Skip Header

What Do We Know About The Undercount of Children?

Working Paper Number POP-WP039
Kirsten K. West and J. Gregory Robinson
Component ID: #ti659381514

There is a pressing need for statistics to inform policy makers of the size and characteristics of the nation's children. Decennial data provide most of these statistics either directly or indirectly when used as population controls for surveys. In this paper, we summarize what we know about the undercount of children. While considerable amount of attention has been devoted to the undercount for the U.S. population overall and to specific subgroups such as young black men in inner cities, less attention has been given to the undercount of children. We bring together data from the 1990 Post Enumeration Survey and Demographic Analysis results in an attempt to learn as much as we can from these sources about the coverage of children in the decennial census.


The authors wish to thank Patrick Cantwell, Roxanne Feldpausch, Alfredo Navarro, Michael Starsinic and Machell Town, Decennial Statistical Studies Division for assistance with special Dual System Estimates tabulations for the age group 0-17.

An earlier version of this paper was presented at the annual meeting of the Southern Demographic Association in Annapolis, Maryland, October 30-31, 1998.

This paper reports the results of research and analysis undertaken by Census Bureau staff. It has undergone a more limited review than official Census Bureau publications. This report is released to inform interested parties of research and to encourage discussion.

Component ID: #ti1421127017


In recent years, the interest in monitoring and understanding the lives of children has grown rapidly. Consequently, there is a pressing need for information on the number and characteristics of the nation's children. Decennial census data provide most of this information either directly or indirectly by serving as controls for surveys.

It is well known that censuses are subject to undercount. The 1990 Census was no exception. The count fell somewhat short compared with independent population estimates and post-enumeration survey (PES) results. The 1990 PES estimated the net undercount1 to be about 1.6 percent for the total population. Substantial differences occurred among population groups. For example, the net undercount was estimated to be 0.7 percent for non-Hispanic Whites, but 4.6 percent for Blacks, and 5 percent for Hispanics (Hogan, 1993). The undercount was also estimated to be higher in the nation's large cities and was disproportionately concentrated in the inner (central) areas of those cities.

The PES estimates indicate that more than half of the net undercount in the total population resulted from the net undercount of children in the age groups less than 18 (Census Bureau, 1997). Thus, in 1990, more than two million children were not counted. Furthermore, the distribution for undercounted children mirrored that for the total population in that the undercount differed for population subgroups. While the undercount rate for the non- Hispanic White population age 0-17 was estimated to be 2 percent, it was 7.1 percent for Black children and 5 percent for Hispanic children.

A differential undercount of children may mean potential inequities in fund allocations for programs and in the provision of social services for children and their families.2 A differential undercount may also affect how policies related to children are written, how data are interpreted and the conclusions drawn from studies on children.

In this paper, we summarize what we know about the undercount of children in the 1990 Census. In the first part of the paper, we present empirical data to examine how many children are undercounted and who they are. In the ensuing discussion, we offer potential explanations for the undercount of children drawing on what is known about causes of census coverage errors of the total population.

Component ID: #ti1421127016


Evaluation is an important element of the census process. Traditionally, the Census Bureau has depended on two methods to evaluate the completeness of coverage: 1) a PES methodology and 2) Demographic Analysis (DA). Each method is described briefly below.

The PES is a sample survey conducted after the census. The PES approach is based on a case-by-case matching of two population samples. One sample is drawn from an address list compiled independently of the census. This sample provides an estimate of gross underenumeration. Another sample is drawn from the census. This sample provides a basis for matching and for estimating correct and erroneous enumerations. The data from the samples are combined with the census to obtain dual system estimates (DSE) of the true population size and thus, of a net undercount. The DSE estimates are calculated separately for geographic and demographic subgroups thought to have the same probability of being captured in the census. These subgroups are also known as the post-strata. A separate undercount estimate is calculated for each post-stratum (Hogan, 1993).

The DA method relies on data sources essentially independent of the current census to reconstruct the population. Administrative data on annual births, deaths, and international migration, and analytical estimates of emigration and undocumented immigrants, and administrative records such as Medicare enrollment are combined to estimate the size of the overall population, as well as the population by age and by broad race categories. The undercount in the decennial census is the difference between the population size as estimated by the DA and the census count. The DA method is considered a useful tool for estimating the undercount at the national level and serves as a tool to measure changes in net undercount rates between censuses (Robinson, 1993). There is no sampling variance associated with the DA undercount estimates, although other errors may be present.

