American Community Survey (ACS)

Skip Navigation Menu for Past Notices on ACS

Notice About Using the 1999 American Community Survey Estimates

Special Limitations of the 1999 American Community Survey Estimates

Significant Changes from 1998 to 1999

Special Limitations of the 1999 American Community Survey Estimates

The 1999 American Community Survey estimates have some limitations, which they will not have in future years.

1. First, these estimates do not reflect certain kinds of change in the racial and Hispanic origin distributions of the population in selected counties.
2. Second, potential problems are caused by differences between the race and Hispanic origin questions on the American Community Survey questionnaire and the questions on the 1990 census.

Insensitivity to Certain Changes in Race and Hispanic Origin

As with most household surveys, the American Community Survey data are controlled so that the numbers of people in categories defined by age, sex, race, and Hispanic origin agree with the Census Bureau's official intercensal population estimates. The American Community Survey measures the characteristics of the population, but the official count of the population comes from the previous census, updated by the official intercensal estimates program. As for all surveys conducted by the Census Bureau, the American Community Survey data were controlled to agree with the official population estimates by age, sex, race, and Hispanic origin. In the case of the American Community Survey, controls are at the county level. Such adjustments are important to correct the survey data for potential under-representation of hard-to-enumerate demographic groups compared with the level of coverage achieved in the preceding census.

The methods used to produce the official population estimates, at the county level, are not yet able to reflect changes in the race or Hispanic origin characteristics of people who migrate into and out of the county, if the changes do not follow the pattern for the state as a whole. Consequently, after being controlled to the official population estimates, the American Community Survey data will also not reflect changes that differ from the average for the state.

For example, before the population controls were applied to the American Community Survey data for Rockland County, NY, they showed that a higher percentage of the population identified as Black compared with the official population estimates. This higher percentage corresponds to patterns of immigration known to local experts. With further research and evaluation, after Census 2000 and after the American Community Survey is implemented in all counties, the methods used for official population estimates will be able to incorporate information like this from the American Community Survey, to improve population controls.

A similar situation appears to exist in some of the other sites added in 1999. Specifically, the initial American Community Survey results suggest that there may be a higher percentage of Blacks in Jefferson County, AR; Tulare County, CA; and Broward County, FL; as well as in Rockland County, NY. A higher percentage of Whites may reside in Calvert County, MD; Madison County, MS; and San Francisco County, CA. A higher percentage of Asians may reside in Broward County, FL and Franklin County, OH. However, these differences appear to be considerably smaller than the ones for Bronx County, NY.

The counties in the sites added in 1999 might be expected to have an unusual number of such patterns of change, since many high-growth and especially diverse counties were deliberately selected precisely so that potential problems, such as for survey weighting methods, could be identified and resolved for the future. Some of the higher reporting of Hispanic origin for the American Community Survey may, however, reflect differences in the wording or layout of the questionnaire, or the reporting patterns of respondents, compared with the 1990 census on which the population controls were based. These differences are described in the next section.

Effects of differences between the American Community Survey and the 1990 census

The American Community Survey differs from the 1990 census, both in the questions and in the procedures used to fill in missing information when respondents do not answer a question. Ordinarily, surveys use questions and missing data procedures comparable to the previous census-that is, the 1990 census. However, because the American Community Survey is to be compared with Census 2000, the 1999 American Community Survey uses the Census 2000 questions and procedures. After Census 2000, special intercensal population estimates for 1999 will be re-computed, to be consistent with the questions and procedures of Census 2000, permitting an exact comparison between the American Community Survey data using the two sets of population estimates as weighting controls-the "1990-based" and "2000-based" ones.

In the meantime, the "2000-based" American Community Survey must use the "1990-based" intercensal estimates as population controls for the 1999 data. Our approach has been to force agreement with the intercensal estimates for the total population, and to apply weighting factors to achieve approximate agreement for the numbers of people in the major race groups and for the number who identify as Hispanic. As stated above, this adjustment, which also takes into account age and sex, is necessary to correct for differences in coverage. While the use of weights does assure that survey totals match the official population estimates of broad groups, the number of people with various combinations of race and Hispanic origin is still based on tabulating the actual survey responses.

In all counties but one, the results of this method seem reasonable, with weighting factors falling generally in the ranges of values expected with such weighting methods. Consequently, the resulting "experimental" estimates are likely to be representative of the characteristics of these counties. We will only be able to confirm this conclusion, however, when Census 2000 information becomes available in 2001.

At present, the differences between the 1990 and 2000 questions and procedures are having an effect on the data for Bronx County, NY such that the normal statistical methods are unable to adequately compensate for them. The population controls give considerably too much weight to people who are identified as Black Hispanics, too little weight to people who are identified as Black non-Hispanics or White Hispanics, and much too little weight to people identified as White non-Hispanics. Because the characteristics of these groups differ, inaccuracies in the weights mean that the characteristics of combinations of these groups are also inaccurate.

The experimental estimates for Bronx County, NY remain on the American Community Survey website and the American Fact Finder so that data users can study them, but they should be used with great caution, since they do not necessarily provide an accurate profile of Bronx County. The estimates will be recalculated after Census 2000 data are available and revised weights for 1999 can be produced.

[PDF] or denotes a file in Adobe’s Portable Document Format. To view the file, you will need the Adobe® Acrobat® Reader available free from Adobe. This symbol indicates a link to a non-government web site. Our linking to these sites does not constitute an endorsement of any products, services or the information found on them. Once you link to another site you are subject to the policies of the new site.