ~ ~ ~ o o Oo ~ Oo s.= ~ ~ ~ ~ D=. - ~ H (J O O =0 ~ =0 s.= ~ ~0 _ C~ bC _ C~ _ - B C~ ._ - ._ bC ._ ._ C~ ._ bC C~ ._ . . o

OCR for page 188
RECONCILING DIFFERENT ESTIMATES 195 eligible (accounting for monthly income and certification periods) has been declining over the seven-year period. However, the impact of monthly in- come and certification on eligibility estimates compared with estimates based on annual income is constant throughout the period. For infants, use of monthly income and certification periods increases eligibility estimates by 23 to 29 percent. For children, the estimated increase is very stable at 11 to 13 percent. ADJUNCTIVE ELIGIBILITY AND UNDERREPORTING OF MEANS-TESTED PROGRAMS In Chapter 5, we observed that both the public use CPS and SIPP data may suffer from underreporting of participation in means-tested programs that would create eligibility for individuals in the WIC program. To exam- ine the effect underreporting may have on the number of people who gain WIC eligibility through adjunctive eligibility, we used TRIM data with imputed participation in these programs. The imputation procedure in TRIM makes sure the number of participants for each program matches control totals recorded by the programs' administrative records. Table C-4 presents estimates of the proportion of income and adjunctively eligible infants and children from TRIM imputed data for 1994 and 1996 through 1999 and from SIPP data for calendar years 1997 and 1998. The bracketed numbers represent the ratio of the proportion of eligible people in the row relative to the number eligible from the previous step (the row directly above). For example, in 1994, TRIM estimates that 42.2 percent of infants would be eligible based on their annual income. This proportion increases to 54.3 percent when monthly income and certification periods alone are used in the determination of income eligibility. This represents a 29 per- cent increase in the number of children eligible compared with estimates that use annual income. When adjunctive eligibility is also considered, the proportion of infants rises to 58 percent, or a 7 percent (1.07 = 58.2/54.3) increase in the number of eligible infants due to the marginal addition of adjunctive eligibility to the eligibility determination process.3 3The marginal effect of adjunctive eligibility may be understated while the effect of monthly income may be overstated due to the manner it iS being estimated in Table C-4. The effect of monthly income includes not only a "pure" effect of monthly income but also an interaction effect monthly income has with adjunctive eligibility the effect representing the number of individuals who would qualify for WIC either due to monthly income or enroll- ment in a means-tested program.

OCR for page 188
196 be _' - be - ._ be - o o - GN GN GN Do GN GN GN GN GN GN GN Cal _ o US ~ . . . oo ~ ~ oo ~ ~ ~ ~ o ~ ~ ~ . . . . . . GN oo ~ ~ ~ oo ~ ~ ON ~ ON ON ~ up GN ~ GN ~ . . . . . . O U N 00 GN ~ ~ ~ o . . ~ ~ ~ GN ~ O ~ GN ~ ~ oo O . . . . . . . . . . . . GN O ~ oo ~ O ~ u N GN ~ ~ GN A_ A_ . . . ~ UP ~ ~ - ~ - GN ~ o . . ~ ~ oo ~ u~ ~ ~ o ~ . . . ~ oo ~ ~ o . . ~ o ~ . . . ~ o ~ ~ u~ ~ . - - - - o ~ o ~ o ~ o ~ ~ = ~ = ~ ~ = ~ = ~ ~ ~ ~ 3 H (J

OCR for page 188
RECONCILING DIFFERENT ESTIMATES 197 The marginal effect of considering adjunctive eligibility is smaller than the effect of employing monthly income as opposed to annual income. While the TRIM estimates show a slightly larger marginal effect for both infants (11 and 15 percent increases in 1997 and 1998, respectively) and children (8 and 7 percent) than found in the SIPP data, the differences are explainable. In the SIPP data, enrollment in the means-tested programs is reported by the respondents. The TRIM data imputes individual enroll- ment in these programs so that the data reflects enrollment found in ad- ministrative program data. Table C-5 presents the average monthly counts of the number of infants and children reporting enrollment in Medicaid in SIPP and the CPS survey as well as the corresponding estimates in the TRIM-imputed data. Clearly the larger effects are the result of roughly 50 percent more infants and children having enrollment status in the TRIM data than in the SIPP data, which uses reported participation. While a clear case can be made that SIPP data may understate the marginal effect of adjunctive eligibility due to using reported as opposed to actual enrollment in these means-tested programs, the TRIM-imputed data do not necessarily represent truth. TRIM utilizes characterizations of state Medicaid programs to determine Medicaid eligibility in order to assign enrollment to those who are eligible for benefits. However, these control totals pertain to all children under age 18 years, not to the WIC target group of children under age 5 years. Hence there is no guarantee that TRIM is assigning enrollment of infants and children under age 5 in a way that reflects the true number of that age who are enrolled. They may be over- or understating the true number. In addition, the routine employed by TRIM assigns enrollment to nonreporters in order to hit state-level control totals TABLE C-5 Average Monthly Receipt of Medicaid as a Percentage of Total Population of Infants and Children SIPP CPS TRIM Reported Reported Imputed 1997 Infants 26.2 24.7 38.0 Children (1 to 4 years) 20.6 21.2 31.1 1998 Infants 24.9 22.9 37.5 Children (1 to 4 years) 19.5 19.2 28.9

OCR for page 188
198 APPENDIX C on a completely random basis. The assignment does not take into account the specific income of the family. It is possible that TRIM may be assigning too many high-income infants and children to enrollment status and hence overstating the true marginal effect of adjunctive eligibility.