In the sections that follow, results from both evaluation methodologies are presented. The DA results show general and historical trends in the undercount of children. The PES-based results show details for 357 post-strata.3

Component ID: #ti1421127015


3.1 Historical trends in the undercount

The estimated net national undercount in each census from 1940 to 1990 is shown in Table 1 below and illustrated in Figure 1 (estimates based on Demographic Analysis). In each census, the undercount of Blacks has been disproportionately higher than the undercount of non- Blacks (e.g., a black undercount rate of 5.7 percent in 1990 compared with a non-black rate of 1.3 percent). The difference between the black and the non-black net undercount rate increased steadily between 1940 and 1970, starting at 3.4 percentage points in 1940, 3.6 percentage points in 1950, 3.9 percentage points in 1960, and reaching 4.3 percentage points in 1970. In 1980, the difference decreased to 3.7 percentage points, but in 1990, the difference was again more than 4 percentage points. Thus, in 1990, according to the DA, the rate of net undercount for Blacks is more than four times higher than for non-Blacks (Robinson et al., 1993.)

Table 1. Percent Net Undercount by Race (Demographic Analysis) in the 1940-1990 Censuses

  1940 1950 1960 1970 1980 1990
Total Population 5.4 4.1 3.1 2.7 1.2 1.8
Blacks 8.4 7.5 6.6 6.5 4.5 5.7
Non-Blacks 5.0 3.8 2.7 2.2 0.8 1.3
Difference: Black-Non-Black 3.4 3.6 3.9 4.3 3.7 4.4

Source: Robinson et al., 1993

3.2 Undercount of children by age and sex for Blacks and Non-Blacks

Undercount estimates derived from the DA vary by age and race more than they vary by sex, especially for children (Table 2.) Children younger than 10 are more likely to be undercounted than children in the older age groups. In fact, children in the age group 15-19 are likely to be overcounted (a negative sign indicates overcount). These findings by age appear across both race categories, black and non-black. However, children in the black category are more likely to be undercounted than children in the non-black category.

Table 2. Percent Net Undercount (Demographic Analysis) in the 1990 Census: Population less than 20 by Age Group, Sex and Race

Age Total for
Age Group
0-4 3.7 8.6 8.2 2.7 2.8
5-9 3.5 7.7 7.5 2.7 2.8
10-14 1.2 4.1 4.0 .5 .7
15-19 -1.7 - .2 .4 -2.3 -1.7

Source: Adopted from Robinson et al., 1993: Table 3

3.3 Undercount of children by more detailed characteristics

Next, we look at more detailed findings based on data from the 1990 Post-Enumeration Survey (PES). The percent net undercount and the associated standard errors (s.e.) for children in households classified as non-Hispanic white and other, black and non-black Hispanic are shown in Table 3 below. Within the race and ethnic background categories, the results are shown by tenure status (owner/renter), and degree of urbanization.

First, we see that the children in households classified as renters have higher undercount rates than children in the owner categories. Among non-Hispanic Whites, the undercount of renters is more than three times that of owners (4.0 versus 1.2 percent). A similar pattern is observed among Hispanics (7.5 versus 1.2 percent). Among Blacks the difference between renters and owners is not statistically significant. When location of the rental unit is taken into account, it is further noticed that Hispanic children in renter occupied units in rural areas have significantly higher undercount rates than their counterparts in urban areas (17.4 percent vs. 6.9 and 6.2 percent, respectively).

Among owners, the undercount rates are higher among Blacks than Whites and Hispanics. There is no discernable pattern for white, black or Hispanic children by degree of urbanization.

Table 3. Percent Net Undercount (Dual System Estimates) in the 1990 Census:
Population Less than Age 18 by Race, Tenure and Urbanization

% (s.e.)
% (s.e.)
% (s.e.)
Renter 4.0 (0.67) 8.1 (1.16) 7.5 (1.33)
 Urbanized Areas 250,000+ 3.8 (1.08) 8.5 (1.39) 6.9 (1.63)
 Other Urban 3.5 (0.90) 6.7 (1.62) 6.2 (1.95)
 Rural 4.8 (1.74) 8.7 (7.89) 17.4 (5.73)
Owner 1.2 (0.36) 5.6 (1.16) 1.2 (1.16)
 Urbanized Areas 250,000+ 0.8 (0.50) 6.1 (1.48) 1.9 (1.02)
 Other Urban 1.0 (0.48) 4.1 (1.72) -0.3 (1.98)
 Rural 1.8 (0.79) 5.5 (3.29) 0.6 (6.93)

Source: U. S. Bureau of the Census: Unpublished data

The undercount for children in Asian and Pacific Islander households and for American Indians on reservations are presented in Table 4. For Asian and Pacific Islanders, the undercount for children in the renter category is 8 percent. In comparison, for children in the owner category the undercount is -0.5 percent. Compared with those of other races, children in American Indian households on reservations have high undercount -- 13.8 percent.

