The U.S. Decennial Census and the American Community Survey: Looking Back and Looking Ahead

(April 2011) On March 24, 2011, the U.S. Census Bureau released the final 2010 Census redistricting data files for each state that will be used to redraw federal, state, and local legislative districts. Although the 2010 Census data are still new, the Census Bureau has already started evaluating the data and planning for the next decennial census in 2020.

As part of PRB’s Policy Seminar series, Robert Groves, director of the U.S. Census Bureau, and Frank Vitrano, associate director for the 2020 Census, discussed the quality of the 2010 Census, the future of the American Community Survey (ACS), and early plans for the 2020 Census. View their PowerPoint presentation.

In the first half of the seminar, Groves described several ways to measure the quality of a census, and shared some early indicators of the quality of the 2010 Census data. One evaluation method is to compare census results to alternative estimates such as those from the Census Bureau’s Demographic Analysis program or their annual population estimates program. Groves noted that the 2010 Census count of total U.S. population was very close to both the middle estimate from Demographic Analysis and the vintage 2009 population estimates. In addition, at the state level, the percent difference in total population between the census count and the annual population estimate was much smaller for the 2010 Census than it was for the 2000 Census. Groves also described several positive indicators of the quality of Census 2010 data collection processes, such as fewer duplicate records (people counted twice) and a higher share of records with usable data.

There were changes in some census operations that also seemed to have improved data quality in 2010 compared with 2000. For example, distributing bilingual questionnaires to Spanish-speaking households and sending replacement forms in hard-to-count communities increased participation rates. The Census Bureau also used real-time evaluation in the 2010 Census. When results from a precensus survey showed that 18-to-24-year-olds were not inclined to participate in the census, the bureau ramped up their digital advertising campaign in the days prior to April 1 (Census Day) to boost response rates. The 2010 short-form design also appeared to be effective—there was a participation rate of 72 percent compared with 69 percent using the combined short/long form design in 2000. Finally, Groves observed that the availability of high-quality enumerators—due to the recession—may also help explain the improved results in 2010.

The Census Bureau’s post-enumeration survey is another important evaluation tool, used to measure census undercount and which population subgroups were most likely to be missed. Final post-enumeration survey results will be released in 2012. Researchers can also use estimates from the ACS to help interpret the 2010 Census results and put the numbers into a broader social and economic context.

In the second half of the seminar, Frank Vitrano discussed the future of the decennial Census and the ACS, and outlined early plans for 2020. The cost of the census has been rising rapidly since 1970, and one of the primary goals for the 2020 Census is to reduce cost. Vitrano described four key cost drivers:

  • Increased population diversity and decreased willingness to participate in the census or respond to nonresponse follow up.
  • Limited intercensal updates of address files and maps, leading to last-minute changes.
  • Failure to effectively manage and continuously update decennial census planning, schedule, and budget.
  • Rising costs of producing high-quality data, and stakeholder demands for data that are 100 percent accurate.

Vitrano presented several potential measures to reduce costs, including expanding and automating self-response methods, adding an Internet option, and using administrative records for follow up. Administrative records could potentially fill in some of the data gaps for households that cannot be reached through traditional methods. He outlined potential improvements in program management, budgeting and scheduling and noted the need to build consensus among stakeholders regarding the tradeoff between accuracy and cost.

The Census Bureau is currently considering six designs for the collection of 2020 census data, including three alternatives for information technology infrastructure. Vitrano noted that key decisions must be made by 2015 in order to avoid increased risk and cost due to late design changes.

Redistricting data from the 2010 Census are available through the Census Bureau’s American FactFinder website:

For a schedule of data to be released from the 2010 Census, visit the Census Bureau’s website: