National Academies Press: OpenBook

Estimating Eligibility and Participation for the WIC Program: Phase I Report (2001)

Chapter: 4. Potential Biases in Eligibility Estimates

« Previous: 3. Current Methods for Estimating Eligibility and Full-Funding Participation
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

4
Potential Biases in Eligibility Estimates

Concerns over high coverage rates for infants and postpartum women have led some observers to conclude that the Food and Nutrition Service’s (FNS) estimates of the number of eligible individuals are biased and understate the true number of eligible people. While numerous assumptions are made in the FNS estimation strategy, the panel chose to examine the assumptions likely to have the greatest impact on the estimate of the number of individuals eligible for WIC. In particular, the following estimation concerns are examined in this chapter:

  • The accuracy of the Current Population Survey (CPS) in counting all infants and children;

  • Adjunctive eligibility through Temporary Assistance for Needy Families (TANF), the Food Stamp Program, and Medicaid;

  • Use of monthly income versus annual income to determine income eligibility;

  • Adjustment for 6-month certification periods;

  • Definition of the economic unit; and

  • The number of individuals who are at nutritional risk among those who are income eligible for WIC.

The FNS identified three additional areas of the estimation strategy that could potentially affect estimates of the number of eligible individuals (U.S. Department of Agriculture, 1999a). These include the use of alter-

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

native data sources, such as the Survey of Income and Program Participation; the timeliness of the data used in the estimation; and assumptions concerning breastfeeding rates among postpartum women. In the second phase of the study, the panel plans to consider these issues as well as others, which are outlined in Chapter 6.

ACCURACY OF THE CURRENT POPULATION SURVEY

The primary database for FNS’s annual estimates of the number of income eligible infants and children is the March supplement of the CPS. This nationally representative survey of the population collects demographic and income information from over 55,000 American households. Utilizing the CPS information on the number of family members, family income, and age of persons in the family, FNS makes two core estimates: (1) the number of infants who live in families whose annual income is less than 185 percent of federal poverty guidelines and (2) the number of children who live in families whose annual income is less than 185 percent of federal poverty guidelines. While the accuracy of both of these core estimates is crucial, the accuracy of the estimates of the number of infants is especially important for two reasons. First, the number of income eligible infants is the base from which the number of pregnant and postpartum women eligible for WIC is inferred. Hence any errors in estimating the number of income eligible infants would also be reflected in the estimates of the number of income eligible women in these groups. Second, high estimated coverage rates of infants and postpartum women led the panel to question whether the numbers of eligible people in these groups were being properly estimated.

To consider the accuracy of the CPS estimates of total number of infants and children, the panel asked the Census Bureau to make a presentation at the panel’s Workshop on Estimating WIC Eligibility and Full-Funding Participation. Notes from the presentation by Gregory Spencer of the Census Bureau Division of Population Estimates were given to the panel for its consideration (Spencer, 2001). To assess the accuracy of the CPS estimates, Spencer (2001) compared weighted CPS sample estimates of the numbers of infants and children to the CPS control totals. These control totals are estimated from the Census Bureau’s annual estimates of the noninstitutionalized U.S. population of infants and children (which are produced using birth and death records from vital statistics data with an adjustment for migration) plus an adjustment for the net undercount in

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

the 1990 decennial census. These CPS control totals for infants and children were compared with weighted CPS sample estimates of infants and children. Person weights for the CPS are developed for each person of a given age, race, and gender and should ensure that CPS sample estimates of the population of infants and children match the control totals for those age groups. However, data provided in Spencer (2001) show that weighted CPS sample estimates of infants and children do not match control totals. Table 4–1 shows the percentage difference between weighted counts of infants and children from the CPS and their respective control totals as of March of the given year. These estimates indicate that the CPS weighting scheme utilized by the Census Bureau consistently underrepresents the number of infants, but that since 1994, it consistently overrepresents the total number of children.

To understand why the weighted counts of infants and children do not add up to the control totals, Spencer (2001) provides a detailed description of how the weights are constrained to the control totals. The number of white male and white female infants are constrained to add to the CPS’s control totals for each age with single-year intervals (i.e., separately for age 0, age 1, age 2, etc.). However, the number of nonwhite infants is not required to match totals for single-year age intervals because sample sizes

TABLE 4–1 Percentage Difference Between the Weighted Current Population Survey Counts of Infants and Children and Control Totals from Population Estimates

 

Percentage Difference

Year

Infants

Children

Infants and Children

1992

–2.0

–2.6

–2.4

1993

–1.0

–2.4

–2.2

1994

–1.1

0.4

0.1

1995

–2.9

1.0

0.2

1996

–0.7

0.5

0.2

1997

–2.6

0.4

–0.2

1998

–1.0

0.3

0.1

1999

–4.1

0.9

–0.1

2000

–2.4

0.7

0.1

 

SOURCE: Calculations by panel from estimates provided by Spencer (2001).