Table 4. Percent Net Undercount (Dual System Estimates) in the 1990 Census:
Population Less than Age 18 in Asian and Pacific Islanders Households by Tenure
and in American Indian Households

% (s.e.)
Asian and Pacific Islander 3.3 (2.09)
 Owner -0.5 (2.55)
 Renter 8.0 (3.54)
American Indian on Reservations 13.8 (5.00)

Source: U. S. Bureau of the Census: Unpublished data

In summary, for children ages 0-17, the pattern in the data suggests a higher undercount of children in American Indian, Hispanic and black households than in white households. When examined by tenure status, children in the renter category are undercounted by a higher percentage than children in the owner category-the differential is statistically significant for all race/ethnic categories except black. The rural areas often have higher estimated undercounts than do the urban areas, but the differences are only statistically significant for Hispanic renters. An undercount in excess of 10 percent is reported for two post-strata: children on American Indian reservations, and children in the non-black Hispanic renter category in rural areas.

More detailed regional data are shown in Appendix Table 1. From inspection of this table, black children in the owner category in large urbanized areas in the North East and in the renter category in large urbanized areas in the West also have undercount in excess of 10 percent. Sixteen out of 51 post-strata show undercount ranging from 5 to 10 percent.

3.4 The undercount of children compared with the undercount of the total population

The net undercount rates for children are compared to the undercount rate for the total population in Table 5 and in Figure 2. A disproportionate undercount of children is measured by both the DA and the PES results. This differential is found for both Blacks and non-Blacks.

Table 5. Percent Net Undercount by Race in the 1990 Census (Demographic Analysis and Dual System Estimates)

  Total Blacks Non-Blacks
Demographic Analysis:      
Total Population 1.8 5.7 1.3
Ages 0-9 3.6 8.1 2.8
Dual System Estimates:      
Total Population 1.6 4.4 1.2
Ages 0-17 3.2 7.0 2.5

Figure 2. Percent Net Undercount by Race

The PES results allow us to look more closely at the patterns of undercount of children compared to the total population. The net undercount for children by post-stratum group including region is presented in Appendix Table 1. The net undercount for the total population is presented in Appendix Table 2. In this section, we compare the pattern of the undercount shown in the two tables.

First, we examine the undercount patterns by region for non-Hispanic Whites and Others. In the South, children show significantly higher undercount rates. Among owners in the South, children in other urban areas show significantly higher undercount rates than the total population (2.3 versus 0.5 percent), and, among renters, the undercount for children in the large urbanized areas is high and much higher than the undercount for the total population (8.2 versus 2.5 percent). In the West, children in the combined owner stratum also have a higher undercount than the total population (2.1 versus 0.0 percent).

Next, we examine the pattern for Blacks.4 For all regions combined, the undercount rate of black children in each post-stratum is consistently higher than for the total population (7.1 percent versus 4.6 percent). The undercount of black children is significantly higher than for the total population in the owner post-strata (5.6 percent versus 2.3 percent for all regions combined). For black owners in the North East, where the total undercount is less than two percent, the comparable factor for children is 10.6 percent. For black renters in the large urbanized areas in the West, the already high undercount rate for the total population (10.0 percent) becomes even higher when focusing on children (13.6 percent).

For Hispanics, for Asian and Pacific Islanders and for American Indian households on reservations, the patterns observed for the children are identical to the patterns observed for the total population, with renters showing higher undercount than owners and children having a slightly higher undercount than the total population. The regional variation is low, and the differences are not statistically significant.

Component ID: #ti1421127014


In recent years, a number of efforts have been made to understand the causes of undercount. Several studies have examined coverage errors in the total population and for specific subpopulations (Brownrigg, 1991, de la Puente, 1993, Fein and West, 1989). There has been no systematic attempt to look at reasons for undercounting children. Thus, it is within the framework of the coverage errors and total population that we examine the results for children. Coverage errors take the form of both omissions and erroneous inclusions. In order to understand better how these errors might occur, the discussion begins with a brief overview of the Census Bureau's residence rules.

