Report to Congress
February 1, 1999
U.S. Census Monitoring Board
4700 Silver Hill Road
Suite 1250 - 3
Suitland, MD 20746
For sale by the U.S. Government Printing Office
Superintendent of Documents, Congressional Sales Office, Washington, DC 20402
|U.S. CENSUS |
|4700 Silver Hill Road
Suitland, MD 20746
|Phone: (301) 457-9900 |
Fax: (301) 457-9901
|Gilbert Casellas |
Everett M. Ehrlich
|Phone: (301) 457-5080
Fax: (301) 457-5081
Pursuant to P.L. 105-119, the Presidentially-appointed
Members of the Census Monitoring Board (the "Board") hereby transmit to
Congress our February 1, 1999 report. The report of the
Congressionally-appointed Members is being sent under separate cover.
The Board also has taken testimony and met with numerous outside experts and local community leaders, including members of the National Academy of Sciences ("NAS") panels on census issues, state and local elected officials as well as career civil servants in charge of mapping technology and address list development, representatives of community organizations and civil rights groups; and many others.
|J. Kenneth Blackwell|
A. Mark Neuman
Joe D. Whitley
|As you know, the legislative mandate of the Board is to
"observe and monitor all aspects of the preparation and implementation of
the 2000 decennial census." To date, the Board has held six bipartisan
public hearings; has visited all three Dress Rehearsal sites (Columbia, South
Carolina; Menominee, Wisconsin; and Sacramento, California); has examined the
Census Bureau's data capture headquarters in Jeffersonville, Indiana; has been
briefed by the General Accounting Office team in charge of reviewing the 2000
decennial census and by the Commerce Department's Office of Inspector General;
and has received several extensive briefings by senior managers at the Census
Bureau (the "Bureau") in Suitland, Maryland.
The Board wishes to thank Census Director Ken Prewitt, Atlanta Regional Director and former Acting Census Director Jim Holmes, Associate Director John Thompson and all of the other dedicated career professionals at the Census Bureau for their tremendous cooperation in assisting the Board over the past several months.
Few government endeavors have been subjected to as much scrutiny by so many oversight authorities as the 2000 census. The burden placed on the Bureau must be acknowledged, and the Board wishes to note that throughout our proceedings, the Bureau has never failed to comply with our many requests for written information or in-person briefings.
I. Summary of Findings
Everyone agrees that the 1990 census was a disaster. It was the most expensive census ever and the first in 50 years that produced a larger undercount than the preceding census. More than eight million Americans were missed in 1990; about four million were counted twice.
The resulting net undercount of more than four million was comprised disproportionately of racial and ethnic minorities and children.
In addition, we would note that there is little data on the number of people with disabilities who were missed in the 1990 census, or who were not identified as such. The challenge of fully identifying individuals with disabilities, and collecting census data sufficient to allow government, industry and community service organizations to respond to their needs is a vital one. The Board will examine this issue in future reports and will make recommendations for enhancing the quality of data gathered by the census in this area.
Was the Bureau to blame for the 1990 undercount? Not according to the NAS. The fact is, in a nation of 271 million or so people, 16 percent of whom change their residence each year, it simply is impossible to produce an accurate or complete count using traditional methods of census enumeration.
After 1990, the bipartisan leadership in Washington instructed the Bureau to work with the nonpartisan NAS to develop a more effective methodology for the 2000 census. The goal was to achieve better accuracy at less cost. In the end, the Bureau and NAS agreed on a plan that combined a traditional enumeration of 90 percent of the population with a statistically accurate count of the remaining 10 percent based on sampling of non-respondents during the follow-up phase of the regular head count. The plan also called for the Bureau to conduct a separate nationwide statistical survey based on a sample size of 750,000 households to ensure that the combined head count and sampled results were as accurate and complete as possible in every state.
It is important to note that neither the Bureau nor the NAS ever proposed conducting the decennial census entirely by means of sampling. The original plan for the 2000 census called for the use of scientific estimation only as a supplemental means of finding those people who do not respond to traditional census methods.
The NAS-Bureau plan was endorsed by every relevant scientific organization in the country. But, that did not stop legal challenges to it. On January 25, 1999, the U.S. Supreme Court announced that the legal challenges had some merit. Current law apparently does not permit the Bureau to use statistical methods, even in part, to produce census results for reapportionment of the U.S. House of Representatives. The Court did not deny that the use of statistical "sampling" would produce a more accurate count; it simply ruled that existing law does permit it for the sole purpose of reapportioning the House.