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

for these two groups are too small to do so. In particular, black male and black female weights are each constrained to match totals for 2-year age intervals (i.e., infants and 1-year-olds together; 2- and 3- year olds together; and 4- and 5- year olds together). For males and females of other races, separate gender weights are each constrained to totals for 5-year age intervals (i.e., all infants and children aged 0 through 5). The net effect of not controlling the weights to single-year age intervals for nonwhite infants appears to be the primary reason for the systematic underrepresentation of infants on the CPS.

Over the nine years of data presented in Table 4–1, the CPS underrepresents the total number of infants by an average of 2 percent each year. Because the incomes of black families are, on average, lower than incomes of white families, it is therefore likely that the number of income eligible infants is understated by more than 2 percent. This inference requires further investigation, however. In addition, the most appropriate method to rectify this problem needs to be examined. Currently, the panel foresees two options. One option is to reestimate the CPS person weights for infants and children to reflect the Census Bureau’s population and undercount estimates for narrowly defined age groups that are relevant for WIC eligibility estimates. The second option is to construct an adjustment factor that could be applied to any calculations made from the CPS to reflect the underrepresentation of infants.

Finally, the discussion above assumes that the Census Bureau’s control totals reflect an accurate estimate of the number of infants and children in the population. Analysis of the 2000 census would provide valuable insights into whether this assumption is valid. Such an analysis is beyond the scope of the panel’s charge, however.

ADJUNCTIVE ELIGIBILITY

Current methods used to estimate the number of people who are income eligible for WIC do not account for those adjunctively eligible through participation in TANF or the Food Stamp Program and make only a small adjustment for Medicaid adjunctive eligibility. This small adjustment has been recognized as inadequate by FNS. The panel concurs that the current FNS methodology inadequately accounts for adjunctive eligibility. This section presents simulations of the number of people who are adjunctively eligible for WIC through the TANF, Food Stamp, or Medicaid programs.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

Eligibility Rules for WIC, Medicaid, TANF, and Food Stamps

There are several notable differences in WIC income eligibility rules and the income eligibility rules of Medicaid, TANF, and food stamps. These differences are important in this context because some people who are not income eligible for WIC may be eligible for one of the other three programs, and could then be adjunctively eligible for WIC if they enroll in one of the other programs. Most notable are those differences between Medicaid and WIC eligibility because the income thresholds for Medicaid are higher than those of TANF and food stamps, and in some states, higher than the WIC threshold. As of 1998, the income thresholds for Medicaid for infants were above the 185 percent of poverty guidelines of WIC in eight states: Arkansas, California, Hawaii, Minnesota, Rhode Island, Tennessee, Vermont, and Washington. With the exception of California, each of these states also have income eligibility thresholds for children that are also above 185 percent of poverty. Furthermore, all but Arkansas and Washington also have income eligibility thresholds for pregnant women that are above the WIC threshold. There are other important differences between WIC and Medicaid eligibility rules. As of 1996, 37 states have medically needy programs that allow subtractions of medical expenses from income for determining eligibility. Medicaid also allows other income disregards, whereas WIC considers only gross income. Therefore, some people may be adjunctively eligible through Medicaid if their net income is below but their gross income is above 185 percent of poverty. In all, 13 states had Medicaid net income limits for infants, 10 states had net income limits for children age 1 to 5, and 11 states had net income limits for pregnant women that exceeded 185 percent of poverty guidelines as of 1998. There are also differences in the Medicaid and WIC definitions of families, periods of certification, and eligibility redetermination. Lewis and Ellwood (1998) discuss the differences in Medicaid and WIC eligibility rules in more detail. Because of these differences in program eligibility, some observers have criticized FNS’s small adjustment for Medicaid adjunctive eligibility (Greenstein and Ku, 2000).

Differences between the Food Stamp, TANF, and WIC eligibility rules have received less attention with respect to adjunctive eligibility (probably because the income thresholds for these programs are lower than the threshold for WIC), although some differences in rules are noteworthy. First, both the Food Stamp Program and TANF have monthly certification periods, while WIC has 6-month and yearly certification periods. Current

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

FNS methodology to estimate WIC eligibility uses annual income instead of monthly income. If a family has a month or two in which their income is low, they might apply for food stamps or TANF and become certified for those months and hence eligible for WIC. Use of annual income for WIC eligibility estimation may not count these people as eligible.

Second, the income concept employed for WIC eligibility includes payments from means-tested cash programs such as TANF and Supplemental Security Income (SSI). The inclusion of payments from TANF in the measure of income used to assess income eligibility for WIC raises the possibility that infants and children who receive TANF are determined to be income ineligible even though they would be adjunctively eligible because of their participation in TANF. Currently no adjustment is made for infants or children whose annual income places them above 185 percent of the poverty guidelines, but who may actually be certified as eligible because of their participation in TANF or the Food Stamp Program.

Estimates of the Number of Infants and Children Adjunctively Eligible for WIC

To estimate the number of infants and children who are eligible to receive WIC benefits, information about the income of the child’s economic unit is not sufficient to determine their eligibility. WIC eligibility can also be gained through enrollment in means-tested programs (TANF, food stamps, and Medicaid). While the CPS collects both income and program participation data on individual families and households, these measures are insufficient for two reasons. First, the CPS collects only annual income data. The use of annual income as opposed to monthly income is believed to understate the number of infants and children who would be eligible on the basis of their monthly income. Second, individuals tend to underreport their participation in means-tested programs to surveys such as the CPS.1 The direct use of the survey data on program participation would then in turn understate the number of infants and children who would be adjunctively eligible for WIC because of their participation in other means-tested programs, particularly Medicaid.

To rectify the deficiencies in the CPS data for purposes of predicting

1  

See Bavier (1999), Primus et al. (1999), and Wheaton and Giannarelli (2000), for recent accounts of underreporting of transfer program participation and income in the CPS.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

program eligibility in WIC and other means-tested programs, researchers have resorted to modifying the survey data at the individual level through the use of modeling techniques known as microsimulation models. For this study, the panel employed data produced by the Transfer Income Microsimulation 3 (TRIM3) model. (TRIM is explained in more detail in Appendix D.) This model was developed and is maintained by the Urban Institute with funding from the Department of Health and Human Services. It has been used for over 30 years to analyze changes in eligibility rules for means-tested programs such as AFDC and Medicaid, as well as major welfare reforms including the 1996 welfare reform act leading to the formulation of the TANF program.

To examine the possible magnitude of the number of people adjunctively eligible through participation in TANF, the Food Stamp Program, and Medicaid, the panel requested the Urban Institute to extract data from the TRIM model. These data are based on the March 1999 CPS and contain the person records of 12,708 infants and children. Each record contains information on the number of members of the family unit, income of the family (as defined by the census money income definition), and whether the child was covered by any private or governmental health insurance. These variables reflect the values reported by the family to the CPS but with an adjustment the TRIM model makes to ensure that the data match data on program participation from administrative records collected as part of the programs: the number of months the child was enrolled in TANF, the number of months enrolled in the Food Stamp Program, and the number of months enrolled in Medicaid. Utilizing data from the CPS and information on state Medicaid programs, the TRIM model imputed the number of months that the child would be eligible for Medicaid. Based on each family’s annual income and some assumptions about seasonal income patterns, the TRIM model also imputed estimates of each month’s income through the year.2Appendix D provides additional details on how monthly income is simulated from the CPS in the TRIM model.

Current FNS methodology for estimating eligibility was applied to these data to replicate the counts of income eligible infants and children.

2  

Income from sources other than transfer programs also tends to be underreported in surveys. TRIM does not make adjustments for underreporting of income from these othersources.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

TABLE 4–2 Adjunctive Eligibility Adjustments and Simulated Estimates of the Number of Income Eligible Infants and Children (counts are in millions)

 

Infants

Children

 

Counts

% Change from Baseline

Counts

% Change from Baseline

Baseline (using current FNS methodology with TRIM3 data)

1.475

 

6.307

 

Annual incomes less than 185 percent of poverty guideline or participated in TANF or Food Stamp Program

1.619

9.8

6.645

5.4

Include Medicaid participants

2.146

45.4

7.640

21.1

Include uninsured Medicaid eligibles

2.200

49.2

7.775

23.3

Include all Medicaid eligibles

2.422

64.2

8.673

37.5

 

SOURCE: Calculations by panel using data provided by the Urban Institute.

We estimated that 1.475 million infants and 6.307 million children in the nonterritorial United States would be income eligible for WIC during 1998.3 This closely approximates FNS’s estimates of 1.488 million infants and 6.359 million children for 1998.4 We will call the 1.475 infants and 6.307 children the baseline estimates; they are presented in Table 4–2.

The panel next examined the extent to which this procedure excluded

3  

“Unrelated children” were identified as foster children and hence deemed to be eligible for WIC regardless of the foster family’s income in accordance with FNS methodology.

4  

Our estimates based on TRIM data differ from the FNS estimates by 0.8 percent. This difference is due to the fact that TRIM data on income have been adjusted to account for underreporting of transfer income in the CPS data. Income data have been adjusted to reflect “simulated” public assistance that has been controlled to match aggregate state administrative totals. Hence there are fewer individuals who fall under the 185 percent poverty guidelines compared with the CPS public use files used by FNS.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

infants and children who were enrolled in TANF or the Food Stamp Program at some point during the year. If we add the infants and children who enrolled for at least one month in either one of these programs during 1998, then the estimate of WIC eligibility rises to 1.619 million for infants and 6.645 million for children. This represents an increase of 9.8 percent in the infant population and 4.5 percent in the child population from the baseline estimates.

To estimate the number of additional infants and children who are eligible for WIC through Medicaid adjunctive eligibility, the panel provides an upper bound, a lower bound, and an intermediate estimate of eligibility. The WIC regulations state that to gain adjunctive eligibility through Medicaid, the infant or child must be enrolled and not merely eligible for Medicaid. However, all those eligible for Medicaid are essentially also eligible for WIC because all they must do is enroll in Medicaid to be considered adjunctively eligible for WIC. Therefore, an estimate of the number of people who are eligible for Medicaid is an upper-bound estimate of the number of infants and children who could gain WIC eligibility through enrollment in the Medicaid program. A lower-bound estimate for the number who could gain adjunctive eligibility would be those infants and children who were enrolled in Medicaid for at least one month in 1998. While these scenarios may provide both a lower and an upper bound for the impact of Medicaid adjunctive eligibility, both have problems. The upper bound of all eligible persons regardless of their enrollment status actually includes some people who are already covered by private health insurance. The lower bound does not include WIC applicants with incomes greater than 185 percent of poverty who are not enrolled in Medicaid but are eligible. This group can always apply for Medicaid and become adjunctively eligible for WIC. The panel also obtained an estimate that is in between the two bounds. This estimate added infants and children who were eligible for Medicaid but were not covered by private health insurance during the year. The estimates from these three scenarios are presented in Table 4–2.