4.1 Census Bureau residence rules

The Census is based on contacting each household in the nation rather than every individual in the population. The Census Bureau conducts the census under the de jure residence principle. In other words, householders are asked to report on behalf of themselves and the other members of their households, that is, those who reside in their home as opposed to the de facto approach that would enumerate people where they are on a specified date. To help respondents understand what it means for somebody to reside in a household on a given date, census forms have rules listed and guidelines for the respondent to follow.

The 1990 Census form used an "include-exclude" list of items to instruct the respondent how to roster the household. The respondents were instructed to include a newborn baby still in the hospital, a foster child, a roomer, a housemate, a boarder, anyone staying with the household on Census Day with no other permanent place to live, and anyone staying with the household most of the time while working, even if he or she had another place to live. Similarly, the respondent was instructed not to include anyone living away while attending college, in a correctional facility, nursing home, or mental hospital on Census Day, in the Armed Forces and living somewhere else, or anyone who lives or stays somewhere else most of the time.

Thus, in order for the respondent to provide a complete and accurate roster of the people who live at a given address on Census Day, he or she must be capable and willing to follow the above guidelines. To the extent that the respondent has trouble reading, has difficulty with the language in which the form is printed, or lives in a household structure that does not clearly fit the Census Bureau's guidelines, coverage errors are likely to occur. To the extent that the respondent is unwilling to follow the Census Bureau's guidelines, coverage errors are likely to occur. These errors may disproportionately affect children.

4.2 Omissions

At the household level, four different sources of omissions can be distinguished: 1) omission of some household members in an otherwise enumerated household, also referred to as "within household nonmatch;" 2) omission of an entire household in an enumerated housing unit, also referred to as "whole household nonmatch;" 3) omission of a household because the housing unit in the building was missed, although the building was counted in the census, or 4) omission because the building and all its housing units were missed by the census.

Table 6 presents the distribution of nonmatches by these categories. It also shows census processing errors.5 It should be noted that standard errors are not available for these statistics. The results may not express statistically significant differences, and therefore, only the general patterns in the data are discussed. As seen in the table, for the total population in 1990, approximately one third of the cases could be characterized as omission of a person, another third as omission of a whole household, and, finally, close to one third as omissions involving housing units and housing structures. Five percent of the error is estimated to be processing errors. All race categories follow this pattern, but Blacks and Asian and Pacific Islanders have a higher percentage individual person and whole household omissions than Whites and Hispanics--close to 75 percent compared to around 60 percent.

In the categories involving missed housing units, Hispanics have the highest percentage followed by Blacks and Asian and Pacific Islanders. Whites have the lowest percentage. The distribution of missed housing structures indicates this to be an important reason for missing non-Hispanic Whites (27.7 percent). Among Asian and Pacific Islanders this category of cases accounts for only 8 percent.

Table 6. Percent Distribution of Types of Nonmatches and of Total Resolved Cases

Population Percent of Total Nonmatches
Total 30.5 33.9 8.5 22.0 5.1
Non-Hispanic White 27.1 34.1 6.4 27.7 4.3
Black 36.1 39.5 10.1 10.9 3.4
Hispanic 33.0 27.0 14.0 16.0 10.0
Asian/Pac. Islander 34.7 40.0 10.7 8.0 6.7

Source: Hogan, 1993, Table 6

4.3 Erroneous inclusions

While the ultimate interest is in undercount, it should be kept in mind that the estimates shown are estimates of net undercount. Net undercount is a function of both erroneous omissions and inclusions. The percent distribution of erroneous enumerations is provided in Table 7 below. The "other counting error" category is the largest--38 percent. Most of these "other counting errors" stem from enumerations of people that moved to the address after Census Day. These people should have been counted elsewhere. About 28 percent of the erroneous inclusions can be attributed to duplicates. Another 21 percent is listed as unmatchable because the census enumeration did not provide enough information for matching. A small number of erroneous errors are attributed to geocoding error (6 percent) and fictitious records (2.6 percent). A geocoding error occurs if a person was enumerated in an area that was more than two blocks away from the block where he or she should have been enumerated.

Table 7. Percent Distribution of Types of Erroneous Inclusions

  Total Percent of Total
Total 5.8 100.0
Duplicate 1.6 28.2
Fictitious .2   2.6
Geocoding error .3   6.0
Other counting error 2.2 38.0
Unmatchable 1.2 20.8
Imputed .3   4.5

Adopted from Hogan, 1993, p.1057

The coverage errors are discussed below in more detail as they relate to the undercount of children.

4.4 Coverage errors of children

Coverage errors are likely to occur because the respondent has difficulty rostering his or her household. Unique household and living arrangements may contribute to this difficulty. There are many living arrangements involving children that could fall in this category. Some examples are:

  1. a child who resides in a diverse household structure and in a unique living arrangement among multiple nuclear families. For example, it is estimated that in 1990, as many as two million children lived in households maintained by their grandparents (Casper and Bryson, 1998).
  2. a child who resides in a situation where he or she is only loosely connected to the household; perhaps without a stable place of residence for long periods of time.
  3. a child who is not a biological offspring or otherwise related to the respondent.
  4. a child who has more than one residence.
  5. a child living away at a boarding school or perhaps at a college part of the year.
  6. a child in a hospital or an institution.

While most of these factors can result in census undercount, factors such as d), e) and f) may also lead to erroneous inclusions. As noted earlier, according to Demographic Analysis results, children in the age group 15-19 appear to be overcounted (see Table 2).

Difficulty with the census questionnaire may also lead to coverage errors for children. The census rules of residence instruct that the person in whose name the house or apartment is owned, being bought, or rented be listed as person 1 on the form. The respondent is then asked to identify members of the household in relation to person 1. This often contradicts the respondent's notion of family or household. Family may also be interpreted as meaning "all family." This may cause erroneous inclusions (Martin and Griffin, 1995). Finally, in 1990, while the roster of household members requested at the beginning of the questionnaire had room to list 12 persons, the form itself allowed for detailed data on only seven persons. To the extent that children are listed last, a respondent with a large family may fail to include them.

Assuming the respondent has no difficulty understanding the Census Bureau's residence rules and whom to list as usual residents of the households, the respondent must still be willing to do so. Historically, a number of factors have hampered attempts to count certain subgroups in the population. In situations where several families live together to share the rent, but do so against housing contract rules, the respondent may not be willing to report members who do not appear on the lease. Listing some members of the household may have other negative consequences. For example, a respondent may fear that disclosure of certain members of the household will affect eligibility for social services, that children illegally in the country will be deported, or that the whereabouts of a child in hiding from a custodial parent will be detected. There are many situations where the Census Bureau's promise of confidentiality of the data is not trusted (Bailar and Martin, 1987; Hainer et al, 1988; Fein and West, 1988).

Not all coverage errors stem from omissions or erroneous inclusions of people. As seen in Table 7 above, about a third of the omitted cases were classified as coverage errors associated with the housing unit. Housing units are missed because they were never on the address list used in the Census, or because they were erroneously deleted from the list. Entire households could also be missed if housing units are misclassified as vacant rather than occupied. Sometimes a housing unit is missed because it is difficult to locate the dwelling (hidden from public view). This might happen in rural areas as well as highly urbanized areas. Sometimes a housing unit is missed because it is not easily recognized as such. For example, converted garages or basement apartments may be difficult to detect. To the extent that households that reside in housing units that fit these descriptions have more children than households in other types of structures, differential undercount of children is expected.

Component ID: #ti1421127013


In the 1990 Census, the PES estimated a net undercount of the population that amounted to more than four million persons. PES estimated that more than half of these persons were children under the age of 18. DA results suggest a somewhat lower proportion of the total net undercount as being undercount of children, but both methods agree that the undercount of children is disproportionate to their share in the population. The PES estimated that in the 1990 Census, children in the following categories were more likely to be undercounted than their counterparts:

  • renter occupied units
  • American Indian, Hispanic and black households
  • rural areas or large metropolitan areas

Furthermore, estimates generated by DA suggest that differential coverage errors are not associated with the gender of a child, but that age of child is an important factor. Children under 10 are more likely to be undercounted than older children. Children, age 15-19, may be overcounted.

Statisticians and politicians agree: it is a challenge to conduct the census. Not only do we have to count the nation's children, we have to count them in the households and at the addresses where they are defined to reside on Census Day. Many factors contribute to the undercount of children and to the differential patterns of undercount. Possible explanations for differential undercount of children mirror the explanations offered for coverage errors in the general population. The propensity for coverage errors for children may be exacerbated if the types of households, the types of living arrangements and the types of housing units we are most likely to miss in the census include a disproportionate number of children.

Component ID: #ti1421127012


Bailar, Barbara A. and Elizabeth A. Martin. (1987). "Report on Meetings in Los Angeles, Chicago and Denver." Unpublished Census Bureau Memorandum.

Brownrigg, Leslie A. (1991). "Irregular Housing and the Differential Undercount of Minorities." Paper prepared for the Census Advisory Committee Meetings in Alexandria, Virginia, November 13-15. United States Department of Commerce, Bureau of the Census.

Casper, Lynne M. and Kenneth R. Bryson. (1998). "Co-Resident Grandparents and Their Grandchildren: Grandparent Maintained Families." Paper presented at the Annual Meeting of the Population Association of America, Chicago, Illinois, May.

Census Bureau. (1997). "Report to Congress The Plan for Census 2000." United States Department of Commerce, Bureau of the Census, July.

Childers, Danny R. (1993). "The Impact of Housing Unit Coverage on Person Coverage." Housing Unit Coverage Study (HUCS) Results Memorandum Number 2. Distributed with cover memo from Ruth Ann Killion to Thomas C. Walsh (June 24, 1993). United States Department of Commerce, Bureau of the Census.

Citro, Constance F. and Michael L. Cohen (eds.). (1995). The Bicentennial Census. Washington, D.C., National Academy Press.

de la Puente, Manuel. (1993). "Why are People Missed or Erroneously Included by the Census: A Summary of Findings from Ethnographic Coverage Reports." Report prepared for the Advisory Committee for the Design of the Year 2000 Census Meeting, March 5. United States Department of Commerce, Bureau of the Census.

Fein, David J. and Kirsten K. West. (1988). "The Sources of Census Undercount: Findings from the 1986 Los Angeles Test Census." Survey Methodology, Vol 14, Number 2, 223-241.

Hainer, Peter, Catherine Hines, Elizabeth Martin, and Gary Shapiro. (1988). "Research on Improving Coverage in Household Surveys." Proceedings of the Fourth Annual Research Conference. United States Department of Commerce, Bureau of the Census: 513-539.

Hogan, Howard. (1993). "The 1990 Post-Enumeration Survey: Operations and Results." Journal of the American Statistical Association, Vol. 88, No 423: 1047-1061.

Martin, Elizabeth A. and Deborah H. Griffin. (1995). "The Role of Questionnaire Design in Reducing Census Coverage Error." U.S. Bureau of the Census. Working Paper in Survey Methodology, SM95/08.

Robinson, J. Gregory, Bashir Ahmed, Pritwis Das Gupta, and Karen Woodrow. (1993). "Estimation of Population Coverage in the 1990 United States Census Based on Demographic Analysis." Journal of the American Statistical Association, Vol. 88, No. 423: 1061-1077.

Component ID: #ti1421127011


  1. The term 'coverage error' refers to both undercount and overcount. Overcount is defined as the erroneous inclusion or double counting of an individual as opposed to undercount or failure to include an individual. Net undercount is the difference between the overcount and the undercount.

  2. Every year, billions of dollars are allocated by the federal government or by states to aid their local governments. Many allocations are formula based, and particularly for small areas such as counties and school districts, are based on estimates from the census or on surveys that employ census controls. Research on the specific effects of errors in the census on the distribution of federal funds to an area is limited. Much will depend on how the fund allocation formulas are written. However, in general, two factors determine the extent to which the errors result in inequitable distributions of monies. First, whether the funds are distributed on a per capita basis or as shares of a fixed total sum. If they are distributed on a per capita basis, errors in coverage have direct effects. If they are allocated based on the share of population, inequitable distribution of funds will occur only if the eligible areas experience significant different rates of net undercoverage. Second, whether other factors beside the population count are used to determine fund distribution. If other factors such as eligibility thresholds dominate the formula, errors in coverage have less effect (Citro and Cohen, 1985).

  3. The method used in 1990 to estimate adjusted counts down to the census block level consisted of defining 1,392 post-strata on the basis of age, sex, race or ethnicity, tenure, type of place and geographic location. It was a criticism of the census adjustment estimates that the post-strata were too heterogeneous, especially geographically. Thus, in July of 1992, the original 1,392 strata were revised into 357 homogeneous groups. The new stratification reduced the number of cells with small sample sizes and high variances These post-strata were considered for adjusting the postcensal base. (For a full discussion of the stratification scheme, see Hogan, 1993).