So, where do we go from here?
For starters, whether one agrees with the Court's divided 5-4 decision or not, the Court has spoken and the Bureau's plan to save time and money through the use of "sampling for non-response follow-up" was barred for 2000.
Some would argue that Congress should change the law to allow for the explicit use of statistically adjusted numbers for reapportionment. We would agree, as the primary goal of the census every 10 years should be to produce the most accurate portrait of America -- and the consensus among nonpartisan experts is that statistical adjustment is the only way to produce such a count. But, a statutory change seems unlikely in the near term. Therefore, the burden now is on the Bureau, the Commerce Department and the Congress to reach a consensus on how to conduct the 2000 decennial census in the most successful manner possible, consistent with the Court's ruling.
Fortunately, the Bureau has spent the past year developing an alternative ("Track II") to its original operational plan ("Track I") for the 2000 census. The Track II plan, which incorporates many if not most of the aspects of Track I, was released just two weeks ago on January 15, 1999. At first glance, it is a sophisticated and thorough combination of traditional enumeration efforts, along with traditional follow-up efforts to locate nonrespondents (in place of Sampling for Non-Response Follow-Up, or "SNRFU"), plus several new coverage improvement programs. It includes extensive pre-census consultation with local governments to ensure that the Bureau's Master Address File is as up-to-date as possible before Census Day -- April 1, 2000. It includes a first-ever $100 million professional advertising campaign to encourage greater citizen participation. And, it proposes to open 520 local field offices that will be staffed by as many as 500,000 temporary census workers, making Census 2000 the largest peacetime mobilization in our nation's history.
As you can imagine, the Board has spent much of its time over the past several months analyzing the Bureau's proposed SNRFU plan -- which is now moot. We are now turning our attention, just as the Congress and everyone else is, to the Bureau's traditional follow-up plans and coverage improvement programs under Track II. We intend to examine these aspects closely and to hold a public hearing on it in the near future. Our next scheduled report, due April 1, 1999, will cover the Bureau's Track II preparations thoroughly.
It is worth noting now, however, that the Track II plan will cost substantially more than Track I, and, like Track I, it will enable the Bureau to meet the December 31, 2000 deadline for delivery of reapportionment numbers.
The Track II plan also anticipates conducting a Post-Enumeration Survey ("PES") to evaluate the size and characteristics of the expected undercount. Such a PES was done in 1990 and is considered an essential part of any modern census by the Bureau and others in the statistical, demographic, business and local government communities. Some partisans are already lining up against the PES because they do not want the Bureau to produce a second set of "adjusted" census results that could then be used for redistricting purposes within states.
That is a political issue outside of the Board's purview. We would note, however, that the Supreme Court's January 25, 1999 ruling holds that existing law actually requires the Bureau to produce adjusted numbers "if feasible" for all non-apportionment purposes, including redistricting within the 50 states, the distribution of at least $182 billion in federal funds for each of the next 10 years, and for all other federal statistical reporting requirements (Bureau of Labor Statistics, et al.). As Justice O'Connor wrote,
"In holding that the 1976 amendments did not change the prohibition on the use of sampling in determining the population for apportionment purposes, we do not mean to suggest . . . that the 1976 amendments had no purpose . . . Rather, the amendments served a very important purpose: they changed a provision that permitted the use of sampling for purposes other than apportionment into one that required that sampling be used for such purposes if 'feasible.'"(1)
This issue will now be debated by Congress and the Administration. But, regardless of how or when adjusted numbers are ultimately used, we can report with certainty that a PES is deemed vital to the Bureau's efforts to achieve "maximum possible accuracy at all levels of geography." Indeed, since the Court's ruling, a number of respected outside experts and community leaders have communicated with the Board on precisely this point:
"Regardless of how sampling for non-response is used or not used in Census 2000, a large and well carried out Post Enumeration Survey is critically important for this particular census. A PES is our best tool for assessing the coverage of Census 2000 and, hence, knowing how well we have done." (Dr. John Rolph, University of Southern California, Chair, Committee on National Statistics, National Academy of Sciences)
(1) Department of Commerce v. U.S. House of Representatives, No. 98-404 (U.S. January 25, 1999), at 23.
