American Community Survey (ACS)

The American Community Survey: The Census Bureau's Plan to Provide Timely 21st Century Data

Mary Ellen Davis and Charles H. Alexander, Jr.

U.S. Census Bureau
Washington, DC 20233

The paper was printed in the "Delaware Dataline", Summer 1997.

In a rapidly changing society, many see the decennial census as too slow and antiquated. Data from a census taken once every ten years becomes stale. To counter this, the US Bureau of the Census is implementing a process of continuous measurement called the American Community Survey (ACS).

The American Community Survey will not replace the decennial census; information from what is commonly called the "short form" will still be collected every ten years. The ACS will replace the "long form" census questionnaire starting in 2010.

The Census Bureau's method for collecting detailed socioeconomic data (traditionally decennial census) will be re-engineered. Data will be provided throughout each decade. The goal of continuous measurement is to blend the strength of census-based small-area estimation with the quality and timeliness of continuing surveys.

Current plans call for a continuous monthly survey covering most of the US beginning in 2000 and extended to every county beginning in 2003. The ACS will cover topics determined by the decennial census content process. Over the course of each year starting in 2003, three million different addresses will be selected for interview.

At its full sample size, the ACS will provide descriptive profiles annually for all states, and for all communities of 65,000 persons or more, based on the interviews from the previous year. For less populous communities, such as rural areas, city neighborhoods, or very small population groups, the sample size will be too small to make good estimates from just one year's sample.

At three million households per year, it will take about five years to sample the same number of households as sampled in the 1990 decennial census. Therefore the full five-year accumulation of data will be needed to provide accurate information for small geographic areas such as census tracts.

Once the ACS is in full operation, estimates will be provided each year for every governmental unit, for every component of the population, and for census tracts and block groups based on refreshed multiyear accumulations.

The goals of the American Community Survey are:

• To provide users with geographically detailed socioeconomic data throughout each decade
• To aid state and local officials in meeting new responsibilities such as welfare reform
• To replace the 2010 census long form
• To improve the infrastructure for the federal statistical systems

Plans for the ACS include releasing an annual microdata file, patterned after the five percent Public Use Microdata Sample, (PUMS) file of the 1990 decennial census records. In addition, the ACS will provide summarized data similar to the Summary Tape Files (STF) of the 1990 decennial census. The microdata files, tabulated files, and associated documentation will be available through the Internet and on CD-ROMs.

Users will be able to obtain annual summaries down to the block-group level, with the flexibility to tabulate and analyze data in various ways. However, the recommended profiles to describe small communities would be based on multiple years of data, the number of years depending on population.

As originally planned, the ACS was to have an extra-large sample during the years 1999-2001 to allow direct comparison of three-year accumulations of ACS data with the 2000 long form estimate for every census tract. The Census Bureau believes that comparison of the ACS with the long form is essential if the ACS is to replace the long form as the basic profile of the nation's communities.

Given tight budgets, however, it was not realistic to conduct an expanded ACS and a census long form at the same time. The Bureau has developed an alternative that reduces and redesigns the sample while providing a good opportunity for comparison.

The new plan reduces the ACS comparison sample from five million addresses per year for 1999-2001 to 940,000 addresses per year in 2000-2002. Instead of spreading the entire sample evenly across the nation, the 940,000 sample will be divided into two parts. The first will be a nationwide sample of 700,000 addresses per year covering most metropolitan areas and a sample of rural counties.

The second will consist of about 40 selected "comparison counties" from across the nation, where an annual sample totaling 240,000 addresses will be concentrated to allow detailed scrutiny of the ACS estimates to understand where and why they differ from the 2000 long form estimates. These comparison counties will be in the sample starting in 1999 and will be sampled at the same annual rate (roughly 5%) that was used nationally in the original proposal, so direct comparisons for every census tract can be made in these counties.

From the national sample of 700,000 addresses, the Census Bureau plans to make a direct comparison of the ACS and the 2000 long form estimates for all states, large cities, and large sub-state areas. For smaller areas, such as small counties, small towns, or census tracts, the ACS sample will be too small to produce a 2000 estimate. Instead, statistical modeling will be used to produce a modeled, "ACS-like" estimate. Some differences are expected due to collecting data throughout the year rather than just in April as well as to differences in population coverage and non-response rates. Mid-decade updates from the full ACS will be compared to the modeled, "ACS-like" estimates which have been adjusted for differences in collection procedures.

During 1998, the Bureau will continue to collect data for the eight areas (in Oregon, New York, Florida, Pennsylvania, Nebraska, Texas, Ohio and New Mexico) currently used as evaluation sites. They will produce 1997 estimates for these counties and estimate changes between 1996 and 1997 for the four counties which were first set up to evaluate the ACS in 1996. They will also collect data for two counties in South Carolina that overlap with counties in the 1998 Decennial Census Dress Rehearsal to investigate the effects of conducting both the ACS and the census simultaneously.

In 1999, they plan to increase the number of county/city sites in the sample to approximately forty and begin expanding the national sample. Between 2000 and 2002, 700,000 addresses will be included annually in the national sample in addition to the 240,000 addresses in the forty bench-mark sites. The size of the 2000 sample will allow the Census Bureau to provide descriptive profiles for all states, all congressional districts, and cities or counties of 250,000 persons or more.

Because it will give current estimates for small geographic areas, the Census Bureau believes that the ACS can be used to improve statistical estimates for states and counties from other federal surveys such as the Current Population Survey, the Survey of Income and Program Participation, the Local Area Unemployment Statistics program, and the National Crime Victimization Survey. As a result, these surveys will be able to provide information with much greater geographic detail than in the past.

Making this small-area data available continuously should provide an essential and consistent base for evaluating the effects of welfare reform, changes in educational programs, enterprise or empowerment zones, and other programs that affect local communities.

For additional information about the American Community Survey, contact the Demographic Statistical Methods Division of the Continuous Measurement Office, US Bureau of the Census at 888-456-7215 or cmo.acs@census.gov.

An American Community Survey Timeline

The proposed general timing for the American Community Survey between 1997 and 2010 is designed to lead to the eventual replacing of the census long form by the ACS for the 2010 census: