Connecting information professionals across upstate New York.

Categorized | Members

Introducing the American Community Survey

by Pam Baxter & Warren Brown

For those who follow activities in the federal statistical agencies (and who doesn’t!), you may have noticed some prominent headlines about the upcoming decennial Census in recent months:

“Census Switch Worries Some,” (Washington Post, July 8, 2008) regarding the recent decision by the Census Bureau to replace use of handheld computers with pencil and paper collection instruments to follow up with non-respondent households.
“Census Damage Control,” (New York Times editorial June 23, 2008) on agency preparation and funding shortfalls provided by Congress.
“Out for the Count,” (Boston Globe, July 27, 2008) highlighted a Census Bureau decision to count legally married same-sex couples in Massachusetts as “unwed partners.”

So this seems an appropriate time to highlight a statistical product from the Census Bureau that will come into its own in 2010—The American Community Survey.

But what’s wrong with the Census we’ve all grown to love?

Directed by the US Constitution (Article 1, section 2), the Census conducts a count of residents of the United States every ten years. Of course, there are inherent shortcomings in a survey conducted on a ten-year cycle. The information on employment, housing values and rent, and income becomes dated and cannot accommodate emerging information needs. Furthermore, concerns about accuracy of the survey are always important to political interests as the decennial Census data determines both Congressional reapportionment and legislative district boundaries, as well as influence dispersal of funds from a host of federal programs.

The 1990 and 2000 surveys were littered with bureaucratic and political fallout. As always, cost was a primary concern. Sending human enumerators to follow up with those households who did not respond to the mailed Census form (on average, 30% of households in 2000) was a large part of that cost.  The issue of the “undercount” or those traditionally unrepresented in the Census (urban dwellers, the poor, isolated rural residents, and the homeless) also loomed large. A 1993 proposal to apply statistical estimation methods (to address both cost and the undercount problems) was decried by advocates of the “actual enumeration” called for in the Constitution. Additionally, some groups objected to the intrusiveness of questions on the “long form,” a questionnaire distributed to approximately 16% of the U.S. population that collected detailed housing and demographic information.

Still, the U.S. Census is extremely important to academic and business researchers, policy makers and advocates, as well as the demographically curious. Certainly detailed demographic and economic characteristics are collected by other federal surveys. However, researchers rely on the Census for information on small areas such as the New York towns Altamont, Corinth, and Whitney Point.

Introducing the American Community Survey

Enter the American Community Survey (ACS), a Census Bureau undertaking that has been under development for more than 10 years. Although the ACS will not entirely replace the decennial Census, it will serve as a substitute for the Census “long form.”

A primary goal of the ACS is to provide continuous measurement to address problems with a decennial data collection cycle. A strength of the Census “long form” has been its ability to provide demographic, social, and economic snapshots of small geographic areas. When fully implemented, the ACS will supply a moving picture of the American population for the same small geographic levels that the Census “long form” has done previously.

How does it work?

The ACS surveys are distributed monthly to a sample of approximately 250,000 addresses. The responses are tabulated and released on an annual basis. (Another 36,000 addresses are included for a parallel Puerto Rico Community Survey.)

Beginning in 1996, the Census conducted a series of dress rehearsals to test and refine sampling and survey techniques. The first real data collection began in 2000, with full nationwide implementation in 2005. The Bureau has produced profiles based on ACS data for the country, states, and large geographic areas during the development phase of 2000 to 2004, and since full implementation in 2005.

Small area geographies have fewer people to sample, and as such they require a longer period of time to accumulate a critical mass of survey responses for inclusion in the ACS. Therefore, reports of sub-state areas will be produced depending on the size of their populations. The more densely populated the area, the earlier the ACS data will be released.

