June 13, 2001
Elaine Murakami, Nandu Srinivasan and Ed Christopher
The hearing included an introduction by Chairman Dan Miller (R – Florida) followed by two panels. Fore the first panel, William G. Barron, Jr., Acting Director, and U.S. Bureau of the Census presented a prepared statement and answered questions. The second panel consisted of four expert witnesses who were present and two others who submitted written statements for the record.
Chairman Miller, Rep. Carolyn Maloney (D-New York), Rep. Bob Barr (R-Georgia), Rep. Wm. Lacy Clay (D-Missouri), and Rep. Chris Cannon (R-Utah) questioned the presenters.
The text of the presentations can be accessed from Chairman Miller’s website at http://www.house.gov/danmiller/census/hearing/hearing1.html#6.13.01hearing
The Honorable Dan Miller, Chairman of the Subcommittee, opened the hearing by remarking on the success of Census 2000. He said that ACS would simplify the 2010 Census, and bring us closer to a “postcard census”. The other major advantages of the ACS are its ability to provide up-to-date and timely social, economic, demographic, and housing information.
Chairman Miller noted that Congress has some issues on ACS to which
he hoped the Census Bureau (CB) would respond. These include:
1. Cost: CB asked for $131 million in the FY03 budget for ACS.
2. Content: The ACS questionnaire asks 69 questions, while the Census 2000 long form asked only 53. This raises issues of intrusiveness and privacy. Chairman Miller wanted the establishment of a predetermined and definitive process by which to alter the ACS questionnaire for additions or deletions of questions.
3. Requiring ACS to be Mandatory: What are the implications if ACS was a voluntary survey?
4. Avoiding duplicating other ongoing survey work, particularly the Current Population Survey (CPS).
Panel 1: Prepared Statement of William Barron
Mr. William Barron reported that the ACS is one of three key components for re-engineering the 2010 Census:
1. Improving the accuracy of the geographic database and Master Address File.
2. Eliminating the long form from 2010 by collecting those data through ACS.
3. Re-engineering the census process through early planning.
He also reported that the CB initial estimates of life-cycle costs for a re-engineered Census (including ACS, and only the short form for 2010) will demonstrate “cost neutrality” when compared with the cost of repeating Census 2000 for 2010 (without ACS). The estimated cost of repeating Census 2000 for 2010 is $12 billion (using standard assumptions). He said that the advantage of the re-engineered Census is that at the same cost, much more data will be produced throughout the decade.
In response to the questions from members of the subcommittee, Mr. Barron pointed out that the first “Census Tract” level data from ACS will be available in 2008, and that the Census Bureau and Office of Management and Budget (OMB) are going through a process for redefinition, and re-examination of questions asked in the ACS.
Chairman Miller asked if a large survey such as ACS can eliminate some of the other surveys such as the CPS. Mr. Barron replied that the two surveys have very different objectives, and the ACS cannot eliminate the need for CPS. However, the ACS complements the data collected by the CPS. Chairman Miller also wanted to know if funding the ACS fully means that Census for 2010 will be mailed as a “post card”. Mr. Barron replied that it will be a much shorter Census, but the exact survey instrument has not yet been identified.
Rep. Maloney asked Mr. Barron why the ACS data for Bronx County, New York was retracted. Mr. Barron replied that the data was not retracted, but the Census Bureau found that the data could be reported more precisely by using better weights. Initially, the Census Bureau weighted the ACS data based on the State and County population estimates program. In the case of Bronx County, CB found that the ACS results using these estimates are significantly different than results from Census 2000. So, the CB used ACS as an input into the State and County population estimates, and then re-weighted the data.
In reply to another question, Mr. Barron replied that the CB has no plans to reprogram remaining funds from decennial census to the “State and County Population Estimates” program. However, the Census Bureau Demographic Analysis Program team is working hard to ensure that the estimates are closer to reality.
Rep. Maloney also questioned the CB on the program to remove duplicate
records from households and group quarters for Census 2000. She wanted
to know why the Census Bureau was focusing on duplicate records and not
equally researching missing records.
Rep. Barr questioned the length of the ACS questionnaire, and stressed his concern for privacy and the intrusiveness of the ACS.