"The critical need for such a large scale PES is greater in the absence of sampling for non-response follow-up than it would have been in the original Census 2000 sampling plan, because all evidence suggests that a so-called traditional enumeration in 2000 will be demonstrably worse than in 1990, in terms of both omissions and erroneous enumerations." (Dr. Stephen Fienberg, Carnegie Mellon University, former Chair, Committee on National Statistics)
"While it is unfortunate that a narrow majority of the Supreme Court reached the conclusion that current law prohibits the Census Bureau from using sampling for apportionment purposes, it is critical to understand that the Supreme Court did not rule that scientific sampling methods cannot be employed in the 2000 census for other purposes, including the disbursement of billions of dollars of federal funds." (Wade Henderson, Executive Director, Leadership Conference on Civil Rights)
"Since the Supreme Court decision did not rule out using the PES for other purposes such as funding, I think it is really important to do a large scale Post Enumeration Survey to obtain an accurate estimate of the undercount." (Dr. Barbara Bailar, Senior Vice President for Survey Research, National Opinion Research Center, University of Chicago)
"If a PES is not done, the information will never be available regardless of the political or judicial outcome." (Mayor Mike Morrison, Waco, Texas)
"With a burgeoning Asian Pacific American population that is very much one of the populations that have been undercounted in the past, we believe it is essential for this nation, in light of the recent Supreme Court decision, that the Census Bureau conduct a full and accurate post enumeration survey in the year 2000." (George Ong, National President, Organization of Chinese Americans)
"While we are disappointed by the Supreme Court decision, we are encouraged that there is still a window of opportunity for the Census Bureau to be allowed to use sampling in ensuring a fair and accurate count in the Latino and other minority communities who were undercounted in the 1990 Census." (Artura Vargas, Executive Director, National Association of Latino Elected Officials)
"Our hope is that a bipartisan consensus can occur that will see to it that we have an efficient and accurate count in the year 2000. As we know, billions of dollars in resources did not reach communities that gravely needed them in the areas of education, health, law enforcement and civil rights enforcement." (Hilary Shelton, Director of the Washington Bureau, NAACP)
This does not mean Congress and the Bureau should not make every effort to improve coverage in the hardest-to-count areas of the country. We support proposals to hire additional enumerators and to spend additional advertising dollars to target areas with historically high undercount rates.
But, based on all the testimony we've heard from within and outside the Bureau, a traditional census will inevitably result in a huge undercount of the most vulnerable members of our society no matter how much money Congress appropriates to the task. Why? Because the social and demographic trends that produced the undercount in 1990 have increased over the past 10 years (for example, the growth of minority populations and persons with limited English proficiency relative to the entire population).
Consider the just-released dress rehearsal results. Despite finishing on schedule, on budget and on target in terms of public response, the Bureau has determined that the problem of the undercount has gotten worse since 1990, not better. For example, in Sacramento, an undercount of 6.3 percent -- which was much larger than expected -- was found after the PES was conducted and analyzed during the rehearsal. The undercount in Menominee was also significant. (The data in Columbia have not been finalized.)
So, the outcome of the 2000 census -- in light of both the dress rehearsals and the Court decision -- really is not all that uncertain. Assuming Congress provides the necessary funds, which we recommend it should, the career professionals at the Bureau will conduct a sophisticated field operation in 2000. They will utilize the latest data capture technology, the most extensive local promotion and outreach program in census history, and numerous coverage improvement programs to find and count the hardest-to-count segments of the population.
The Bureau obviously is dedicated to producing the most accurate census possible. But, when all is said and done, the Bureau acknowledges that total coverage will be significantly less than 100 percent -- and, at best, only on par with the 1990 results. And, as in 1990, a disproportionate number of those missed in 2000 will be the urban and rural poor, people with disabilities, racial and ethnic minorities and children.
This outcome -- another massive undercount -- should be unacceptable to all Americans.
There has been much talk in recent months about the risk of a "failed" census in 2000. But, now that the Court has resolved the operational question of whether the more efficient sampling techniques can be used to produce numbers for reapportionment, there is no reason for the census to fail. There is plenty of time for the Bureau to refine its plans accordingly.
But, Congress and the Administration still need to decide whether to fund a second set of adjusted and more accurate census results, as current law apparently mandates.
We urge Congress and the Administration to resolve this issue as soon as possible. We happen to agree with the nonpartisan experts that a PES is essential and that an undercount must be avoided at all costs. And, we would propose that the Secretary of Commerce agree in advance to release the adjusted data regardless of the results in order to remove any question as to whether his decision is motivated by partisan politics.