For example, areas with populations of 65,000 or more use single-year estimates (sampling from one year of data from the ACS) whereas smaller areas like Census blocks will need 5 years of sampling data collected before being released. The 2005 survey data for the large population areas was released in 2006, and the 2006 survey data was released in 2007. Collection for the mid-sized population areas requires three years of data collection before being released. Data collection for these mid-sized population areas started in 2005 and were first be released in December 2008. Estimates for the smallest geographic areas (block groups, tracts, small towns, and rural areas) depend on five years of survey data and will be released in 2010 for the first time, and then annually thereafter.


Placed in the New York perspective, ACS profiles for areas with populations greater than 250,000, such as the Town of Brookhaven, Monroe County, and the City of Buffalo, go back to data collected each year since 2000. Tompkins County and Ithaca metropolitan area profiles were released in 2006 based on the 2005 data collection. Data on characteristics of population and housing for Lewis County, with an estimated population more than 20,000, will be based on 3-year period estimates and released in December 2008. And Hamilton County, estimated population of approximately 5,000 won’t have ACS profiles published until 2010.

Most people associate the decennial Census and ACS as a survey of households and the people who live in them. However, both surveys also cover group quarters such as college dormitories, nursing homes, correctional facilities, and military barracks. Additional funding enabled the Census Bureau to incorporate group quarters in the ACS survey cycle beginning in 2006.

How is this information different than the data from the long form?

It’s important to keep in mind at least two important factors regarding ACS data products. One is sampling error and the other is the practical efficacy of continuous measurement.

Any results based on surveys distributed to a sample of a population (as opposed to a survey in which 100% of all respondents are surveyed) are subject to sampling error. As the size of the sample increases relative to the entire population, the sampling error decreases. Tabulations based on the decennial “long form” surveys (for example, education attainment, income, value of home) were all subject to sampling error because not all persons in the U.S. received and completed the long form. ACS tabulations are subject to larger sampling errors because the survey sample is smaller.

For example, the 2007 ACS profile of the City of Albany indicates that the median household income was $37,298. The margin of error is +/- 3,070. That is, there’s a 90% probability that the 2007 median household income is between $34,228 and $40,368.

A second item to keep in mind, and one that’s related to the sampling error issue, is the role of continuous measurement. Although the continuous measurement of the ACS is an improvement over the static snapshot decennial Census, how ACS information is used has real implications for local governments and others who rely on accurate and timely profiles. For example, New York State’s Empire Zones program has used decennial Census data on poverty and unemployment for census tracts to determine areas that qualify for assistance. With the ACS, these data will be based on 5-year period estimates beginning with the release in 2010 and will be released each year thereafter. In the past, eligibility was determined once a decade, but with the ACS releasing estimates each year, frequent changes in eligibility could raise a problem.

The verdict?

Obviously, the ACS will not entirely replace the decennial Census, which is constitutionally mandated and used for benchmarking other federal statistical products. The ACS will serve as a substitute for the Census “long form,” and in doing so will be able to release information in a timely way that could not have been done with the previous method.

For more information: 
The American Community Survey: An Overview is a web-based PowerPoint program that addresses many of the most common questions.

About the Data has infomation on sampling error and continuous measurement reports.

ACS Compass Products, a new series of educational tools and handbooks.

For background on development of the ACS, see:
Alexander, C. (2000). American community survey. In M. J. Anderson (Ed.), Encyclopedia of the U.S. Census. (pp. 26-28). Washington, DC: CQ Press.

Author information:
Pam Baxter
Librarian and Data Archivist, Cornell Institute for Social and Economic Research

Warren Brown
Director of the Applied Demography Program at the Vinson Institute of Government, University of Georgia

Comments are closed.

Save the Date!

Mark your calendars for the Spring 2018 UNYSLA Conference: Lead From Where You Are


When: April 13th, 2018

Where: Rochester, NY


Check back soon for more details!

Get Involved with UNYSLA

There are several ways to get involved including conference planning, board positions, networking events, and so much more! Send any inquiries to anyone listed on our About page or fill out our Contact Us form.

Quick Links

Thank you to our sponsors!

Thanks to our Sponsor!

Thanks to our sponsor, Ebsco Health! Interested in becoming a sponsor? Find out how.

Archived Posts