APPENDIX B
Comparison of Census and CPS Estimates of Poverty
There are many similarities and some differences between the CPS and the census systems for measuring poverty. They both base poverty estimates on the official definition of poverty for the nation, which compares the income of a family (or an unrelated individual) to a given poverty threshold, but they differ in the amount of data they collect, how they collect and process the data, and the frequency of data collection.
1990 CENSUS
A census of the U.S. population is conducted once every 10 years, and 1990 is the most recent one. In the 1990 census, income data—the basis for measuring poverty—were collected from a sample of 15 million households: a sample of about I in 6 households spread systematically across the country, except that very small counties and places (with estimated 1988 populations under 2,500) were sampled at a 1-in-2 rate, and very populous census tracts (or equivalent areas) were sampled at a 1-in-8 rate.
Data are collected in the census mainly by self-enumeration, whereby respondents fill out questionnaires received in the mail. Enumerators follow up those households that fail to return a questionnaire and collect the information through direct interviews. In 1990 approximately 74 percent of U.S. households returned their questionnaires with some or all of the requested information (Edmonston and Schultze, 1995:189). Data from the balance of the population were obtained by personal interviews. The follow-up enumerators are usually inexperienced temporary workers who are given very limited training.
The income data in the 1990 census are based on eight questions on various
components of income. (The census form also included a total income question, which was intended to permit respondents to enter a single amount if they could not provide amounts by source.) Nonresponse rates are higher for income than for most other items in the census. When household income information is missing, the Census Bureau uses statistical techniques to impute it on the basis of nearby households with similar characteristics. For the 1990 census, on average, 19 percent of aggregate household income was imputed (Edmonston and Schultze, 1995:387).
All censuses are subject to undercount—that is, failure to count everyone. There are no direct estimates of the undercount for poor children. For 1990, the net undercount was estimated at 1.8 percent for the total population, but there were substantial differences among population groups. For example, the net undercount was estimated at 5.7 percent for blacks and 1.3 percent for nonblacks. The net undercount also varied significantly by age: black girls and boys aged 5–9 were missed at a rate of 7.5 percent and 7.7 percent, respectively. Almost two-thirds of the estimated omitted population consisted of two age groups: children under age 10 and men aged 25–39 (Robinson et al., 1993:13). The undercount was also higher in large cities than in other areas, and it was disproportionately concentrated in the inner areas of those cities. These are also the areas where poverty is high. Thus, it seems likely that the undercount for poor children aged 5–17 is larger than the undercount of all children aged 5–17.
CURRENT POPULATION SURVEY
The CPS is a monthly labor force participation survey. For the period from 1990 to 1994, about 60,000 housing units were eligible for interview every month, and about 57,400 of them were found to be occupied by households eligible for interview.1 Of these 57,400 households, an interview was not obtained for various reasons for about 2,600 households—a noninterview rate of 4.5 percent.
Part of the CPS sample is changed each month: in the rotation plan, three-fourths of the sample is common from one month to the next, and one-half is common for the same month a year earlier. Each March, supplementary questions are asked about money income received the previous year. To obtain more reliable income data for the Hispanic-origin population, all November CPS households with one or more Hispanic persons are reinterviewed in March if they still include a Hispanic person. This procedure adds about 2,500 Hispanic households to the sample in March.
The CPS sample design, which is a multistage probability sample design, is revised about once every 10 years on the basis of the results of the latest census.
1 |
Starting in 1996, about 50,000 households nationwide (a sample of about 1 in 2,000 households) were eligible for interview every month—a reduction of about 17 percent from the early 1990s. |
From 1986 to 1994, the CPS sample design included 729 sample areas consisting of about 1,300 counties. These areas were chosen on the basis of 1980 census data to represent all 3,141 counties (in 1990) and independent cities in the 50 states and the District of Columbia. A design based on the 1990 census was phased in between April 1994 and July 1995: it included 792 sample areas consisting of about 1,300 counties, chosen to represent all 3,143 counties (in 1994) and independent cities in the 50 states and the District of Columbia.
In general, larger states have larger CPS sample sizes. The largest states, however, have CPS sample sizes that are smaller than their proportionate share of the U.S. population, and the smallest states have proportionately larger sample sizes. For example, California, with 12.2 percent of the U.S. population, has 9.9 percent of the CPS sample; Wyoming, with 0.18 percent of the U.S. population, has 1.3 percent of the CPS sample. This sample design means that estimates of poverty rates in large states are generally more precise than those in smaller states. The largest states, however, have larger relative errors due to sampling variability than would be expected if the CPS sample were allocated to the states in proportion to their population; the reverse holds true for smaller states.
The sample is designed to meet specific reliability criteria for the nation, each of the 50 states and the District of Columbia, and the substate areas of New York City and the Los Angeles-Long Beach metropolitan area. The CPS is not designed for direct county estimates. More than one-half of U.S. counties do not have sample households in the survey.
The CPS is carried out by permanent, experienced, and well-trained interviewers, initially by personal direct interviews, with subsequent interviews by telephone. For the March Income Supplement, the CPS asks household respondents about their money income received during the previous year, using a detailed set of questions for identifying about 28 different sources. About 20 percent of aggregate household income is imputed (about the same percentage as in the census)—that is, the data are missing and therefore constructed from information from similar households (Citro and Kalton, 1993:Table 3-6).
Like other household surveys, the CPS exhibits population undercoverage at higher rates than the census itself. The coverage ratios for the CPS show the magnitude of the population undercoverage relative to the census. Coverage ratios are defined as the estimated survey population before ratio adjustment to census-based population controls divided by the census-based population controls. (Beginning with the March 1994 CPS, the population controls reflect an adjustment for the undercount in the census itself.) For March 1994, the ratio of the CPS estimated population to the population control total (all ages) was 92 percent; for the age group 0–14 and the age group 15–19 years, the ratios were 94 percent and 88 percent, respectively (Bureau of the Census, 1996:Table D-2).
CPS undercoverage is corrected by ratio adjustments to the survey weights that bring the CPS estimates of population in line with updated national population controls by age, race, sex, and Hispanic origin. However, the ratio adjust-
ments do not correct for other characteristics on which the undercovered population might be expected to differ from the covered population. For example, the ratio adjustments reweight equally the sample households within an age-race-sex-Hispanic origin category, when research suggests that it is likely that lower income households within a category are more poorly covered than higher income households.
DIFFERENCES BETWEEN CENSUS AND CPS DATA
In comparing the census and the CPS as data sources for estimates of income and poverty, one difference is the definition of the universe for which the numbers in poverty are estimated; see Table B-1. Residence rules differ for the two data sources: in the census, students attending colleges away from their parental homes are counted at their college location; in the CPS, they are usually counted at their parental home. Also, the census excludes all unrelated individuals under age 15 in households from the poverty universe; the CPS excludes only those unrelated individuals under age 15 in households who are not part of an unrelated subfamily. (Unrelated subfamilies are made up of people who are related to each other but not to the householder, such as the family of a resident employee.) For
TABLE B-1 Poverty Universes for the 1990 Census and the March 1990 CPS
Component |
1990 Census |
March 1990 CPS |
||
Total resident U.S. population |
248,709,873 |
250,180,762a |
||
|
Population not covered in CPS |
-3,998,221 |
-3,989,762 |
|
|
|
Institutionalized |
-3,334,018 |
—b |
|
|
Armed Forces in barracks |
-589,700 |
—b |
|
|
Unrelated individuals under 15 in group quarters |
-74,503 |
—b |
|
Unrelated individuals under 15 in households |
-780,235c |
-199,000 |
|
|
College dormitory residents |
-1,953,558 |
0d |
|
Poverty universe |
241,977,859 |
245,992,000 |
||
a Reflects 1980 census-based population estimates, which estimated a higher resident population in 1990 than the 1990 census. b Intercensal estimates used to derive CPS population controls for survey weighting are not available by component. The sum of these three components was 3,989,762. c The large difference between census and CPS estimates reflects the census practice of excluding children under age 15 in unrelated subfamilies from the poverty universe. d The CPS includes household members away at school who are living in dormitories as family members and includes them in the poverty universe. SOURCE: Data from Housing and Household Economic Statistics Division, Bureau of the Census. |
estimates of related children in poverty (who are children in families but not children in unrelated subfamilies), the CPS and census poverty universes differ less than the CPS and census poverty universes for the total population. Essentially, the only difference is the treatment of college students in dormitories.