The inclusion of all Medicaid eligible people has a large impact on the number of WIC eligible infants and children. Using these upper-bound estimates, the number of eligible infants rises to 2.422 million from 1.619 million, while the number of eligible children rises to 8.673 million from 6.645 million compared with estimates that account for annual income tests and adjunctive eligibility through TANF and Food Stamp Program participation. For infants, this represents a 50 percent increase in eligibility

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

estimates compared with the estimates that take food stamps and TANF adjunctive eligibility into account and a 64 percent increase from the baseline estimates. For children, this represents a 31 percent increase in eligibility estimates compared with the estimates that take Food Stamp and TANF adjunctive eligibility into account and a 38 percent increase from the baseline estimates. The large increases in the estimated numbers of eligible infants and children are not solely the result of increases in the Medicaid income limits beyond 185 percent of the poverty guidelines in some states. Rather, Giannarelli and Morton (2001) provide evidence that a large proportion of the increase in WIC eligibility actually comes from infants and children who live in states with Medicaid limits that are 185 percent of poverty or less. The reason is that Medicaid allows families to subtract from their gross incomes certain allowable deductions which results in a number of families with gross incomes exceeding 185 percent of poverty who are eligible for Medicaid.

In reality, not all applicants who are eligible for Medicaid will enroll, and some may already be covered by private insurance. A more conservative approach provides lower-bound estimates of those who are adjunctively eligible for WIC. These lower-bound estimates were constructed by including only those infants and children who had at least one month of enrollment in Medicaid. Results in Table 4–2 show that even this conservative approach represents a substantial increase in the number of income eligible infants and children. Compared with the estimate determined by annual income and adjunctive eligibility through TANF or the Food Stamp Program, the number of eligible infants rose from 1.619 to 2.146 million and from 6.645 to 8.673 million for children. For infants, this represents a 33 percent increase in eligibility estimates compared with the estimates that take Food Stamp and TANF adjunctive eligibility into account and a 45 percent increase from the baseline estimates. For children, this represents a 15 percent increase in eligibility estimates compared with the estimates that take Food Stamp and TANF adjunctive eligibility into account and a 21 percent increase from the baseline estimates.

If those infants and children who are Medicaid eligible but lack health care coverage are included, the number of income eligible infants rises from 1.619 to 2.200 million and from 6.645 to 7.775 million for children compared with the estimate determined taking adjunctive eligibility through food stamps and TANF into account. For infants, this represents a 36 percent increase in eligibility estimates compared with the estimates that take Food Stamp and TANF adjunctive eligibility into account and a 49 percent

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

increase from the baseline estimates. For children, this represents a 17 percent increase in eligibility estimates compared with the estimates that take Food Stamp and TANF adjunctive eligibility into account and a 23 percent increase from the baseline estimates.

Considering the total effect of adjunctive eligibility (including those adjunctively eligible through any one of the three programs) results in a large impact on the estimated number of eligible infants and children. Even if the conservative approach of including only those who participate or are enrolled in TANF, the Food Stamp Program, or Medicaid is taken, the effects are large. There is a 45 percent increase in the number of estimated eligible infants for WIC compared with the 1998 baseline estimate: the estimated number increases to 2.146 million from 1.475 million. For children, the increase is not as large but still sizable. The number of estimated eligible children increases 21 percent from 6.307 to 7.640 million. The substantially larger understatement of the number of infants relative to the understatement of the number of children helps explain why current estimates of coverage rates for infants are high both in absolute and relative terms to the coverage rates for children.

On the basis of these simulations, it is apparent that there are substantial numbers of infants and children who are adjunctively eligible for WIC but are not otherwise counted as eligible for WIC given current methods used to estimate eligibility. Current FNS methodology does not account for the substantial effect that adjunctive eligibility has on the total number of estimated eligible people. Therefore, current estimates of the number of income eligible infants and children are underestimated. Furthermore, because the number of eligible pregnant and postpartum women are derived from the number of eligible infants, it is probable that these numbers are also underestimated.

Conclusion: Not fully accounting for adjunctive eligibility results in a substantial underestimation of the number of people eligible for WIC.

The panel has not fully explored alternatives for estimating how many people are adjunctively eligible for WIC but future work will give this issue more consideration. For the remainder of this report, however, we take the conservative approach and count only those participating in other programs as adjunctively eligible. The estimations of eligible infants and children resulting from this approach are called the new baseline estimates in the remainder of the report. We do note that estimates of the number of

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

adjunctively eligible for WIC could be even larger if estimates counted all those eligible for Medicaid as adjunctively eligible for WIC regardless of their current Medicaid enrollment status, since more people are eligible for Medicaid but are not enrolled.

USE OF ANNUAL VERSUS MONTHLY INCOME

Use of annual income to estimate income eligibility has been highlighted as one possible barrier to accurately estimating WIC eligibility (U.S. Department of Agriculture, 1999a; Gordan et al., 1997). While the WIC regulations are vague about the time period for determining family income, many observers suggest that using monthly income of the family would be closer to the rules employed by states and local WIC personnel. Given the variability of income over the course of the year, and especially around the birth of a child, the use of annual income or average monthly income will tend to overstate the family’s income at the time of application for WIC. As an alternative to the use of annual income, the panel employed a monthly income test based on the family’s worst income month or the month of the year in which the family’s income was lowest. In this situation, if the family’s lowest month’s income was less than 185 percent of the poverty guidelines for a month, then the infant or child was considered income eligible for a full year.5

First, the panel compared the use of the worst month income test to the use of an annual income test, without including those individuals adjunctively eligible for WIC through enrollment in other transfer programs. These results are presented in Table 4–3. Adopting this difference only, the number of income eligible infants rose by 25 percent (from 1.475 to 1.845 million), while the number of income eligible children rose by 21 percent (from 6.307 to 7.612 million) compared with the original baseline estimates for each category. These results compare nicely to the estimates from Gordon et al. (1997), who used actual (nonsimulated) Survey of Income and Program Participation (SIPP) data. They performed two sets of calculations. One set constructed an estimate of annual income from the monthly SIPP data from three calendar years (1990–1992) and

5  

Infants are certified as eligible for a year, while children and postpartum women are certified for 6-month periods. Later, we employ a 6-month certification period for children and reestimate the number of eligible children.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

TABLE 4–3 Monthly Income Adjustments and Simulated Estimates of the Number of Income Eligible Infants and Children (counts are in millions)

 

Infants

Children

 

Counts

% Change from Baseline

Counts

% Change from Baseline

Baseline

1.475

 

6.307

 

Monthly income using “worst month” and no adjustment for adjunctive eligibility

1.845

25.1a

7.612

20.7a

New baseline (with adjunctive eligibility)

2.146

 

7.640

 

Monthly income using “worst month” with an adjustment for adjunctive eligibility

2.230

3.9b

8.306

8.7b

aPercent change from baseline without accounting for adjunctive eligibility.

bPercent change from new baseline that takes adjunctive eligibility into account.

SOURCE: Calculations by panel using data provided by the Urban Institute.

used it to determine income eligibility. The other set employed the monthly income data and determined eligibility based on the worst month. They found that use of the worst month income test raised eligibility estimates by 25 percent for infants and 26 percent for children compared with estimates that used the constructed annual income measure. Hence, the TRIM-imputed monthly income flows are of the same order of magnitude as those calculated from actual (nonsimulated) data.

The adjunctive eligibility issue is closely related to the issue of income variability over the year. Consider a family that experiences a few months of unemployment and subsequent financial hardship within a year, but otherwise has income that is stable and above the WIC income threshold. This family may receive food stamps, TANF benefits, or Medicaid health insurance coverage during the months of hardship, making them adjunctively eligible for WIC as well. Furthermore, since WIC certification periods are for either 6 months or 12 months, these families may actually receive WIC longer than they receive food stamps, TANF, or Med-

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

icaid. Thus, estimates of how many infants and children are adjunctively eligible based on enrollment in these programs already captures some who would be eligible if the worst income month criterion was used. Since adjunct eligibility is part of the WIC regulations, in the panel’s judgement this should be accounted for in the eligibility estimations. This is why we have chosen to estimate adjunctive eligibility before estimating eligibility using alternative definitions of income. The marginal impact of the use of monthly income is then measured from the base that takes account of adjunctive eligibility first (the new baseline estimates). Doing so, we estimate that 2.230 million infants and 8.306 million children would be eligible for WIC in 1998. The marginal impact of monthly income represents an increase of 4 percent for infants and 9 percent for children, compared with the new baseline estimates. Thus, once adjunctive eligibility is accounted for, the marginal impact of monthly income, while still sizable, is not as large as was previously suggested.

SIX-MONTH CERTIFICATION PERIOD FOR CHILDREN

All of the estimates provided thus far in this report have assumed that an individual found to be eligible for WIC at a point in time is certified to be eligible for an entire year. For infants, this does indeed reflect WIC regulations. However, children must be recertified for eligibility every 6 months (as must postpartum women). Thus the estimates for adjunctive eligibility and the impact of monthly income may be overstated for this group.

To explore the effect of this shorter certification period for children, the panel conducted the following simulation. The Urban Institute data contain information about the number of months that the child participated in TANF, the Food Stamp Program, and Medicaid, as well as the number of months the family passed the income test. We considered the hypothetical situation in which monthly certification was employed to determine WIC eligibility. Under this hypothetical, an approximation to the number of months that the child was eligible for WIC was the maximum of the number of months the child had passed the income test, was adjunctively eligible, or both. To approximate 6-month certification, it was assumed that if a child had a minimum of 6 months of WIC monthly eligibility, then he or she was eligible for the entire year. However, if a child had less than 6 months of eligibility, then he or she was considered certified for only 6 months.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

Employing this definition of certification, the average number of children eligible for WIC during 1998 was 7.913 million (these results are not presented in a table). Compared with previous estimates employing monthly income and adjunctive eligibility but annual certification for all children, this estimate is 5 percent lower; however, it is 4 percent higher than the estimate for children that employed annual income and adjunctive eligibility.

DEFINING THE ECONOMIC UNIT

The current FNS methodology employs the Census Bureau’s family definition to represent the WIC economic unit. A census family is defined as all persons related by blood or marriage who live together. For example, if a mother with an infant and a child lives with her two parents, then the FNS methodology would consider all five persons to constitute an economic unit for determination of WIC eligibility. However, as noted above, the regulatory definition of the economic unit allows considerable discretion on the part of WIC personnel. The staff member could determine that the mother, infant, and child are economically independent of her parents and hence would count only the income of this three-person unit, not the five-person unit, in determining eligibility for WIC. While the census family represents a broad definition of the economic unit, the panel recognized that a narrower definition of the economic unit could result in more individuals being identified as being eligible for WIC. The panel explored the use of an alternative definition of the economic unit that includes only parents and children under the age of 18 years. In our example, this alternative definition considers only the mother, her infant, and her child as the economic unit. For a lack of a better term, we denote this definition as the narrow family compared with a broad family definition that would consider the two parents of the mother (grandparents of the children) as part of the economic unit.

The panel used Urban Institute data and the TRIM model to examine the sensitivity of the estimated number of income eligible persons to the definition of a WIC economic unit. Two scenarios reflect alternative ways that WIC staff might assess different living arrangements. Under a restrictive scenario, we considered the infants and children to be eligible only if they were eligible under both the narrow and the broad definitions of a family. Under a more generous scenario, we considered them eligible if the family meets income eligibility requirements for at least one of the definitions.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

The estimates of income eligible infants and children under the panel’s new baseline estimates (including adjunctively eligible participants in the Food Stamp Program, TANF, and Medicaid) and these two scenarios are presented in Table 4–4.

The use of the restrictive scenario of the economic unit has only a small negative effect on the number of income eligible people. The estimated number of eligible infants falls by 0.2 percent, while the estimated number of children falls by 0.3 percent. The more generous scenario of the economic unit has a larger impact on the estimates of the income eligible people, but still a modest one overall. The number of eligible infants rises by 1 percent, while the number of eligible children rises by 1.5 percent from the new baseline estimates.

Giannarelli and Morton (2001) present estimates of the effect of these alternative unit definitions that suggest a much larger impact on the number of income eligible infants and children. However, the baseline they employed did not account for adjunctive eligibility. Our estimates employ a baseline that does account for adjunctive eligibility. The impact of these alternative definitions appears to be much more modest once adjunctive eligibility is accounted for in the estimates. Thus, the definition of the

TABLE 4–4 Definition of the Economic Unit and Simulated Estimates of the Number of Income Eligible Infants and Children (counts are in millions)

 

Infants

Children

 

Counts

% Change from New Baseline

Counts

% Change from New Baseline

New baseline

2.146

 

7.640

 

Eligible under both broad and narrow definitions of the economic unit

2.140

–0.2

7.614

–0.3

Eligible under at least one definition of the economic unit

2.166

1.0

7.754

1.5

 

SOURCE: Calculations by panel using data provided by the Urban Institute.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

economic unit does not appear to be of much significance for estimating income eligibility and thus is not further explored here.

ESTIMATION OF THE PREVALENCE OF NUTRITIONAL RISK

To be fully eligible to receive WIC benefits, applicants must also be found to be at nutritional risk. This requires meeting at least one of the many risk criteria for the state in which the applicant lives.6 To account for this final eligibility requirement in estimating WIC eligibility, the FNS methodology adjusts the estimated number of income eligible persons downward by a constant percentage to account for those who are income eligible but not nutritionally at risk. Currently used adjustment factors by category are: 0.95 for infants, 0.752 for children, 0.913 for pregnant women, 0.933 for nonbreastfeeding postpartum women, and 0.889 for breastfeeding postpartum women. These adjustment factors were based on estimates of nutritional risk for income eligible individuals from the first WIC Eligibility Study (WES I) conducted in the early 1980s (U.S. Department of Agriculture, 1987). The WES I study developed a “modal” set of nutritional risk criteria (a list of the criteria most commonly used by the states) based on the operational definitions of the criteria used in each state at that time. Using this modal set of risk criteria with data on income, diet, and health status from the 1980 National Natality Survey (NNS) and the 1978–1980 National Health and Nutrition Examination Survey (NHANES II), WES I estimated the number of income eligible persons who were also at nutritional risk. The adjustment factor for infants was updated in 1991 from 72 percent to 95 percent to account for the fact that infants whose mothers participated in WIC are automatically considered nutritionally at risk.