This seems to be the only fair and balanced resolution of this political controversy. Individual states can then decide which numbers to use for redistricting, but the distribution of Federal funds to states and localities, much of which is devoted to programs for the poor, the disabled and others who typically go undercounted in the census, would not be skewed unfairly. An adjusted set of numbers also would allow the business community and others with a vital stake in accurate census data to make decisions based on reality, not deficient census methodology.
But, if Congress does not fund the PES, then "failure" is inevitable. The 2000 Census would result in a record undercount of minorities and children, and it would not be the Census Bureau's fault.
The remainder of this report covers several of the specific areas Congress asked us to monitor, including the Dress Rehearsal results, a discussion of the Master Address File and Local Review, an analysis of the suggested use of Administration Records as an alternative census methodology, and the status of the Bureau's preparations regarding Computer Modernization, Outreach and Promotion, Field Offices and Staff Recruitment.
II. Dress Rehearsal Results
Most census observers agree on the basic problem: How do we eliminate the differential undercount of minorities and children?
By way of background, in 1940 the Census Bureau began to systematically estimate the undercount in the decennial census, and for 50 years the net undercount fell steadily. In 1990, however, things changed. As shown in Exhibit 1, the 1990 decennial census was the first in 50 years to be less accurate than the one before.
Based on the results of the 1990 PES, the Bureau's Committee on Adjustment of Postcensal Estimates ("CAPE") reported a net 1990 undercount of roughly 1.6% -- a bit more than 4 million people.(1) This net figure includes a total of 8.4 million people who were missed, and 4.4 million people who were counted twice.
(1) Assessment of Accuracy of Adjusted Versus Unadjusted 1990 Census Base for Use in Intercensal Estimates, Report of the Committee on Adjustment of Postcensal Estimates (CAPE), August 7, 1992, p. 3.
If the undercount in the Census were evenly distributed across the country and among the nation's racial and ethnic groups, its impact might be largely irrelevant to the distribution of government funds and the allocation of political representation among and within the states. But, the undercount is not evenly distributed. In 1990, the undercount among non-Hispanic Whites was 0.7%, and White homeowners were actually overcounted by .3%.(1) As shown in Exhibit 2, undercount rates for other racial and ethnic groups were much higher.
Second, the 1990 Census produced a substantial undercount of the most vulnerable members of our society, and those most in need of government services: children.2 It is estimated that just over half the people missed in the 1990 census were children(2), with an unfortunate effect on education and other services.
As Sacramento City Councilwoman Lauren Hammond told the Monitoring Board,
"In my whole district there is no middle school. So what does that mean, there are no children that are 13 or 14? Really, it is ridiculous. I have four high schools and eight elementary schools and not one darn middle school because there aren't the numbers to reflect that."(1)
Her sentiments echo those of other community leaders around the country, especially those in state and local government. In a letter to Census Monitoring Board Co-Chair Ken Blackwell, Buck McVeigh, Administrator of the Wyoming Department of Administration and Information's Division of Economic Analysis, wrote:
"Attempts to locate and enumerate everyone have fallen short as prior censuses have shown. People were left out and costs were exorbitant. Sampling will allow the Bureau to make scientific estimates of the population . . ."(2)
Jan Perry, Executive Director of Census 2000 for the city of Los Angeles, wrote:
"(I)t is the City's position that the Bureau's proposal to use the [statistical adjument] process to correct the differential undercount is the best method to ensure a complete count for the City of Los Angeles."(3)
The 1990 census produced a substantial differential undercount between states. The undercount ranged from a high of 3.4 percent in the District of Columbia to a low of .13 percent in Rhode Island.(4) These declines in the quality of the census data came despite the fact that the $2.6 billion 1990 census was the most expensive decennial census ever conducted, 65 percent more expensive than the 1980 census.(5)
In 1997 and 1998, the Bureau carried out three major field tests of its operational plans for the 2000 census. Included in those test plans were a variety of strategies designed to improve operations and to decrease the undercounts expected for 2000.
(1) Testimony of the Hon. Lauren Hammond, Transcript, Hearing of the U.S. Census Monitoring Board, 12/16/98, Sacramento, California, p. 99.
(2) Letter from Buck McVeigh to the Hon. J. Kenneth Blackwell, September 8, 1998.
(3) Letter from Jan Perry and Jessica F. Heinz to the Hon. J. Kenneth Blackwell, December 9, 1998.
(4) Report of the Committee on Adjustment of Postcensal Estimates (CAPE), Attachment 4.
(5) U.S. General Accounting Office, Decennial Census: 1990 Results Show Need for Fundamental Reform, 1992, p.p. 34-35.