Overall imputation rates for income nonresponse are about the same in the census and the CPS, and the amount of income imputed is similar. Table B-2 presents aggregate income estimates for different income components from the two sources. The more detailed set of income questions in the CPS—28 compared with 8 in the census—and the direct interviewing methodology in the CPS would be expected to provide more comprehensive and accurate income data. It is believed that overreporting of some components of income, such as wages and salaries, occurs in the self-reported census data. (Another reason for the overestimate of wages and salaries in the census, compared with the independent benchmark, may be the editing procedures that were applied to responses to the total income question.)
The net effect of the differences between the CPS and the census in data collection, processing, and other aspects of the two systems is that there are dif-
TABLE B-2 Household Income by Type, 1989:1990 Census and March 1990 CPS
ferences between them in estimates of income and poverty. Aggregate estimates of total income and income by type when compared with independent benchmarks differ between the CPS and census. For example, total income from the 1990 census (for income types for which independent estimates can be constructed) is 91.6 percent of the benchmark; the corresponding figure from the March 1990 CPS is 88.7 percent. For income from such sources as Social Security and Railroad Retirement and retirement, disability, and survivor income, the CPS is closer to the benchmark than the census.
Estimates of median household income in 1989 by state differ between the March 1990 CPS and the 1990 census by amounts that are statistically significant for 18 states; see Table B-3. For 15 of the 18 states, the census estimates are higher than the CPS estimates; for some states, the differences are as much as 10–15 percent. For three states, the census estimates of median household income are significantly lower than the CPS estimates, in the range of 6–8 percent. Estimates of poverty rates by state also differ between the CPS and the census: statistically significant differences are observed for seven states; for six of them, the census poverty rates are significantly higher; see Table B-4.
Estimates of poverty rates for related children aged 5–17 differ between the CPS and the census. The CPS estimate that 18 percent of all related children aged 5–17 were poor in 1989 (based on a 3-year average of data from the March 1989, 1990, and 1991 CPS) exceeds the census estimate by 1 percentage point (5.9% of the census estimate), a difference that is statistically significant at the 10 percent significance level; see Table B-5. (The CPS estimate of the number of poor school-age children also exceeds the census estimate by a statistically significant amount; see Table 2-4.)
