Statement of Linda Jacobsen, vice president, Domestic Programs, Population Reference Bureau, before the Joint Economic Committee, U.S. Congress, July 21, 2009

Good afternoon, Chairwoman Maloney and members of the Committee. Thank you for the opportunity to testify today before the Joint Economic Committee. I will share my perspective on the benefits and challenges of using the Census Bureau's American Community Survey.

Without a doubt, the American Community Survey (ACS) is fundamentally changing the way we collect and use data to assess the nation's population and housing. While the traditional census long form collected detailed socioeconomic data just once a decade, the ACS is a continuous survey that provides updated demographic, economic, and housing data every year. As the pace of change has accelerated in the U.S., so has the need for timely and reliable data. The ACS has replaced the census long form to meet that need.

The ACS is already providing substantial benefits to federal agencies, nongovernmental organizations, and state and local governments. For example, the Department of Veteran Affairs uses ACS data to evaluate the need for educational, employment, and health care programs for veterans, while the Council on Virginia's Future relies on ACS data to monitor annual trends in the travel time to work. At PRB, we have used ACS data to track changes in the number, location, and well-being of children in immigrant families. We have also used the ACS to produce a database and wall chart on the U.S. labor force, including state and metropolitan area estimates of people working in high-tech and other science and engineering jobs.

ACS data are also contributing to planning for the 2010 Census. ACS data from 2005, 2006, and 2007 were used to validate and enhance population segmentation for the Census 2010 Integrated Communications Campaign. The Census Bureau is also using 2005-2007 ACS data on language spoken at home and English-language ability to select census blocks that will receive a Census 2010 bilingual English and Spanish form.

There are some important differences between the census long form and the ACS that are essential to understand in evaluating the benefits and trade-offs in the switch to the ACS. Foremost is the fact that the sample size of the current ACS is much smaller than the sample size of the 2000 Census long form. As a result, ACS data from multiple years must be combined to provide reliable estimates for geographic areas with smaller population sizes. The ACS provides 1-year estimates for areas with populations of at least 65,000, 3-year estimates for areas with populations between 20,000 and 65,000, and 5-year estimates for areas with less than 20,000 people. This last group includes small counties, cities, and towns as well as census tracts and block groups. The ACS was fully implemented nationwide in 2005, so the first 3-year estimates for 2005-2007 were released last December. The first 5-year estimates for 2005-2009 are scheduled for release in 2010.

In 2000, the long form was sent to approximately 18 million addresses, resulting in 16.4 million final interviews. This represented about 1 in every 6 households, the same share of households that received the long form in 1990. In contrast, the ACS is sent to about 3 million addresses each year, resulting in about 2 million final interviews. When combined over five years, then, the ACS will only be sent to 15 million addresses, resulting in about 10.5 million final interviews. This represents only about 1 in every 9 households. Of course, the number of households in the U.S. continues to increase every year. Between 2000 and 2007, the number of households increased by 7 million. While the decennial long form sampled the same proportion of households in 1990 and 2000, the ACS samples the same number of households each year. Maintaining a fixed sample size over time necessarily means that ACS data will be collected from a smaller share of U.S. households each year.

As a result of the smaller sample size, estimates from the ACS also have higher levels of sampling variability than estimates from the 2000 Census long form. This means the ACS estimates are less precise or less reliable, particularly for small geographic areas and population subgroups. Several evaluation studies have reported that combining 5 years of ACS data did not provide reliable estimates for census tracts in some counties. In a recent PRB analysis of 2005-2007 data for 26 states, we found that one-fifth (20 percent) of counties with a population of 20,000 or more did not have a reliable estimate of the share of working families that are below 200 percent of the poverty level. Although this clearly constitutes a smaller population subgroup, there were still more than 9.5 million such families nationwide in 2007, and they are an important group for policy considerations. To achieve the objective of fully replacing the long form, the ACS must provide a comparable scope of reliable data for smaller geographic areas, including census tracts. The current sample size of the ACS is the result of funding constraints. Based on their experience to date, a growing share of data users are calling for an increase in the sample size of the ACS to improve the reliability of estimates for smaller geographic areas and subgroups. The ACS has tremendous potential to provide the timely, detailed data critical for evidencedbased policy and program design and implementation. Additional funding could significantly increase the likelihood the ACS will realize this potential.

Thank you for this opportunity to testify and I will be happy to answer any questions the Committee may have.