New Estimates of the Prevalence of Nutritional Risk

The WES I nutritional risk analysis was recently reconsidered in the second WIC Eligibility Study (WES II) (U.S. Department of Agriculture,

6  

Prior to 1999, criteria used by states varied widely and were unstandardized. However, states have now adopted standardized anthropometric, medical, predisposing, and certain dietary risk criteria. These are described in WIC Policy Memorandum 98–9, Nutritional Risk Criteria. An expert panel of the Institute of Medicine, the Scientific Basis for Dietary Risk Criteria for WIC Programs Committee, is currently examining dietary risk criteria in the still unstandardized category “Failure to Meet Dietary Guidelines.”

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

1993, 1996). WES II used data from the 1988 National Maternal and Infant Health Survey (NMIHS) and Phase I (1988–1991) of the National Health and Nutrition Examination Survey (NHANES III) to estimate the prevalence of nutritional risk among income eligible WIC populations. These estimates were produced by the firm Sigma One under contract with FNS in a report entitled “Nutrition Risk and Eligibility for WIC,” in February 1999 (U.S. Department of Agriculture, 1999b). This study shows increases in the proportion of income eligible persons who are at nutritional risk for each category. Specifically, the study estimates that 90.4 percent of income eligible children, 88.4 percent of income eligible infants, and 95.2 percent of income eligible women were also at nutritional risk. These estimates are higher than those from the WES I study, except that the percentage of infants at nutritional risk is lower than the 95 percent adjustment factor currently used.

Concerns about the methods and inconsistencies in the report led the panel to conclude that the new estimates of the prevalence of nutritional risk in income eligible persons should not be adopted without further investigation. Two problems in the Sigma One report arise that warrant skepticism about its findings: the method used to combine data on nutritional risk prevalence from two datasets and the relationship between the estimated risks for infants and women.

WES II used the NHANES III Phase I survey to estimate the percentage of income eligible women who met at least one of the modal nutritional risk criteria mentioned previously.7 NHANES III estimates show that 94 percent of women are at dietary risk (i.e., did not consume at least the minimum number of servings from food groups listed in the dietary criteria) and an additional 4.3 percent met the anthropometric and medical criteria. WES II also estimates the prevalence of medical risk for income eligible women from the NMIHS 1988. This survey includes measures of medical risk not found in NHANES III, but does not include measures of dietary risk. Estimates from NMIHS show that 72.1 percent of income eligible women are at medical risk compared with 83.4 percent in NHANES III (when dietary risk is not considered). To come up with a total estimate of nutritional risk, WES II averages these two medical risk

7  

WES II considers only women of childbearing age (ages 15 to 46) and does not provide estimates separately for pregnant, nonbreastfeeding postpartum, and breastfeeding postpartum women.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

estimates.8 Taking the midpoint of the two estimates of medical risk does not seem appropriate. Since each dataset measures prevalence of medical risk based on different criteria, the combined number estimated at medical risk should not be lower than the larger of the two. For example, if NHANES measures anemia and finds 30 percent at risk and NMIHS measures miscarriages and finds 20 percent at risk on this basis, the combined estimate should be higher than 30 percent, since some who are anemic may not have also met the medical risk criteria on the basis of having a miscarriage. Furthermore, data from NHANES III alone indicate that 98.3 percent of women met at least one risk criterion. Yet the final estimate combining medical risk data from NMIHS is only 95.2 percent.

A second problem in the WES II study concerns the relationship between the estimates of nutritional risk for income eligible women and infants. Current WIC regulations state that an infant is automatically deemed at nutritional risk if the mother was at risk during pregnancy. Hence the risk of infants cannot be lower than the risk of pregnant women. Yet the Sigma One methodology has ignored this relationship. In their previous study, the nutritional risk for infants was estimated to be 72 percent, while the risk rate for pregnant women was 91.3 percent. This inconsistency led FNS in 1991 to revise their assumptions for infants to 95 percent. The WES II study estimates do not capture the WIC regulations in this respect.

Updating the Nutritional Risk Prevalence Estimates

It is the panel’s view that the estimates of nutritional risk prevalence for the categorically and income eligible WIC population currently in use should be reexamined. The recent efforts to standardize nutritional risk criteria across states and the availability of more recent data motivate a revision of the estimates of nutritional risk eligibility that are currently in use. Standardized anthropometric, medical, and predisposing risk criteria have already been adopted by the states, but they have not been incorporated into the model used to estimate the prevalence of nutritional risk

8  

An addition for the percentage found to be at dietary risk only (from NHANES III) and adjustments for age criteria that render an applicant nutritionally at risk (under age 18 and over age 36) results in a final estimate of 95.2 percent of income eligible women at nutritional risk.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

among income eligible groups. Recommendations concerning dietary risk criteria are to be released soon by the Institute of Medicine, but they will become standard only to the extent that they are adopted by FNS. If newly standardized nutritional risk criteria differ from the modal criteria used in WES II, it is possible that some substantial changes could occur in the proportions of women and children found to be at nutritional risk.