Panel 2: Prepared Statements of the expert witnesses
The main concern of Dr. Paul R. Voss was that the CB should base rural ACS data on sufficiently large sample for the data to have a statistical precision similar to that provided by the Census long form. The very smallest governmental units in the country are now sampled at 7.5% annually, yielding an effective sample size of 37.5% in five years as opposed to the 50% sample used in the Census long form. The current plans at the CB aim for uncertainty levels around 33% larger than long form estimates. The other key issues raised were:
1. The precision of the data at small area geography.
2. Boundary changes and how they may affect the presentation of the results of the survey.
3. The implications of the differences in data collection procedures between the ACS and the Census long form.
4. Multi-year averaging of data.
Marilyn M. McMillen distributed copies of a paper documenting
the uses of Census data by the Department of Education. She stressed
on four topics:
1. Statistical Reporting on critical topics on education
2. Enhancing current data
3. Enhancing on going data collection schemes, and
4. Importance of ACS data to ensure fair and equitable distribution of education funds.
She said most of the Census data used by the National Center for Education Statistics (NCES) come from the CPS. The overlap of ACS and CPS will allow to “drill down” CPS data and obtain data at lower levels of geography on key items. NCES sees ACS as a means of expanding the range of topics about which data are collected. “The size of ACS ensures that sufficient numbers of households containing “rare populations” could be identified throughout the decade.”
Dr. Donald J. Hernandez said that the Population Association of America and the Association of Population Centers strongly endorse the on-going development and the evaluation of the ACS. He highlighted the value of the census long form data and the ACS data, and then reviewed the potential advantages of ACS. Dr. Hernandez stressed that the most important condition is that the ACS must be fully funded during each year of the present decade. He added that the sample size for the ACS should not be constant every year, but should account for anticipated changes in population throughout the decade.
Taking the extraordinarily large and complex nature of ACS, he identified
five major issues for evaluating success:
1. Topical content
2. Evaluation of the test data
3. Response rates
4. Weighting and inter-censal estimation
5. Full-scale evaluation
He said one of the most challenging issues for the CB is to develop effective weighting and estimation procedures, if the ACS were to provide high-quality data for small geographic areas and population groups. The quality of the inter-censal estimates used by the CB in weighting the ACS deteriorates over the course of the decade between decennial censuses. Moreover, CB comparison of 2000 Census results with inter-censal estimates strongly suggest that the quality of the migration component of demographic analysis has declined during the past decade. He stressed that inaccurate survey weights would lead to inaccurate estimates of population characteristics.
He proposed that the CB should:
1. Convene a standing advisory committee to propose innovative evaluative approaches and analysis of existing and future ACS data.
2. Create a mechanism for identifying and funding researchers.
3. Convene an annual conference devoted to ACS.
4. Publish results of the research on ACS for wider distribution and comment.
5. Develop a formal mechanism for making changes to ACS based on research findings and experiences.
Linda Gage said that it is premature to endorse the ACS as the preferred method for collecting long form data in 2010. She said that the full development of ACS is contingent upon adequate funding, maintenance of a current and comprehensive master address file, and successful implementation of the survey over the next seven years. The CB will release ACS data for areas with population less than 20,000 people only in 2008. These areas constitute 92% of all the cities in California. Thus, any realistic assessment of ACS can take place only in 2008. The key points in Ms. Gage’s presentation were:
If the current sampling plan of ACS were continued through the decade without accounting for the growth in population, the ACS may go to only 1 in 7 of the households nationwide as opposed to the 1 in 6 sample used in the decennial census.
The Census Bureau estimate for Census Day, April 1, 2000 was 6.9 million lower than the number counted in the Census. Thus, if the ACS were not controlled to accurate population estimates, the long form data produced will be seriously flawed.
3. Evaluation scheme
The timeliness of ACS data is a benefit only if the data are accurate and comparable for all areas. There has been no rigorous evaluation of the ACS long form data collection success measured against the 2000 Census data.
Success in the thirty-one comparison sites and in the nation’s states and largest jurisdictions will be encouraging, but not definitive. She said that the ACS would benefit from a published evaluation plan and schedule including: results to date, case studies, and evaluations, comparisons of the survey results based on 1990- and 2000-based estimates with actual census 2000 data items, comparison of data quality with independent data sources, currency and completeness of the master address file, possible improvements to survey procedures, response rates, sample size, and coverage by race or ethnic group.
She strongly recommended that a decision date, along with milestones and critical measurements, be established and monitored to support a recommendation and decision to use ACS as the instrument for collecting long form data.
Questions posed to panel 2:
Rep. Clay asked the panelists if funding ACS is the best decision Congress can make. The panelists replied that it would be a smart decision. The panelists felt that the ACS fulfils important statistical needs of the country.
The panelists rated the CB ACS staff highly on their response to user
questions and comments.