  4. For Blacks and non-black Hispanics, separate estimates do not exist for Other Urban and Rural Areas by geographical region.

  5. An example of a census processing error would be a failure to count a valid questionnaire that was returned.

Component ID: #ti1421127010

Appendix Table 1. Percent Undercount And Standard Errors by Post-Stratum: Children 0-17

Post-Stratum North East South Midwest West Total
Non-Hispanic White and Other 0.5 (0.66) 3.4 (0.59) 1.1 (0.62) 2.4 (0.69) 2.0 (0.32)
  Owner -0.3 (0.65) 2.2 (0.61) 0.6 (0.77) 2.1 (0.85) 1.2 (0.36)
    Urbanized Areas 250,000+ -1.0 (1.05) 1.9 (0.95) 0.4 (0.64) 1.7 (1.33) 0.8 (0.50)
    Other Urban -0.8 (0.91) 2.3 (0.88) 0.4 (0.68) 1.5 (1.42) 1.0 (0.48)
    Rural 1.0 (1.03) 2.3 (1.12) 0.9 (1.97) 4.0 (1.38) 1.8 (0.79)
  Renter 2.6 (1.77) 6.3 (1.37) 2.7 (0.92) 3.0 (1.29) 4.0 (0.67)
    Urbanized Areas 250,000+ 0.6 (1.81) 8.2 (2.30) 4.7 (1.92) 1.2 (2.18) 3.8 (1.08)
    Other Urban 3.1 (2.19) 3.8 (2.05) 2.8 (0.94) 4.2 (1.64) 3.5 (0.90)
    Rural 6.8 (6.60) 6.6 (3.01) 0.3 (1.87) 6.7 (2.78) 4.8 (1.74)
Black 7.9 (1.97) 6.3 (1.19) 6.4 (1.40) 11.7 (3.29) 7.1 (0.87)
  Owner 10.6 (4.56) 4.6 (1.28) 3.2 (1.17) 9.2 (5.19) 5.6 (1.16)
    Urbanized Areas 250,000+ 11.2 (5.21) 4.3 (1.37) 3.0 (1.37) 9.9 (5.82) 6.1 (1.48)
    Other Urban --------------- --------------- --------------- --------------- 4.1 (1.72)
    Rural --------------- --------------- --------------- --------------- 5.5 (3.29)
  Renter 6.7 (1.93) 7.6 (1.74) 8.2 (2.04) 12.9 (4.04) 8.1 (1.16)
    Urbanized Areas 250,000+ 6.6 (2.10) 7.8 (2.48) 8.5 (2.41) 13.6 (4.41) 8.5 (1.39)
    Other Urban --------------- --------------- --------------- --------------- 6.7 (1.62)
    Rural --------------- --------------- --------------- --------------- 8.7 (7.89)
Non-Black Hispanic 4.2 (3.96) 5.6 (1.31) 2.5 (1.99) 5.1 (1.03) 4.9 (0.95)
  Owner -0.9 (6.44) 1.7 (1.40) -3.8 (1.83) 2.1 (1.12) 1.2(1.16)
    Urbanized Areas 250,000+ -1.2 (7.85) 2.9 (1.23) -6.0 (2.43) 3.2 (1.16) 1.9 (1.02)
    Other Urban --------------- --------------- --------------- --------------- -0.3 (1.98)
    Rural --------------- --------------- --------------- --------------- 0.6 (6.93)
  Renter 5.6 (4.49) 9.4 (2.09) 7.7 (3.10) 7.2 (1.51) 7.5 (1.33)
    Urbanized Areas 250,000+ 5.3 (5.08) 9.4 (2.98) 7.1 (4.28) 6.4 (1.92) 6.9 (1.63)
    Other Urban --------------- --------------- --------------- --------------- 6.2 (1.95)
    Rural --------------- --------------- --------------- --------------- 17.4 (5.73)
Asian and Pacific Islander 3.4 (2.12) 2.7 (2.05) 3.0 (2.07) 3.4 (2.11) 3.3 (2.09)
  Owner --------------- --------------- --------------- --------------- -0.5 (2.55)
  Renter --------------- --------------- --------------- --------------- 8.0 (3.54)
American Indians on Reservations --------------- --------------- --------------- --------------- 13.8 (5.00)

Component ID: #ti1421127009

Appendix Table 2. Percent Undercount and Standard Errors By Post-Stratum: Total Population

Post-Stratum North East South Midwest West Total
Non-Hispanic White and Other -0.3 (0.55) 1.3 (0.37) 0.2 (0.39) 1.3 (0.53) 0.7 (0.22)
  Owner -1.5 (0.62) 0.5 (0.38) -0.4 (0.43) 0.0 (0.41) -0.3 (0.23)
    Urbanized Areas 250,000+ -2.1 (1.08) 0.7 (0.71) -0.3 (0.39) -0.3(0.65) -0.5 (0.38)
    Other Urban -1.1 (0.49) 0.5 (0.42) -0.1 (0.40) 0.6 (0.58) 0.1 (0.23)
    Rural -0.5 (0.70) 0.2 (0.69) -0.7 (1.18) 0.3 (0.69) -0.2 (0.47)
  Renter 2.5 (1.10) 3.6 (0.93) 2.0 (0.85) 3.9 (1.17) 3.1 (0.50)
    Urbanized Areas 250,000+ 1.2 (1.39) 2.5 (1.48) 2.3 (1.61) 3.2 (1.62) 2.3 (0.75)
    Other Urban 3.4 (1.51) 3.2 (1.74) 1.2 (1.09) 4.5 (1.34) 2.9 (0.77)
    Rural 6.5 (4.20) 6.2 (1.71) 2.9 (1.51) 6.1 (1.81) 5.3 (1.13)
Black 5.8 (1.13) 3.9 (0.76) 3.6 (0.85) 8.0 (1.80) 4.6 (0.53)
  Owner 1.7 (1.72) 2.4 (0.73) 0.9 (0.74) 5.7 (1.72) 2.3 (0.56)
    Urbanized Areas 250,000+ 1.6 (1.91) 2.2 (0.90) 0.8 (0.87) 6.1 (1.91) 2.2 (0.63)
    Other Urban --------------- --------------- --------------- --------------- 1.3 (0.98)
    Rural --------------- --------------- --------------- --------------- 3.5 (1.90)
  Renter 8.1 (1.50) 5.4 (1.30) 5.7 (1.43) 9.4 (2.47) 6.5 (0.82)
    Urbanized Areas 250,000+ 8.4 (1.61) 6.3 (1.90) 6.0 (1.68) 10.0 (2.72) 7.3 (0.98)
    Other Urban --------------- --------------- --------------- --------------- 4.2 (1.18)
    Rural --------------- --------------- --------------- --------------- 4.6 (5.33)
Non-Black Hispanic 5.4 (2.74) 5.5 (1.10) 2.8 (1.65) 4.9 (0.97) 5.0 (0.77)
  Owner 0.9 (3.63) 2.2 (0.77) -2.3 (1.73) 2.4 (0.74) 1.8 (0.67)
    Urbanized Areas 250,000+ 0.7 (4.45) 2.5 (0.90) -4.3 (2.58) 2.9 (0.87) 2.0 (0.72)
    Other Urban --------------- --------------- --------------- --------------- 0.9 (1.64)
    Rural --------------- --------------- --------------- --------------- 2.7 (2.69)
  Renter 6.8 (3.18) 9.3 (1.92) 7.2 (2.47) 6.7 (1.51) 7.4 (1.18)
    Urbanized Areas 250,000+ 6.7 (3.51) 9.3 (2.59) 6.6 (3.26) 5.9 (1.84) 7.0 (1.36)
    Other Urban --------------- --------------- --------------- --------------- 6.6 (2.74)
    Rural --------------- --------------- --------------- --------------- 15.8 (5.01)
Asian and Pacific Islander 2.9 (1.45) 2.4 (1.35) 2.9 (1.44) 2.1 (1.32) 2.4 (1.36)
  Owner --------------- --------------- --------------- --------------- -1.5 (1.50)
  Renter --------------- --------------- --------------- --------------- 7.0 (2.52)
American Indians on Reservations --------------- --------------- --------------- --------------- 12.2 (4.73)

  Is this page helpful?
Thumbs Up Image Yes    Thumbs Down Image No
Comments or suggestions?
No, thanks
255 characters remaining
Thank you for your feedback.
Comments or suggestions?
Back to Header