A question is whether CPS and census estimates of poverty rates and numbers of poor related children aged 5–17 differ in terms of geographic distribution, perhaps because of differences in completeness of income reporting that reflect the income mix in different areas or other reasons. In preliminary research conducted by the panel, no statistically significant differences were found in the ratios of CPS to census estimates of the number of poor school-age children among various geographic groupings of counties and states. For example, when counties were grouped by size, there were no significant differences among the groups in their ratios of CPS to census estimates of the number of poor school-age children.
In summary, there are many factors reflecting differences between the census and CPS concepts and procedures that may account for variations in their estimates of poverty levels, rates, and distributions. It is important to keep these factors in mind, particularly when attempting to measure changes in poverty at specific levels of geography since 1990.
TABLE B-3 Median Household Income in 1989 by State: 1990 Census and March 1990 CPS
|
Income (in $) |
|
|
|
|
|
1990 Census |
|
March 1990 CPS |
Percentage Difference: Census-CPS as Percent of CPS |
|
State and National |
Median |
Standard Error |
Median |
Standard Error |
|
U.S. Total |
30,056 |
7 |
28,906 |
159 |
4.0* |
Alabama |
23,597 |
58 |
21,284 |
1,070 |
10.9* |
Alaska |
41,408 |
165 |
36,006 |
1,378 |
15.0* |
Arizona |
27,540 |
72 |
28,552 |
1,210 |
-3.5 |
Arkansas |
21,147 |
50 |
21,433 |
915 |
-1.3 |
California |
35,798 |
27 |
33,009 |
618 |
8.4* |
Colorado |
30,140 |
55 |
26,806 |
1,398 |
12.4* |
Connecticut |
41,721 |
77 |
42,321 |
1,592 |
-1.4 |
Delaware |
34,875 |
151 |
32,068 |
1,133 |
8.8* |
D.C. |
30,727 |
156 |
26,752 |
1,015 |
14.9* |
Florida |
27,483 |
35 |
26,085 |
475 |
5.4* |
Georgia |
29,021 |
53 |
27,542 |
1,021 |
5.4 |
Hawaii |
38,829 |
173 |
35,035 |
1,328 |
10.8* |
Idaho |
25,257 |
79 |
24,654 |
953 |
2.4 |
Illinois |
32,252 |
32 |
31,300 |
623 |
3.0 |
Indiana |
28,797 |
50 |
25,898 |
1,022 |
11.2* |
Iowa |
26,229 |
46 |
26,265 |
792 |
-0.1 |
Kansas |
27,291 |
56 |
26,862 |
908 |
1.6 |
Kentucky |
22,534 |
55 |
23,283 |
1,206 |
-3.2 |
Louisiana |
21,949 |
49 |
22,861 |
1,857 |
-4.0 |
Maine |
27,854 |
82 |
28,221 |
1,389 |
-1.3 |
Maryland |
39,386 |
73 |
36,016 |
1,187 |
9.4* |
Massachusetts |
36,952 |
58 |
36,086 |
704 |
2.4 |
Michigan |
31,020 |
32 |
30,775 |
790 |
0.8 |
Minnesota |
30,909 |
37 |
30,185 |
1.278 |
2.4 |
Mississippi |
20,136 |
59 |
19,917 |
947 |
1.1 |
Missouri |
26,362 |
42 |
26,497 |
746 |
-0.5 |
Montana |
22,988 |
98 |
23,692 |
1,311 |
-3.0 |
Nebraska |
26,016 |
60 |
26,319 |
1,521 |
-1.2 |
Nevada |
31,011 |
92 |
29,340 |
8,455 |
5.7* |
New Hampshire |
36,329 |
97 |
37,532 |
1,371 |
-3.2 |
New Jersey |
40,927 |
51 |
39,120 |
948 |
4.6* |
New Mexico |
24,087 |
89 |
22,602 |
1,028 |
6.6 |
New York |
32,965 |
41 |
31,496 |
453 |
4.7* |
North Carolina |
26,647 |
38 |
26,406 |
517 |
0.9 |
North Dakota |
23,213 |
94 |
25,229 |
903 |
-8.0* |
Ohio |
28,706 |
37 |
29,021 |
655 |
-1.1 |
Oklahoma |
23,577 |
57 |
23,667 |
1,236 |
-0.4 |
|
Income (in $) |
|
|
|
|
|
1990 Census |
|
March 1990 CPS |
Percentage Difference: Census-CPS as Percent of CPS |
|
State and National |
Median |
Standard Error |
Median |
Standard Error |
|
Oregon |
27,250 |
55 |
28,529 |
1,435 |
-4.5 |
Pennsylvania |
29,069 |
32 |
28,690 |
684 |
1.3 |
Rhode Island |
32,181 |
120 |
30,124 |
1,354 |
6.8 |
South Carolina |
26,256 |
56 |
23,798 |
1,059 |
10.3* |
South Dakota |
22,503 |
89 |
24,108 |
999 |
-6.7* |
Tennessee |
24,807 |
50 |
22.611 |
1,305 |
9.7 |
Texas |
27,016 |
27 |
25,886 |
559 |
4.4* |
Utah |
29,470 |
86 |
30.717 |
1,014 |
-4.1 |
Vermont |
29,792 |
110 |
31,295 |
1,136 |
-4.8 |
Virginia |
33,328 |
60 |
34,118 |
1,205 |
-2.3 |
Washington |
31.183 |
44 |
31,961 |
1,472 |
-2.4 |
West Virginia |
20,795 |
61 |
21,677 |
843 |
-4.1 |
Wisconsin |
29,422 |
53 |
29,123 |
1,240 |
1.0 |
Wyoming |
27,096 |
133 |
29,521 |
1,289 |
-8.2* |
* Statistically significant difference from 0 at the 10 percent significance level. SOURCE: Data from Bureau of the Census. |
TABLE B-4 Poverty Rates for All Persons by State, 1989:1990 Census and March 1990 CPS
|
1990 Census |
|
|
March 1990 CPS |
|
|
|
State and National |
Population (000s) |
Total Poverty Rate |
Standard Errora |
Population (000s) |
Total Poverty Rate |
Standard Error |
Difference Between Rates |
U.S. Total |
241,978 |
13.12 |
0.012 |
245,992 |
12.82 |
0.2 |
0.30 |
Alabama |
3,946 |
18.34 |
0.096 |
4,074 |
18.90 |
2.0 |
-0.56 |
Alaska |
532 |
9.00 |
0.140 |
488 |
10.45 |
1.5 |
-1.45 |
Arizona |
3,584 |
15.74 |
0.086 |
3,556 |
14.12 |
1.9 |
1.63 |
Arkansas |
2,292 |
19.07 |
0.116 |
2,419 |
18.27 |
2.0 |
0.80 |
California |
29,003 |
12.51 |
0.027 |
29,346 |
12.85 |
0.7 |
-0.35 |
Colorado |
3,213 |
11.68 |
0.080 |
3,258 |
12.06 |
1.8 |
-0.38 |
Connecticut |
3,188 |
6.82 |
0.057 |
3,136 |
2.87 |
1.0 |
3.95* |
Delaware |
645 |
6.71 |
0.086 |
676 |
10.06 |
1.7 |
-1.35 |
D.C. |
571 |
16.87 |
0.233 |
569 |
17.93 |
2.4 |
-1.06 |
Florida |
12,641 |
12.69 |
0.042 |
12,762 |
12.46 |
0.9 |
0.23 |
Georgia |
6,300 |
14.65 |
0.069 |
6,197 |
14.99 |
1.9 |
-0.34 |
Hawaii |
1,071 |
8.25 |
0.119 |
1,036 |
11.23 |
1.8 |
-2.98* |
Idaho |
966 |
13.25 |
0.153 |
1,014 |
12.43 |
1.6 |
0.82 |
Illinois |
11,144 |
11.91 |
0.033 |
11,559 |
12.72 |
0.9 |
-0.81 |
Indiana |
5,372 |
10.68 |
0.051 |
5,453 |
13.74 |
1.9 |
-3.06 |
Iowa |
2,677 |
11.48 |
0.074 |
2,835 |
10.26 |
1.5 |
1.22 |
Kansas |
2,392 |
11.48 |
0.074 |
2,835 |
10.26 |
1.5 |
1.22 |
|
1990 Census |
|
|
March 1990 CPS |
|||
State and National |
Population (000s) |
Total Poverty Rate |
Standard Errora |
Population (000s) |
Total Poverty Rate |
Standard Error |
Difference Between Rates |
Kentucky |
3,582 |
19.03 |
0.093 |
3,576 |
16.11 |
2.0 |
2.92 |
Louisiana |
4,101 |
23.58 |
0.108 |
4,080 |
23.28 |
2.3 |
0.29 |
Maine |
1,190 |
10.80 |
0.108 |
1,233 |
10.38 |
1.6 |
0.72 |
Maryland |
4,661 |
8.27 |
0.054 |
4,567 |
9.00 |
1.6 |
-0.73 |
Massachusetts |
5,812 |
8.93 |
0.050 |
5,831 |
8.80 |
0.8 |
0.14 |
Michigan |
9,077 |
13.12 |
0.040 |
9,297 |
13.21 |
0.9 |
-0.09 |
Minnesota |
4,259 |
10.22 |
0.053 |
4,268 |
11.20 |
1.7 |
-0.98 |
Mississippi |
2,503 |
25.21 |
0.147 |
2,574 |
21.99 |
2.0 |
3.22 |
Missouri |
4,971 |
13.34 |
0.055 |
5,193 |
12.59 |
1.8 |
0.75 |
Montana |
777 |
