PREPARED STATEMENT OF
Mr. Chairman, Mrs. Maloney, and Members of the Subcommittee:
I am pleased to be here today to discuss the American Community Survey (ACS). The ACS, part of our Continuous Measurement program, is the single most important improvement in federal household statistics since the middle of the last century and I believe it is the cornerstone of our effort to keep pace with ever increasing demands for timely and relevant data. If the Census Bureau and the Congress are to continue leadership in providing information to the knowledge society, we must invest in full development of the ACS. In our web-based society of the 21st century, where information is increasingly available at the public's fingertips, it is no longer acceptable that planners and policy-makers must use data that are up to 12 years out of date. Not only will the ACS provide more frequent detailed data for small geographic areas, but it will also revolutionize the way we take the decennial census.
To understand fully the need for the ACS, we must discuss it in the context of the United States' longstanding need for detailed social, economic, and housing data to further the business of governing. The very first census in 1790 went beyond the bare constitutional need for a headcount. Since 1820, each census has included questions that describe the condition of our society and economy. In fact, it is no coincidence that "statistics" and "state" share the same common root. As the United States grew to become a powerful nation, statistics came to play an increasingly larger role and, by the late 1800's, the compendious census questionnaires reflected that. The 20th century can be described as the first "measured century," and this helped to build our democracy and our free economy.
The Census Bureau introduced a long form on a sample basis into the 1940 census, and we have used this approach in each decennial census since then. The selection of a sample, based on established scientific methods largely developed by Census Bureau professionals in the 1930's, means that not everyone is asked every question; the large majority receive only a short form, thus, lessening the burden on the public. The long form for Census 2000 includes the questions asked on the short form and additional questions, for a total of 52. In contrast, the long form included 57 questions in 1990, and 81 questions in 1940.
Over the years, the long form included in the decennial census has met the Nation's need for detailed data essential for governing. The decennial census long form has provided reliable, consistent data for small geographic areas--as small as city neighborhoods--and it has done so in a very efficient manner. So why are we here today talking about an alternative to the long form?
In the early 1990's, two streams of argument converged to make a powerful case that change was needed. First, was the demand from a wide variety of data users, including those in Congress, for more timely and relevant data than that provided by a long form collected only once every 10 years. I will talk more about this need for up-to-date information later. Second, was the idea that the long form was a drag on the decennial census, a complication in carrying out the constitutional purpose of the census to obtain the most complete and accurate count of the population possible. The solution was to simplify the census by eliminating the long form.
To eliminate the long form without an alternative method of collecting detailed population and housing characteristics would have been irresponsible. Every question has a specific federal legislative or judicial mandate or requirement. When the Census Bureau reported to Congress on March 30, 1998, as required by law, on the specific questions that would be asked in Census 2000, the Census Bureau provided selected citations from about 130 laws that use specific long form items, and that was not a complete list. As one example, there are 25 citations alone for the disability questions. These include the Americans with Disabilities Act; the Rehabilitation Act; the Disabled Veterans Outreach Program; the Old Age, Survivors, and Disability Insurance Program; and the Vocational and Applied Technology Education Act. Without the decennial long form or an appropriate alternative, such as the ACS, Congress would not have the data it mandated to run the programs it has written into law. In effect, the Congress would not have a rudder to guide the ship of state. Indeed, we are aware of at least 35 new bills that have been introduced in the 106th Congress that would require long form data.
The ACS, then, is the statistical community's alternative to the decennial census long form. If fully funded and implemented it will lead to a simpler decennial census, provide more timely and relevant data, and also lead to more widespread efficiencies throughout the federal statistical system. I would like to discuss each of these items in more detail now.
A Simpler Decennial Census
If the ACS is fully implemented, we can eliminate the long form as part of the 2010 decennial census. As I mentioned earlier, elimination of the decennial census long form has been one of the key goals in developing the ACS. If the Census Bureau has the resources to conduct early planning, it will be able to re-engineer many aspects of the decennial census with a view to achieving cost savings related to not having a long form, enhancements to our geographic programs, and changes in other interrelated areas, such as forms design, printing, labeling, questionnaire delivery, enumerator training, nonresponse followup, data capture, data editing, and data tabulation. Estimates of potential cost savings are currently being examined and will be presented to the Administration as soon as available and, subsequently, to the Congress as part of the FY2002 appropriations cycle.
The ACS will also permit the Census Bureau to focus exclusively on conducting the basic count. The ACS will lead to better coverage of addresses in the census because it will require a continuously updated address list and ongoing interaction with local officials. The ACS will also provide updated data to permit the Census Bureau to more effectively target areas where special procedures will be required. As I have previously testified, the data used to determine hard-to-enumerate areas for Census 2000 was 1990 data. This did help, but up-to-date data would have been of much greater value. Census Bureau staff are also looking at ways to use the highly trained and seasoned ACS field staff as a cadre of key supervisory staff for the 2010 census. Because they will have been collecting ACS data successfully in hard-to-enumerate areas, they will have developed relationships with local leaders that will prove invaluable in conducting the 2010 census. So, in simplifying the decennial census by eliminating the long form, we also introduce opportunities for substantial cost savings and improved coverage.
I also believe the ACS, spread out as it is over 120 months, will reduce public concern about privacy. The long form for Census 2000 seems to have caught many people by surprise, with the result that there has been considerable discussion of the long form as an "invasion of privacy." I believe this was partly due to the way the long form was administered to some 20 million housing units all at one time. The ACS provides a different atmosphere and, thus, different opportunities than the census.
Instead of 20 million housing units receiving the census long form all at one time, in each monthly survey, only 250,000 housing units will receive the ACS questionnaire at any one time. Instead of having a large army of temporary enumerators, the ACS field interviewers will be highly trained permanent staff who will be better prepared to deal with the public's questions about the form. As I said at an earlier hearing, I am confident that once the public understands that their answers are protected by law and that every question asked serves an important purpose, they will perform their civic duty to respond. In fact, our experience thus far with the ACS has been that our trained interviewers have achieved good cooperation with few complaints from the public. I believe it is hard to sustain the argument that government data collection is an invasion of privacy when there are such strong protections of the data, when they are used only for statistical purposes--not for regulation or law enforcement--and when each questionnaire item is linked to a program that the people's representatives have enacted.
Having said that, the Census Bureau takes questions and concerns about intrusiveness and privacy seriously. We are aware of the time pressures confronting people and of the concerns we all have about privacy and confidentiality. We also recognize and respect those who have philosophical objections to government activities, government programs, and government data collection. Our Nation has struggled with these profound issues from its inception. I do not believe the Congress has invaded privacy by establishing government programs to help people or by requiring good data to administer those programs, but I respect the views of those who do. I also pledge that the Census Bureau will do all it can to work with the Congress as you struggle with these weighty issues. It is not up to the Census Bureau to decide what kind of society we should be or even whether we should have timely and relevant data to make that vision possible. It is our role, however, to inform the Congress about the most efficient, effective, and modern ways to collect the data, once the Administration and the Congress have determined which data are necessary. That is what we are doing today by presenting our plans for the ACS.
I want now to say a word about the process for developing ACS content, which I expect to involve an active partnership between the Congress, federal agencies, and the statistical community. Currently, the ACS content closely resembles the Census 2000 long form. That's because, as I will explain in more detail later, we need to compare responses to the ACS with those from Census 2000. We need to understand how differences in data collection methods between the two affect the results. But we also realize that what is justified for a once-a-decade data collection needs to be reassessed.
One step in this reassessment will be the Interagency Committee on the American Community Survey, which the Office of Management and Budget (OMB) has established and which met for the first time last week. This group will be the primary mechanism for proposing changes to the ACS questionnaire. The group will work with all the departments and agencies in the Executive Branch to examine whether information is needed for small geographic areas (census tracts, small towns, and rural counties) to determine how frequently the data are needed.
Close collaboration with the Congress will be another factor in determining ACS content, and I look forward to any suggestions you may have for how we can accomplish this. One problem we all share is how to balance concerns about minimizing respondent burden with requests for additional information. Just as some in Congress want minimal data collected, others have expressed the need for additional questions. The only new question added to Census 2000--that concerning grandparents as caregivers--was required as part of the welfare reform legislation. The Census Bureau determined that the data should be collected on a sample basis, although some Members supported asking it of all households. And, there was a sense-of-the-Senate resolution expressing concern about an item that had been moved from the short to the long form. So we must always look for ways to balance these competing pressures and we look forward to working with the Congress to do that.
Timely and Relevant Data for Small Geographic Areas
In addition to permitting improvements to the decennial census, the ACS will provide timely and relevant data for small geographic areas. Data users have demanded more timely data that provide consistent measures for all areas. Decennial census data are already out-of-date 2 years after the census is taken--when long form data are first released--and their relevance declines with every passing year. The ACS will provide essentially the same information every year that the decennial census has provided only once every 10 years. In this new century, especially with the trend to devolution of responsibilities from the federal to the state and local level, communities will need accurate, timely information to chart their social and economic condition. And federal agencies need these data to monitor and evaluate the effectiveness of the programs that are funded at the national level but designed and implemented by states and localities.
The ACS is not designed to provide a headcount. That is what the decennial census is for. The Census Bureau has a population estimates program, conducted in partnership with the Federal State Cooperative Program for Population Estimates, to provide updated population totals for years following the most recent census. The ACS will serve a similar role in providing updated population and housing characteristics, which, among other uses, would enable the allocation of billions of government and business dollars among states, communities, and population groups.
The ACS will also provide timely data to help federal program managers and the "end users" in every community across America. It recognizes the continuing roles of Congress and federal agencies while also elevating partnership with other levels of government and those directly affected by the decisions based on data.
Let me give one example at the Federal level: The Individuals with Disabilities Act allocates the federal portion of the cost of educating children with disabilities based upon a formula that includes disability information now gathered on the decennial census long form. As you know, this information becomes outdated quickly. The last long form data (from the 1990 census) on children with disabilities in the 1989-1990 school year will continue to be used until the 2002 or 2003 school year--after a complete generation of students has come and gone. They reflect the situation of older children, teenagers, and young adults--individuals who are at least 10 years old today. None of the children 9 years old or younger who receive assistance from this program were included in the 1990 census. In the next decade, with the ACS, this problem of using old data will be solved because up-to-date information will capture the changes in characteristics between censuses.
Similarly, the ACS can identify changes in an area's population and give an up-to-date statistical picture that communities can use to:
track the well-being of children, families, and the elderly;
determine where to locate new highways, schools, and hospitals;
show a new or relocating corporation it has the workforce the company needs;
evaluate the quality of available housing;
evaluate programs such as welfare and workforce diversification; and
monitor and publicize the results of their programs.
The up-to-date data that we will issue from the ACS will help communities establish goals using objective information, increase program accountability by measuring results, and measure program performance for future planning. It will enable communities to see how they are doing compared with others, to identify problems and find solutions, and to make programs more flexible to meet the needs of different groups in the community.
Let me give one specific example of how ACS is already proving itself more useful than the decennial census long form at the community level because it is updated and available every year instead of just once a decade. In Oregon, grants and tax credits are provided to developers of affordable housing and housing designed for the elderly and disabled. To obtain grants, developers have to demonstrate need based on demographic characteristics such as the numbers of low-income, elderly, or disabled households, or the size of other disadvantaged populations. This detailed information comes only from long-form type questions. In the past few years, a developer wanting to qualify for such grants would only have had 1980 and 1990 long form data available. Because of concern that extrapolations of such data would not reflect current conditions, the Population Research Center (PRC) of Oregon State University was asked to determine the need for low-income multi-family housing.
Knowing that the ACS had recently been conducted in Multnomah County, Oregon, the PRC devised an assessment tool to compare results from trend analysis based on 1980 and 1990 long form data with results from the 1996 ACS. They compared findings on age, race and ethnicity, household size and type, employment status, income, vacancy and tenure of housing units, gross rent, gross rent as a percentage of household income, quality and age of housing, and other related variables. This comparison showed that many of the actual changes in the area would not have been identified by extrapolating based on 1980 to 1990 trends. Data from the 1996 ACS showed a need for more affordable housing units, whereas the 1980-1990 trend data did not indicate any need.
A More Efficient Statistical System
The ACS offers the potential for dramatic improvement in the infrastructure of the federal statistical system. It is now costly and time consuming to create samples for new surveys or to adjust samples for existing surveys to provide data to meet the needs of policymakers in unforeseen areas.
The ACS can make the entire federal statistical system more efficient by serving as a screening mechanism for new surveys. One example in which the ACS is already doing this is with the National Epidemiological Survey of Alcohol and Related Conditions, which the Census Bureau conducts on a reimbursable basis for the National Institute on Alcohol Abuse and Alcoholism. This survey is designed to gauge alcohol consumption, as well as to provide psychiatric classifications of alcohol use disorders. Ultimately, the data are helpful in determining the causes of alcoholism, and in understanding the behavior patterns and potential treatments or other courses of action to help sufferers, their families, and ultimately, society combat this disease. In 2001, this survey will become the first to use data from the ACS to draw its sample and expects to realize a major efficiency by eliminating the need for interviewing a large number of households just to find the target population groups. For example, by using race and ethnicity information collected by the ACS, we can easily include the sample target groups with no additional respondent burden or field work.
