GEOG 482 Project 3: Acquiring Geographic Data

Topic G: American Community Survey

Don Krysakowski

The old saying that knowledge is power meets with the notion that in Washington, DC, power is money.  The American Community Survey (ACS) will provide information to local governments, organizations, and businesses to assess local needs of such disparate items as roads, schools, and community centers. (U.S. Census Bureau, ACS Overview, 2005)  It will also be used to determine the allocation of federal resources (read: money) to the states on a much more timely basis.

1.  What is the American Community Survey?  What need(s) does it address?

The American Community Survey (ACS), as well as its companion Puerto Rico Community Survey (PRCS), is a survey program designed to capture information about the nation's changing population and meet the demands for timely and relevant data about population and housing characteristics.  The ACS is replacing the "long form" census questionnaire starting with the 2010 census, though elements of the program have already been initiated on a test basis.  Starting this year, a rolling, random sample of housing unit addresses throughout the United States and Puerto Rico began receiving the American Community Survey questionnaire in the mail.  Each year, about 1 in 40 addresses will be surveyed.

In 2006, data will be available annually for all areas of the nation with populations greater than 65,000.  For smaller areas, it will take three to five years to accumulate a large enough sample to produce annualized data.  Areas in size between 20,000 and 65,000 people will have data averaged over a three year period, and areas of less than 20,000 people will utilize data averaged over a five year period. (U.S. Census, ACS Overview, 2005)

In its American Community Survey, A Handbook for State and Local Officials, The United States Census Bureau explains that the traditional decennial census has two parts: 1) the short form, which counts the population; and 2) the long form, which obtains demographic, housing, social, and economic data.  Information from the long form is used for the administration of federal programs and forms the basis for the distribution of billions of federal dollars to states and local communities. (U.S. Census, Handbook for State and Local Officials, 2004)

Since the census is conducted only once every 10 years, long-form information quickly becomes out of date.  The American Community Survey is designed to provide the equivalent of long form data but on a regularly updated basis, i.e. every year, instead of once every ten years.  Further, the ACS also is intended to replace the long form completely beginning with the 2010 census. (U.S. Census, Handbook for State and Local Officials, 2004)

The goal of the American Community Survey is to provide an up-to-date statistical view of a community in order to assist state and federal officials distribute block grants and to track the well-being of children, families, and the elderly.  It will be used to assist with the determination of where to locate highways, schools, and hospitals, and it will be available to both public and private enterprises needing community-based statistical information. (U.S. Census, Handbook for State and Local Officials, 2004)

Instead of surveying approximately one sixth of the U.S. population every ten years with a detailed set of questions about income, housing, and household information using the "long form," the ACS will collect data on approximately 2.5% of American households (1 in 40) every year. (U.S. Census, ACS Overview, 2005)

The methodology for collecting this data is to mail out questionnaires to a targeted list of households.  This targeted list is based upon the U.S. Census Bureau's Master Address File (MAF) which was created prior to the 2000 Census from U.S. Postal Data as well as from addresses supplied by local governments and from other locally-based updates to the Census Address Program.  MAF data is directly linked to the Topographically Integrated Geographic Encoding and Referencing (TIGER) files containing census-required map features and attributes including geocoded data on states, counties, census tracts, and census blocks. (U.S. Census Bureau, American Community Survey Operations Plan, 2003, pp 9-10)

The maintenance of the MAF data, particularly for rural areas without city-style addresses, has proven to be a significant problem.  These rural areas cannot be updated directly from the U.S. Postal Service's Delivery Sequence File (DSF) resulting in poor data quality for about half the land area of the United States and about 15% of the U.S. population. (U.S. Census Bureau, American Community Survey Operations Plan, 2003, pp 9-10)

The U.S. Census Bureau has responded to this challenge by creating the Community Address Updating System (CAUS) that collects rural and rapidly growing community address data by working with community officials as well as by direct field visits by U.S. Census Bureau personnel.  The results of CAUS are fed directly into the MAF in an effort to base the questionnaires mailed out as part of the ACS process on the most current and reliable address data available. (U.S. Census Bureau, American Community Survey Operations Plan, 2003, pp 9-10)

If a questionnaire is not returned by mail within six weeks, the U.S. Census Bureau will then follow up with a Computer-Aided Telephone Interview (CATI) for approximately one in three of these non-responding households.  If the household is unable to be contacted by telephone, the U.S. Census Bureau will then send field representatives out to conduct Computer-Aided Personal Interviews (CAPI) at rates that vary between one in two non-respondents to one in three non-respondents.  These follow-up rates will depend upon the level of responses that have already been received for the census tract or geographic area in question from the mail or CATI processes. (U.S. Census Bureau, Survey Basics Website)

Using the mail, telephone, field visit strategy, the U.S. Census Bureau expects to achieve the necessary levels of completed surveys to have a statistically valid sample data set.

Rick Ayers wrote in the Winter 2001-2002 issue of ESRI News that the purpose of the United States census is two-fold.  First, the census counts the population of the country in order to provide data for the reapportioning and redistricting of representatives to the United States House of Representatives.  Second however, starting in 1940, detailed information on the country's demographic, housing, social, and economic situation has been collected at the same time using the proverbial "long form."  Rick argues that the management of this "long form" data collection process has made the process of taking the census far more complex.  In addition, since this "long form" data is collected every 10 years, the information quickly becomes outdated.  Rick quotes Nancy Torrieri, a senior program analyst for the Demographic Survey Division of the U.S. Census Bureau that, "there was a need for a system that continuously collects data throughout the year and the decade to produce measures of change over time and more current data than the decennial census long form."  (ESRI News, 2002)

When fully operational, the American Community Survey will eliminate the need to collect "long form" data beginning with the 2010 census, instead relying on an ongoing series of surveys that will provide equivalent rolling, or continuously updated, information. (U.S. Census Bureau, ACS Press Release, January 10, 2005)

4.  How is the quality of the American Community Survey data assessed?

The Fall 2004 American Community Survey Fact Sheet indicates that the ACS has been tested since 1996 to ensure the same quality of data as the census long form. (U.S. Census Bureau, ACS Fact Sheet, 2004)

To this end, ACS test data has been collected and evaluated using four measures of data quality: Coverage Rates, Sample Sizes, Response Rates, and Item Allocation Rates. (U.S. Census Bureau, ACS 'Using the Data' Web Site)

Coverage Rates:  Errors in coverage rates exist when either housing units or people have more than one chance to be included in the sample, or when they do not have a chance at all of being selected for the sample.  The ACS population and housing estimates are controlled by independent samples of population and housing data. (U.S. Census Bureau, ACS 'Coverage Rates' Website)

Sample Size:  The ACS publishes confidence intervals, standard errors, and degrees of uncertainty for each data sample so that data users can determine how much confidence can be given to the data.  The ACS has sampled between 800,000 and 900,000 addresses annually from 2000 through 2003 as a part of its testing program. (U.S. Census Bureau, ACS 'Sample Size' Website)

Response Rates:  Non-responsive surveys occur when respondents are unable or unwilling to participate in the survey, when interviewers are unable to locate either the addresses or the intended respondents, or when other barriers exist to completing the survey.  The ACS calculates and publishes survey response rates in an effort to monitor the effectiveness of the survey.  (U.S. Census Bureau, ACS 'Response Rates' Website)

Item Allocation Rates:  The ACS calculates the amount of missing data for a particular question or if the respondent fails to provide an appropriate answer to a question.  These values are collected and published so users can judge the completeness of the data. (U.S. Census Bureau, ACS 'Item Allocation Rates' Website)

Additionally, ACS test data that has been collected between the years 2000 and 2003 are being compared with 2000 Census long form data for selected geographic regions to test the both the quality of the data as well as to provide users with information to make the transition from the long form data to the full ACS data set.  The initial data comparisons will be published in 2006 for areas with populations of greater than 65,000.  (GAO Report, ACS - Key Unresolved Issues, 2004, page 6)

While the American Community Survey has received endorsements from numerous public and private organizations (American Statistical Association 2001, Edward J. Spar Website, Cynthia Taeuber 1998, and Pari Sabety 2004 ), the U.S. Government Accountability Office (GAO) has identified several key issues that it believes need to be resolved.  While each of the issues does not invalidate the ACS, each has profound implications on how the data is interpreted and used.  The GAO found the Census Bureau deficient in explaining how to transition from the current decennial census to the ACS concept of continuous updates.  (GAO Report, ACS - Key Unresolved Issues, 2004, page 19).  The key issues found by the GAO are:

The Concept of Residence:  Previous census data was collected from residents and recorded for a geographic area according to a concept of "usual residence."  This concept meant that persons who lived in Pennsylvania but who resided in Florida part of a year would have their data counted as residents of Pennsylvania.  The ACS has changed this to a concept of "current residence."  The Census Bureau has argued that this is a more fitting approach because the data will be collected on a monthly basis but then used to produce annual averages rather than point-in-time estimates.  Using this approach provides new residence rules for seasonal or migratory individuals and will have repercussions depending upon where an individual is currently residing at the time of the survey. (GAO Report, ACS - Key Unresolved Issues, 2004, page 7)

Reference Period:  Because the ACS is a monthly survey, data will be averaged for yearly values.  The U.S. Census Bureau has assumed that this will be equivalent to data for July 1 of each year in question.  By averaging data over months and years, rapidly growing or declining populations will be temporarily under-represented or over-represented in the data. (GAO Report, ACS - Key Unresolved Issues, 2004, page 8)

Independent Controls for Population and Housing Characteristics:  As mentioned above, the ACS will use population and housing characteristics from independent sources and not from the survey results.  Unfortunately, these independent sources use the "usual residence" concept for seasonal or migratory individuals creating a basic incompatibility with ACS data and for properly interpreting the results.  Further, the GAO found that the independent data sources used do not adequately represent populations in small geographic areas that may result in under-representation of these areas, just the areas the ACS was designed to capture.  (GAO Report, ACS - Key Unresolved Issues, 2004, page 8)

Dollar-denominated Data:  Because the ACS is a rolling survey, information collected that is based on dollar-denominated values such as income, rent costs, or other expenses will need to be adjusted for inflation in order to compare this data over time.  The ACS has chosen to average this monthly data to arrive at annual data values, and the ACS will be using the Consumer Price Index for the annual and monthly adjustments to dollar-denominated data for all areas.  Depending upon the rate at which the CPI changes, dollar-denominated values can be over-inflated or under-inflated at the time of the analysis. (GAO Report, ACS - Key Unresolved Issues, 2004, page 8)

Group Quarters:  The 2000-2004 ACS supplementary/test surveys used to evaluate the ACS methodology, chose to exclude persons living in group quarters, e.g. nursing homes, prisons, college dormitories, military barracks and emergency or transitional shelters for the homeless.  These group quarters situations account for nearly 3% of the total U.S. population and could represent a significant undercount of population and housing data.  The ACS has advised the GAO that it has addressed this issue and will be testing procedures for collecting data from group quarters prior to the 2010 census.  (GAO Report, ACS - Key Unresolved Issues, 2004, page 9)

In addition to these technical issues, the success of the ACS is dependent upon the continued good will of the Congress and the President to provide funding for the effort.  Because data is collected on a rolling basis, reducing the ability of the Census Bureau to collect this data will have profound implications.  This is an effort that once started must certainly continue in order to maintain the consistency of the ACS approach.  In keeping with this concept, the President's FY06 Budget submitted to Congress on February 7, 2005, included a request of $169.9 million to fully fund the continuation of the ACS and the Puerto Rico Community Survey (PRCS).  (American Community Survey Alert, Number 30, 2005)  However, it remains to be seen whether this support will continue in periods of fiscal austerity,

Despite the concerns expressed by the GAO, the American Community Survey represents a significant change to both the way population and residence data is collected as well as to the actual data itself.  By moving to a rolling data collection and publication methodology, individual states and municipalities will see significant changes to their allocation of federal dollars, either through block grant programs or population-based funding formulae.  With continuously changing data, updated on an annual basis, federal funds for such wide ranging items as community policing to road construction will now change on an annual basis.  States will no longer have the luxury of knowing what they can count on for their federal allocations from one year to the next.  While statisticians and community planners may enjoy the availability of continuously updated information to assist with project planning, those responsible for managing the budgets for those projects may not appreciate the uncertainty.  In this kind of environment, I predict that a cottage industry of economic and budget forecasting will emerge based upon the marriage and tighter integration of ACS data with Geographic Information Systems.  States, counties and municipalities with declining populations will lose federal funding allocations at a much faster rate, and states, counties, and municipalities with rising populations will experience increases to their federal funding allocations at a much faster rate.  In terms of federal funding, the rust belt with its declining population will lose federal dollars at a much faster rate; and the sun belt, with its rising population, will benefit from a faster rise in federal dollars.