16.07 |
0.186 |
816 |
15.56 |
1.8 |
0.51 |
Nebraska |
1,531 |
11.14 |
0.097 |
1,602 |
12.73 |
1.6 |
-1.59 |
Nevada |
1,178 |
10.15 |
0.118 |
1,127 |
10.74 |
1.7 |
-0.58 |
New Hampshire |
1,076 |
6.42 |
0.095 |
1,100 |
7.64 |
1.6 |
-1.21 |
New Jersey |
7,563 |
7.58 |
0.030 |
7,623 |
8.16 |
0.8 |
-0.58 |
New Mexico |
1,584 |
20.61 |
0.141 |
1,519 |
19.55 |
2.0 |
1.05 |
New York |
17,482 |
13.03 |
0.040 |
17,938 |
12.57 |
0.7 |
0.46 |
North Carolina |
6,397 |
12.97 |
0.059 |
6,301 |
12.22 |
0.9 |
0.75 |
North Dakota |
614 |
14.38 |
0.210 |
642 |
12.31 |
1.6 |
2.07 |
Ohio |
10,574 |
12.54 |
0.036 |
10,754 |
10.63 |
0.8 |
1.91* |
Oklahoma |
3,052 |
16.71 |
0.091 |
3,126 |
14.72 |
1.8 |
1.99 |
Oregon |
2,776 |
12.42 |
0.080 |
2,915 |
11.22 |
1.8 |
1.21 |
Pennsylvania |
11,536 |
11.13 |
0.033 |
12,141 |
10.39 |
0.8 |
0.84 |
Rhode Island |
964 |
9.61 |
0.121 |
963 |
6.65 |
1.5 |
2.96* |
South Carolina |
3,368 |
15.37 |
0.101 |
3,446 |
16.98 |
1.8 |
-1.60 |
South Dakota |
670 |
15.86 |
0.160 |
696 |
13.22 |
1.6 |
2.64 |
Tennessee |
4,744 |
15.70 |
0.045 |
4,833 |
16.44 |
1.9 |
-2.73 |
Texas |
16,580 |
18.10 |
0.044 |
16,886 |
17.06 |
1.0 |
1.04 |
Utah |
1,694 |
11.36 |
0.104 |
1,683 |
8.20 |
1.4 |
3.16* |
Vermont |
541 |
9.86 |
0.145 |
556 |
7.91 |
1.6 |
1.94 |
Virginia |
5,969 |
10.25 |
0.055 |
6.159 |
10.89 |
1.4 |
-0.65 |
Washington |
4,741 |
10.92 |
0.058 |
4,729 |
9.62 |
1.6 |
1.30 |
West Virginia |
1,755 |
19.66 |
0.134 |
1,799 |
15.73 |
1.9 |
3.93* |
Wisconsin |
4,754 |
10.70 |
0.054 |
4,694 |
8.37 |
1.4 |
2.32* |
Wyoming |
442 |
11.86 |
0.196 |
462 |
10.82 |
1.9 |
1.04 |
* Statistically significant difference from 0 at the 10 percent significance level. a Standard errors of estimates from the 1990 census are calculated using 1980 census design factors. SOURCE: Data from Bureau of the Census. |
TABLE B-5 Poverty Rates for Related Children Aged 5–17 in 1989, by Selected Categories of Counties: 1990 Census and March CPS
|
|
Percent Poor Related Children Aged 5–17 |
|
|
County Category |
1990 Census |
March CPSa |
Difference Between Rates |
|
U.S. Total |
17.0 |
18.0 |
1.0* |
|
Metropolitan |
|
|
|
|
|
Central |
16.4 |
17.9 |
1.5* |
|
Other |
11.4 |
12.5 |
1.1 |
Nonmetropolitan |
20.4 |
19.9 |
-0.5 |
|
Regionb |
|
|
|
|
|
Northeast |
14.3 |
15.5 |
1.2* |
|
Northcentral |
14.9 |
15.8 |
0.9* |
|
South |
20.5 |
21.3 |
0.8 |
|
West |
16.2 |
17.3 |
1.1* |
Population Size |
|
|
|
|
|
Under 2,500 |
22.9 |
22.1 |
-0.8 |
|
2,500–4,999 |
22.2 |
14.6 |
-7.6 |
|
5,000–9,999 |
23.1 |
24.7 |
1.6 |
|
10,000–49,999 |
20.6 |
20.9 |
0.3 |
|
50,000–99,999 |
16.6 |
15.7 |
-0.9 |
|
100,000–499,999 |
14.7 |
15.7 |
1.0* |
|
500,000–999,999 |
14.6 |
15.6 |
1.0 |
|
1,000,000 and over |
19.1 |
21.5 |
2.4* |
* Statistically significant difference from 0 at the 10 percent significance level. a The CPS estimates are 3-year centered averages of data from the 1989, 1990, and 1991 March CPS (reported income in 1988, 1989, and 1990, with population controls derived from the 1980 census). b The Census Bureau's regions are defined in Note c of Table 2-3. SOURCE: Data from Bureau of the Census. |