The ACS can also be a valuable resource to avoid or reduce the cost of new surveys. For example, once the ACS completes a stable year of initial operations, it could add questions on health insurance coverage, and could meet the Congress's expressed need for data on children's health insurance coverage more efficiently than expanding other surveys or adding a new one. This is a good example of how the ACS can be a valuable tool to help Members of Congress get reliable data about emerging issues. Instead of having to develop entirely new surveys, the ACS --often in combination with existing national surveys--would be able to deal promptly with such emerging data needs.
In addition to consulting with the Office of Management and Budget and other federal statistical agencies in developing the ACS, we have worked closely with representatives of state and local governments and private organizations, who also find data collected for federal purposes of immense value. Collaboration with private contractors has been extremely valuable for Census 2000 and we plan to explore ways to reap benefits for ACS as well. Finally, we are also working with our colleagues from private sector survey, research, and data dissemination organizations. Representatives from these organizations were involved in early ACS product development meetings and serve on our standing professional advisory committees. Our standing advisory committees will have a major emphasis on the ACS during the next decade and we have private sector representation on these committees. Up-to-date small area address counts from ACS operations will be beneficial to private companies conducting surveys, small area data on characteristics, such as race and ethnicity, will help them stratify their area samples, and general-purpose data products from the ACS will allow the private sector to develop new specialized data products that contain current, rather than outdated, information.
Mr. Chairman, in your letter of invitation you asked a number of technical questions about how the ACS will be conducted. I would like to address those questions now in the context of
describing what the ACS is, what work has been completed thus far, and the plans for full implementation.
Progress and Plans for the ACS
I want to begin by talking briefly about what the ACS is not. It should not be viewed as a "baby" decennial census. The decennial census requires the careful building of a massive infrastructure of offices, equipment, staff, advertising, and the like--all the things that I have discussed with you over the last several months. The ACS will not require additional offices, but will be run out of our regional offices. Field interviewers, who work on other surveys in addition to the ACS, will use existing laptop computers, and laptops from Census 2000 can be used as replacements. We will not require special data capture centers, as that will be accomplished in our processing center in Jeffersonville, Indiana. We will use the Master Address File for drawing the sample, although a critical element in the overall success of the ACS will be the ability to keep the MAF up-to-date and accurate over the decade and the Census Bureau is building in processes to do that. We are working to provide electronic versions of maps that will reside on the interviewers' laptops; this will reduce the massive requirements for paper copies of maps we encounter in the decennial census. We do not anticipate, at this time, any advertising expenditures for the ACS because its small monthly sample size would make it inefficient to use mass marketing methods such as we needed for the decennial census. We will expect to use community partnerships, which we have found to be very effective in the early test sites. So the ACS will be based on making the most efficient use of existing systems and infrastructure.
The ACS is a large monthly household survey conducted using mailed questionnaires, telephone interviews, and visits from the Census Bureau's field representatives. Sample cases for the ACS will be selected using the Master Address File, which, as I said, will be kept up to date throughout the decade. The Census Bureau will mail questionnaires to each address in the sample and will use purchased lists of telephone numbers to interview households that do not mail back their form. We will select a one-in-three sample of the addresses not interviewed by mail or telephone and conduct personal visit interviews. For those households that do not respond by mail, the quality of data is improved by using well-trained, permanent interviewers with computers to complete interviews. Computer-assisted interviewing allows the Census Bureau to incorporate consistency checks of the data into the collection process. Data quality is further improved by the use of a permanent staff to code write-in entries for processing.
When the ACS reaches full implementation in 2003, there will be approximately 250,000 housing units in each monthly survey or about 3 million per year. The ACS will produce annual data on content items equivalent to the decennial census long form for all communities in the country, including small areas such as census tracts, small towns, and rural counties. The first data for communities of 65,000 people or more will be available in 2004, about 6 months after the 2003 data collection year ends, and every year thereafter. The first data for communities with fewer than 65,000 people will be available over the following 2-5 years because it will take that long to accumulate sufficient sample size for these smaller areas. By 2008, 4-5 years before detailed data could be expected to be available from the 2010 census, the first data will be available for the smallest areas and groups. It will be updated annually from then on.