The dietary data used to establish the proportion of WIC income eligible individuals who are also at nutritional risk are outdated. The estimates currently in use were obtained from 1980 data. The WES II estimates, while somewhat more timely, are based on old dietary information as well; their values are derived from data from NHANES III, Phase I carried out during the 1988–1991 period. When new estimates of nutritional risk are developed, investigators should use the most recent suitable datasets. Possible datasets include the Continuing Survey of Food Intakes by Individuals (CSFII), which can provide an additional means of estimating the proportion of individuals at nutritional risk among those who are categorically and income eligible. The most current version is the 1994– 1996 survey, with a supplementary survey of children that was implemented in 1998. Phase II of NHANES III, which covers the 1991–1994 period is also available. NHANES IV data covering the period, of 1999–2001 is currently not available but will be within the next few years.

Recommendation: Estimates of nutritional risk should be reexamined with more recent data and with additional data sources and should take new state standards of nutritional risk into account whenever possible.

OVERALL CONCLUSION

This chapter has examined six methodological issues for estimating the number of persons eligible for WIC. The undercount of infants in the CPS results in an undercount of infants and women who are income eligible for WIC. Not fully accounting for adjunctive eligibility results in a substantial undercount of the number of people who are eligible for WIC. The use of annual income instead of monthly income also underestimates the number of people who are eligible for WIC at some point during the year. These three results all point to an undercount of the estimated number of eligible persons. The panel also considered three issues for which the effects of different methodological considerations on the total number of estimated

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×

eligibles were more ambiguous and small in size. Using a 6-month certification period for children instead of the 12-month certification period used currently in estimating eligibility results in an undercount of children of 4 percent if annual income is used, but in a 5 percent overcount if monthly income is used. The use of alternative definitions of the economic unit results in very little change in the total number of estimated eligibles, and the direction of the effect depends on the base with which estimates are being compared. Finally, the panel considered current methods for estimating the percentage of income eligible persons who are at nutritional risk and concluded that these estimates should be reexamined with more recent data.

Considering these findings in total, the panel concludes that current estimation methods result in a substantial understatement of eligible persons.

Conclusion: The panel concludes that current methods used to estimate eligibility for WIC substantially underestimate the number of people who are eligible.

The underestimation of eligibility implies that coverage rates are overstated. From the simulation results presented here, the number of infants estimated to be eligible for WIC is underestimated by a total of 54 percent—considering the undercount of infants in the CPS, adjunctive eligibility, and the use of monthly income instead of annual income. The latest coverage rate available for infants is 130.4 percent in 1999. If this rate is recalculated using the increased estimate of eligible infants, the coverage rate falls to 84.7 percent. Presumably the coverage rates of pregnant and postpartum women would also fall similarly. For children, the total underestimation of eligible people is 25 percent (considering an overcount of children in the CPS, adjunctive eligibility, the use of monthly instead of annual income, and a 6-month certification period). The 1999 coverage rate for children was 76.0 percent; when this rate is recalculated with the larger estimate of eligible children, then the coverage rate falls to 60.8 percent. Thus, coverage rates based on the panel’s estimates of eligibility would fall considerably if these estimates pass further scrutiny.

It is important to note that the underestimation of eligible people and subsequent overestimation of coverage rates do not necessarily mean that no ineligible persons are participating in WIC. The panel does not explore this possibility, for it is not part of our charge. We do note that the USDA has recently conducted a WIC income verification study and plans to release the results in late 2001.

Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 30
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 31
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 32
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 33
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 34
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 35
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 36
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 37
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 38
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 39
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 40
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 41
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 42
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 43
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 44
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 45
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 46
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 47
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 48
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 49
Suggested Citation:"4. Potential Biases in Eligibility Estimates." National Research Council. 2001. Estimating Eligibility and Participation for the WIC Program: Phase I Report. Washington, DC: The National Academies Press. doi: 10.17226/10158.
×
Page 50
Next: 5. Estimates of Full-Funding Participation »
Estimating Eligibility and Participation for the WIC Program: Phase I Report Get This Book
×
Buy Paperback | $29.00 Buy Ebook | $23.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Each year the U.S. Department of Agriculture (USDA) must estimate the number of people who are eligible to participate in the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC). These USDA estimates have come under critical scrutiny in part because the number of infants and postpartum women who have actually enrolled in the program has exceeded the number estimated to be eligible by as much as 20 to 30 percent. These high "coverage rates" have led some members of Congress to conclude that some people who participate are truly ineligible, and that funding could be reduced somewhat and still meet the needs of truly eligible persons who wish to participate. But some advocates and state WIC agencies believe that the estimates of the number of eligible persons are too low and more people who are eligible and want to participate could do so.

In response to these concerns, the Food and Nutrition Service (FNS) of the USDA asked the Committee on National Statistics of the National Research Council to convene a panel of experts to review the methods used to estimate the number of people nationwide who are eligible and likely to participate in the WIC program. The panel's charge is to review currently used and alternative data and methods for estimating income eligibility, adjunctive eligibility from participation in other public assistance programs, nutritional risk, and participation if the program is fully funded.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    Switch between the Original Pages, where you can read the report as it appeared in print, and Text Pages for the web version, where you can highlight and search the text.

    « Back Next »
  6. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  7. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  8. ×

    View our suggested citation for this chapter.

    « Back Next »
  9. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!