- Estimates for 2005
- Updated: April 2009
We estimated indicators relevant to early childhood education in Illinois counties, townships, municipalities, and state legislative districts in 2005: the total population; the population of each age cohort from 0 to 5 (including a summation of all six cohorts); the number of linguistically isolated households speaking Spanish and other non-English languages; and the percent of children age 5 and under who are living in households that earn income below the poverty line and below 185 percent of the poverty line, and, for counties and townships, the number of children living in households earning below 400 percent of the poverty line. To do this, we employed data from three sources:
- the 2005 American Community Survey,
- the 2000 decennial census, and
- the US Census Bureau’s Population Estimates Program (in cooperation with the National Center for Health Statistics)
Each source has some drawbacks, and each method we used to estimate the indicators relies on assumptions; however, we have calculated what we feel are the most accurate estimates given the available data. This text will first summarize the five sources of data, their strengths, and their drawbacks. It will then detail the methods used to estimate each of the variables. The appendix contains more detail of the methodology and a table of the counts for each of the variables for each of the counties.
We also projected the number of children age 5 and under in counties in 2011. We derived the number of children below 100, 185, and 400 percent of the poverty line from those projections based on the ratios of children living below those levels of poverty in 2005 in counties. These projections employ data from:
- the Regional Economic Information Service,
- the 2000 decennial census, and
- the Center for Disease Control’s Wonder website.
The pros and cons of these sources and an assessment of the accuracy of these projections can be found in “The right people, the right rates” by Andrew M. Isserman, Journal of the American Planning Association, Winter 93, Vol. 59, Issue 1 (page 45-64). A discussion of the data sources and methods for estimating the 2005 population follows.
We initially tried to find as many data sources covering the indicators as possible. We then narrowed the scope to these four based on their geographic and informational coverage and reliability. All four are based on surveys and censuses conducted by the US Census Bureau, although some also take into account other data and surveys.
2005 American Community Survey
The American Community Survey (ACS) is an annual survey designed to replace the decennial survey (known as the census long form) by 2010. The ACS surveys about 3 million addresses (about 2.5 percent of households) annually based on a Master Address File listing all residential and commercial addresses (US Census Bureau, 2005a; Mather et al.. 2005). The 2005 survey did not survey people in group quarters,1 but they will be included in the 2006 survey (US Census Bureau, 2006a). The first nationwide survey (2003) reported results only for areas of 250,000 people or more. In 2006, the survey reported results for areas of 65,000 people or more (for the 2005 survey). In 2008, the survey is scheduled to report results for areas of 20,000 people or more (for the 2007 survey), and in 2010, it is scheduled to begin reporting results for all areas (Mather et al., 2005). The survey is conducted every month throughout the year, and the results are a compilation of the monthly results over 12 months for locations with more than 65,000 people, over 36 months for locations of between 20,000 and 65,000 people, and over 60 months for locations of fewer than 20,000 people (US Census Bureau, 2006a). Population estimates for 2005 were released in 2006 for 22 counties and 43 townships in Illinois.2 For more specific information, such as language spoken at home, fewer counties and townships were included since the sample may not have been large enough to yield reliable numbers.
The content of the ACS is similar to the decennial long form and includes information on demographic, social, housing, and economic characteristics. The data are released through the US Census Bureau’s Web site,3 similar to the decennial census, in a series of tables. The tables can be accessed by specifying the geographic area the user is interested in and downloaded in a number of spreadsheet formats, including comma-delimited and Excel. Because the tables are pre-defined, only certain ways of cutting the data are available. For example, it is possible to discover how many people live in households with incomes within a certain interval of the federal poverty line (table B17002), and it is possible to discover how many children age 5 and under are under the poverty line (table B17001), but it is not possible to find how many children 5 and under are within a certain interval of the poverty line. Tables of specific information are often not available because the sample may not be large enough to yield reliable results for a phenomenon that affects few people.
2000 US Census
The decennial census is made up of two parts: the census, a 100-percent count of the population in the US; and the long-form survey, a survey of about 19 million addresses (about 15.8 percent of households) (US Census Bureau, 2002; Mather et al., 2005). The two parts ask questions on demographic, social, housing, and economic characteristics similar to the ACS survey and release data for areas as small as block-groups (sub-tract divisions). Like the ACS, the data are released through the US Census Bureau’s Web site4 in a series of tables that can be accessed and downloaded in the same way as the ACS data. The tables, however, sometimes differ from the ACS tables. Usually the decennial census tables allow the user to get more specific tables, but these are still limited.
US Census Bureau’s Population Estimates Program
The Population Estimates Program is a division of the US Census Bureau focused on creating annual national-, state-, county-, municipality-, and township-level estimates of population (though these estimates are not made for precincts, 273 township-like areas that do not have a government associated with them). The same county-level population estimates are also released through the National Center for Health Statistics (Ingram et al., 2003). The data are broken down differently depending on whether they are accessed through the PEP Web site or the NCHS Web site. The data on the NCHS Web site (Bridged-race Vintage 2005 postcensal population estimates for July 1, 2000 – July 1, 2005, by year, county, single-year of age, bridged-race, Hispanic origin, and sex) break down the population at the county level into sex, age, and race cohorts. These estimates are based on revised estimates of population from the 2000 Census and are projected forward using birth, death, and migration information.5 The birth and death data are calculated with data from the Federal State Cooperative Program for Population Estimates and the National Center for Health Statistics. The net domestic migration is calculated with data from the Internal Revenue Service and the Social Security Administration. The net international migration is calculated with data from the American Community Survey and the 2000 decennial census. The net military migration was calculated with data from the Department of Defense. The Population Estimates Program includes people living in group quarters and estimates these numbers separately from the household population with help from the Federal State Cooperative Program for Population Estimates (US Census Bureau, 2005b). The data on the NCHS Web site are for national, state, and county levels only. They do not include estimates for municipalities or townships. The Population Estimates Program releases total population estimates only for these two jurisdictional types through the US Census website.
The methods for estimation vary with the indicator. In some cases our sources provided reliable estimates. In other cases, our sources did not provide estimates at the level we were interested in, or of the exact parameter we were interested in, and so we had to make some assumptions to create the estimate. The following paragraphs discuss the sources, our general methods, and the reliability of the estimates. For more detailed information on our methodology, see appendix A.
For both the total population and the population of people in the age cohorts 0, 1, 2, 3, 4, and 5, we used the estimates created by the Population Estimates Program in cooperation with the National Center for Health Statistics (PEP/NCHS) at the county level. The methodology used to create the estimates is sound, and they appear to be fairly accurate. The ACS covers only 20 counties in Illinois and does not currently release data on single-year age cohorts; however, it does release the number of children under 5 at the county level. Comparing the ACS estimate and the PEP/NCHS estimates for the 20 counties covered by the ACS, we found little variation. The variation seemed to be related to the size of the population under 5 (see chart 1 at the end of this text) with more populous counties having less variation between the two estimates. The county with the most variation was Adams, with the PEP/NCHS estimate about 4.65 percent less than the ACS estimate. For the rest of the counties, the PEP/NCHS estimate was less than 3 percent above or below the ACS estimate. While the PEP/NCHS includes people living in group quarters and the ACS does not, this discrepancy should have little effect on population estimates of children under 5. When the ACS begins to release data for all places in 2010, it will be interesting to compare the two estimates again and look for biases.
PEP creates estimates only for total population in townships in Illinois and not for precincts. For the townships, we assumed that if township X in county A had 23 percent of the total county population in 2005, then it also had 23 percent of the county’s population of 3-year-olds. For the remaining 273 precincts of the 1,710 subcounty divisions in Illinois, we assumed that the county population of children age 5 and under would be distributed between the precincts in the county in the same way it was in 2000. If 21.6 percent of the total Champaign County population of this age group lived in Champaign Township in 2000, we assumed that the same percent lived there in 2005. Likewise, for state legislative districts, we assumed that the distribution of the population in the state between state legislative districts remained the same. For example, if 3.7 percent of the total state population of this age group lived in a particular legislative district in 2000, then we assumed that the same percent lived there in 2005.
For state legislative districts, we assumed that each district maintained the same share of the state’s population in each age cohort. This is similar to the assumption of the relationship between township and county population described above. In addition, since the Population Estimates Program doesn’t publish estimates of total population for these districts, we made the same assumption about total population. State legislative districts are not currently covered by the ACS and so it wasn’t possible to compare our estimates to a reliable data source.
For municipalities, PEP also only creates estimates for total population. Since municipalities are not wholly contained within a county, we assumed that within each municipality, each age cohort maintained a constant share of the population. For example, if 1 percent of municipality X’s population was 1 year old in 2000, then we assumed that 1 percent was 1 year old in 2005 as well. We felt that this would probably be a better estimate, especially for small areas, than assuming a municipality’s share of the state’s total population is the same as the municipality’s share of 1-year-olds (which would be parallel to the township-county assumption above but at a much larger scale and the state legislative district-state assumption). For the 18 municipalities covered by the ACS, the ACS estimate of the number of children age 5 and under and our estimate were fairly close. The percent difference between our estimate and the ACS estimate was between 0 and 42 percent with all but two of our estimates falling within the ACS estimate’s margin of error. The larger the population of children under 5, the closer our estimate was, in percentage terms, to the ACS estimate. It should be noted, however that these comparisons are made among the largest places. The smaller the population, the more variability there is likely to be between our estimate and reality.