Over the last 5 years of phased-in testing, the Census Bureau has demonstrated the feasibility of the ACS methodology. Prototype testing of the ACS began in 1996 in four American communities. By 1999, the number of sites involved in these small scale tests had risen to 31; these 31 sites consist of 36 counties. The ACS has one of the most successful household response rates of any demographic program conducted by the Census Bureau. The final weighted response rate for the 31 sites is over 97 percent.
Next week, the Census Bureau will release data for 21 of the 31 sites included in the ACS in 1999. The data being made available for these sites include long form topics such as housing, education, employment, and income. Comparable data will not be available from Census 2000 until 2002 and 2003. Appendix 1 illustrates the kind of data that will be available by showing a profile from a recent ACS data release.
Data are being collected in the same 31 sites this year, using the $20 million in the FY2000 appropriations for this purpose. The work is conducted out of the Census Bureau's regional offices and involves about 150 field representatives, most of whom also work on other surveys. An operational test to ensure that the ACS can be conducted separately from, but during the same period as, a decennial census, is also underway this year. This operational test is collecting data over 12 months using the ACS questionnaire from a national sample of about 700,000 households in some 1,200 counties. This operation involves about 1,100 field representatives, again, most of whom work on other surveys.
These data will also contribute to a comparison with data from Census 2000 that is necessary because there are differences in methods and definitions between the census and the ACS. Moreover, decision makers will want to compare an area's data to those from Census 2000. Comparisons using data from the operational test and from the 31 sites are essential to determine how much measured change between Census 2000 and future years of the ACS is real and how much is due to operational differences between the ACS and the census.
Data for Small Areas and Groups
The fact that the ACS sample size and design will not provide data for the smallest areas and population groups until late in the coming decade has led some to raise the concern that the ACS may be treating rural areas unfairly. This concern was expressed in report language in earlier appropriations bills. The Census Bureau takes this concern very seriously. Indeed, we wish it were possible to begin providing small areas and groups (and this is not just rural areas but urban census tracts and smaller population groups, whether urban or rural) with high quality, current data right away. But that would basically involve replicating the decennial long form every year, and that is not acceptable in terms of costs or respondent burden. Once again, we are faced with balancing competing demands.
But the Census Bureau takes even the perception that rural areas are being treated unfairly very seriously and has worked with rural data experts to allay those concerns. The Census Bureau contracted with WESTAT to bring together rural experts and get their suggestions for improving the ACS from the perspective of rural areas. The group of a dozen experts, who met in May, 1998, expressed strong support for the ACS and provided several suggestions for further research and changes that might make the ACS even more valuable to rural areas.
The Census Bureau has also worked with the Rural Policy Research Institute, based in Columbia, Missouri. Their Executive Director, Chuck Fluharty, has co-authored a paper presented to the Congressional Rural Caucus entitled "The Importance of the American Community Survey for Improved Rural Community Policy." We have also worked with local planners in two of the smaller rural sites among the 31 sites--Otero County, New Mexico, and Fulton County, Pennsylvania. Even with the understanding that they will receive data later than large urban areas, planners in both of these counties strongly support ACS because it will eventually give them annually updated, current information each and every year--a much better situation than they have now.
So while concerns have been raised about data for rural areas, the Census Bureau believes that the current ACS design is widely seen as a major improvement over the existing situation.
Mr. Chairman, I appreciate this first opportunity to discuss the ACS in some detail before the Subcommittee. I hope we will have many more opportunities to brief you on this innovative development. The ACS represents the best that the Census Bureau and the entire statistical system, working together, can achieve when new strategies are required to address current and future data needs. Over the last several months, we have had extensive discussions about Census 2000. Much work is yet to be done on Census 2000, but it is not too early to begin looking to the future. The ACS is that future and the future is already here.
I credited the cooperation and civic responsiveness of the American people for helping us have a good census. When we begin releasing data from Census 2000, we will begin to give something back to the American people for all they've done. The ACS will continue to give data throughout this coming decade. At the Census Bureau, we like to talk about the census as a "snapshot" of the American people at one point in time; if that is so, then the ACS will be a streaming video of what is going on in our country. It will provide data users who have to make decisions every day, not just every 10 years, the data they need. Mr. Chairman, that concludes my testimony. I will now answer any questions.