Linguistically Isolated households
A linguistically isolated household is defined by the Census Bureau as a “household…in which no member 14 years old and over (1) speaks only English or (2) speaks a non-English language and speaks English very well. In other words all members 14 years old and over have at least some difficulty with English” (note on ACS table B16002). Since the definition is so specific, it can only be found through the ACS or the decennial census. The 2005 ACS covers 17 counties, 18 municipalities, and 43 townships within 11 counties6 with this variable and 96 percent of linguistically isolated households in the state. To estimate how the other 4 percent are distributed throughout Illinois, we assumed the distribution to be the same in 2005 as it was in 2000. We first subtracted the number of linguistically isolated households in the ACS-covered counties from all linguistically isolated households in the state in 2000. We found how the linguistically isolated households were distributed among the remaining counties and applied those percents to the 2005 ACS estimate of linguistically isolated households in Illinois (less the number of linguistically isolated households in the ACS-covered counties).
The townships we estimated in the same way but instead of finding how linguistically isolated households were distributed among townships within the state, we found how they were distributed between townships within a county in 2000 and applied that distribution to the 2005 county total. Where there were ACS estimates for townships, we used those and estimated the remaining townships. Because we felt that the distribution of linguistically isolated households among households within the state would vary too much (or rather small percentage variations could result in large differences in estimates) we assumed that the percent of all households that are linguistically isolated remained constant between 2000 and 2005. Since state legislative districts are not covered by the ACS, we simply found how linguistically isolated households were distributed within the state in 2000 and applied that distribution to the 2005 state total of linguistically isolated households.
To estimate the number of children age 5 and under living under the poverty line, under 185 percent of the poverty line, and under 400 percent of the poverty line, we used tables in the decennial census, ACS, and a table created especially for IECAM by the Census Bureau from the decennial census. Estimates for the number of children living below 400 percent of poverty could only be created for counties and townships due to data constraints. According to the 2005 ACS, 86 percent of children age 5 and under living below the poverty line were living in counties covered by the ACS (ACS 2005, table B17001).7 The ACS does not report age-specific information for other ratios of income to poverty so we can’t know for certain how many children living below 185 percent of poverty were living in the ACS-covered counties or townships. However, we can estimate the number of children living below 185 percent of poverty by using information from the decennial census.
We first applied the percent of all people living below 185 percent of the poverty line in 2005 to the number of children age 5 and under in 2005. When we estimated the 2000 data using this method, we found that our estimates were consistently below the actual number of children living below 185 percent of the poverty line. Therefore, we adjusted the 2005 estimates based on these differences. In the 2000 decennial census, the ratio of income to poverty is reported for different age groups. We found the ratio of the percent of all people living below 185 percent of the poverty line to the percent of children age 5 and under living below 185 percent of the poverty line for each county and township in 2000 and applied this adjustment to the 2005 figures.
To find the number of children living below 400 percent of the poverty line, we assumed that the percent of children living below 400 percent of the poverty line was constant between 2000 and 2005. We found the percent of children age 5 and under living below 400 percent of the poverty line in 2000 and applied it to the population of children age 5 and under in 2005. This information was derived from a table made especially for IECAM by the Census Bureau. They only created these estimates from counties and so we were not able to create estimates of the number of children below 400 percent of the poverty line for municipalities or state legislative districts. This will be remedied in 2010, since the ACS now includes estimates up to 500 percent of poverty and, in the 2010 survey, all areas will be included.
There is a problem in estimating poverty from the ACS using proportions generated by the decennial census. The ACS is carried out throughout the year. In the decennial census, people were asked what their income was in a given year (for example 1999). In the ACS, people were asked how much they made in the last 12 months (for example from April 2006 to March 2007). It appears that people under-reported their income in the ACS (Posey et al., 2003). This explains why the percent of people in poverty appeared to go up in almost all counties covered by the ACS in Illinois between 2000 and 2005 (see chart 2 at the end of the text). It also indicates that our estimates of the number of children in poverty are probably higher than the actual number of children in poverty.
This was particularly noticeable when we estimated the number of children living below 400 percent of the poverty line in another way. We assumed that the number of children living below 400 percent of the poverty line increased at the same rate as children living below the poverty line. The result was that our estimate of the number of children living below 400 percent of the poverty line was frequently larger than the total number of children in the county. However, given the present data availability and without more information on the relationship between income reported in the decennial census and income reported in the ACS, we feel our estimate is reasonable as long as it is understood that the estimate is probably somewhat higher than the actual number of children in poverty.
Chart 1.The relationship between the difference in the American Community Survey (ACS) and the Population Estimates Program/National Center for Health Statistics (PEP/NCHS) estimates of the size of the population under age 5. Cook County was excluded for display purposes (it has an ACS-estimated population under 5 of 399,543, and the PEP/NCHS estimate was 0.22 percent higher).
Chart 2. The percent of children age 5 and under living in poverty went up in almost every county in Illinois covered by the ACS. This increase indicates that there may be something different in the way people reported their income in the decennial census from the way they reported it in the ACS. Note that this is a change in percentage points (2005 percent minus the 2004 percent), not percent change.
Griffin , Deborah H., & Waite, Preston Jay. (2006). American Community Survey overview and the role of external evaluations. Population Research and Policy Review, 25(3), 201-223.
Ingram, Deborah D.; Parker, Jennifer D.; Schenker, Nathaniel; Weed, James A.; Hamilton, Brady; Arias, Elizabeth; & Madans, Jennifer H. (2003, September). United States Census 2000 population with bridged race categories. National Center for Health Statistics (DHHS pub. (PHS) 003-1335). Vital and Health Statistics, 2(135). Retrieved May 30, 2007, from http://www.cdc.gov/nchs/data/series/sr_02/sr02_135.pdf
Integrated Public Use Microdata Series. (2007). Descriptions of IPUMS samples. Retrieved January 15, 2007, from https://usa.ipums.org/usa/sampdesc.shtml#us2005a
Mather, Mark; Rivers, Kerri L.; & Jacobsen, Linda A. (2005). The American Community Survey. Population Bulletin, 60(3), 1-20.
Miller, Ronald E. & Blair, Peter D. (1985). Input-output analysis: Foundations and extensions. Englewood Cliffs, N.J.: Prentice-Hall, pp. 464.
Posey, Kirby G.; Welniak, Edward; & Nelson, Charles. (2003, August). Income in the American Community Survey: Comparisons to Census 2000. Washington, DC: US Census Bureau. Retrieved October 11, 2010, from http://www.census.gov/acs/www/Downloads/methodology/ASA_nelson.pdf
US Census Bureau. (2002, September). Census 2000 basics. Washington, DC: US Government Printing Office. Retrieved May 30, 2007, from http://www.census.gov/mso/www/c2000basics/
US Census Bureau. (2005a). Accuracy of the data (2005). Washington, DC: US Census Bureau, American Community Survey. Retrieved May 30, 2007, from http://www.census.gov/acs/www/Downloads/ACS/accuracy2005.pdf
US Census Bureau. (2005b). Estimates and projections area methodology: County population estimates by age, sex, race, and hispanic origin for July 1, 2005. Washington, DC: US Census Bureau, Population Estimates Program. Retrieved May 30, 2007, from http://www.census.gov/popest/topics/methodology/2005_co_char_meth.pdf
US Census Bureau. (2006a, May). Design and methodology: American Community Survey (Technical paper 67). Washington, DC: US Government Printing Office. Retrieved May 30, 2007, from http://www.census.gov/acs/www/Downloads/tp67.pdf
US Census Bureau. (2006b, August 24). Using data from the 2005 American Community Survey. Washington, DC: US Census Bureau, American Community Survey. Retrieved May 30, 2007, from https://www.census.gov/programs-surveys/acs/
US Census Bureau. (2007). Small area income and poverty estimates: 2001-2004 county-level estimation details. Retrieved May 31, 2007, from http://www.census.gov/hhes/www/saipe/techdoc/methods/cnty/04cntymod.html
Appendix A Method Specifics2005release-appendix-a-1
Footnotes The most common types of group quarters are college residence halls, residential treatment centers, nursing homes, group homes, military barracks, correctional facilities, workers’ dormitories, and homeless shelters.  The counties covered by the 2005 ACS are Adams, Champaign, Cook, DeKalb, DuPage, Kane, Kankakee, Kendall, Lake, LaSalle, McHenry, McLean, Macon, Madison, Peoria, Rock Island, St. Clair, Sangamon, Tazewell, Vermillion, Will, and Winnebago.  Results of the 2005 ACS survey are available through the Census Web site, American Factfinder, at http://factfinder.census.gov. Information from the survey is also available in other formats, such as on CD, but the data used here come from tables created in American Factfinder.  Results of the 2000 Census are available through the Census Web site, American Factfinder, at http://factfinder.census.gov. Return to text.
 Pop05 = Pop04 + births – deaths + net domestic migration + net international migration + net military migration (US Census Bureau 2005b). Return to text.
 The covered counties are Champaign, Cook, DeKalb, DuPage, Kane, Kendall, Lake, LaSalle, McHenry, McLean, Madison, Peoria, Rock Island, St. Clair, Sangamon, Will, and Winnebago.  The counties covered by the ACS for this variable are Adams, Champaign, Cook, DeKalb, DuPage, Kane, Kankakee, Kendall, Lake, LaSalle, McHenry, McLean, Macon, Madison, Peoria, Rock Island, St. Clair, Sangamon, Tazewell, Vermilion, Will, and Winnebago.