Preparations for 2000 intensified last year when the Bureau conducted three full-scale Dress Rehearsals at different locations:
- Sacramento, California -- selected because of its urban setting and the racial and ethnic diversity of its population.
- Columbia, South Carolina -- encompassing the City of Columbia and the surrounding 11-county area. The area was chosen due to its mix of urban and rural areas.
- Menominee, Wisconsin -- conducted in the area surrounding the Menominee Indian Reservation.
At both the Sacramento and Menominee Dress Rehearsal sites, the Bureau was allowed to test the full methodology it had proposed for the 2000 Census, including sampling for non-response follow-up ("SNRFU") and the Integrated Coverage Measurement ("ICM") survey. At the direction of the Congress, the Bureau chose one dress rehearsal site to test a traditional census plan without the use of integrated statistical sampling techniques. Still, this dress rehearsal, which was carried out in the area of Columbia, South Carolina, did include a Post-Enumeration Survey ("PES") like the one used nationwide in 1990.
Based on our observations, it is clear that the regional and local Bureau staff that carried out each dress rehearsal did a very sound job in testing its operational plans, assessing the results and making changes where needed. The final report from the Bureau has not been released yet. However, some preliminary results should be noted:
Schedule and Budget: Activities at all three dress rehearsal sites were completed on-schedule, and with one exception were completed on-budget.
- Columbia: In Columbia, non-response follow-up costs were roughly $2,643,000, 13 percent below budget projections and producing a savings of more than $350,000.
- Sacramento: Non-response follow-up costs in Sacramento were $1,446,000, about $1,300 below the Bureau's original budget estimate.
- Menominee: Menominee non-response follow-up costs proved to be higher than originally projected. Actual total cost was $44,985, which exceeded the Bureau's original budget estimate of $30,754.
Personnel: The ambitious hiring goals the Bureau established for itself in all three dress rehearsal areas proved much easier to meet than had been anticipated. Both the General Accounting Office ("GAO") and the Commerce Department's Office of the Inspector General ("OIG") acknowledged the success of the Bureau's recruiting efforts in accomplishing this goal. In particular, rates of turnover were significantly lower than expected and acceptance rates of individuals offered jobs were much higher. The Bureau had originally projected that 50% of those offered jobs would accept them, and that 100% of employees hired would turn over during the course of dress rehearsal operations. In fact, actual performance during the dress rehearsals far exceeded those projections. One reason for the success in recruitment and retention is that the Bureau instituted a policy of adjusting its wage scales based on local labor market conditions, which we agree is a good idea and should be continued.
- Columbia: 90 percent of candidates receiving job offers accepted them. The turnover rate among Columbia dress rehearsal employees was below 25 percent.
- Sacramento: The job acceptance rate among those offered positions with the dress rehearsal was 78 percent -- 28 percent higher than the Bureau's projections. Turnover among dress rehearsal employees was less than 25 percent.
- Menominee: Seventy-one percent of those offered census jobs accepted them, 21 percent higher than the Bureau estimates. The job turnover rate was 38 percent.
Mail response rates: The Bureau's estimates of the number of households likely to complete and return their census forms proved to be on target, or slightly pessimistic. Mail response rates exceeded Bureau projections in two of the three dress rehearsal sites:
- Columbia: The mailout/mailback response rate for the Columbia dress rehearsal was 54.1 percent, slightly below the Bureau's projection of a 55 percent return rate.
- Sacramento: The Sacramento dress rehearsal's mailout/mailback response rate was 53.7 percent, exceeding the Bureau's original prediction of a 50 percent response to this phase of the operation.
- Menominee: As in Sacramento, actual mailout/mailback response rates exceeded the Bureau's projections. A total of 40.6 percent of households responded during the mailout/mailback phase of the dress rehearsal, slightly higher than the expected 40 percent response rate.
Local Update of Census Addresses ("LUCA"): Participation by local governments in the Bureau's LUCA program produced mixed results. As discussed elsewhere in this report, many local governments felt overwhelmed by the work involved in LUCA. In Columbia, 39 of the 62 two governmental units in the dress rehearsal area signed up to participate in the LUCA process. Thirty-two actually submitted address lists for review. In both the Sacramento and Menominee dress rehearsal areas, only one local governmental unit was involved. Both local governments participated in LUCA during the dress rehearsals.
Most significantly, despite the best efforts of the Bureau and the successful implementation of the dress rehearsal operational plans, the preliminary results released by the Bureau two weeks ago on January 15, 1999 revealed significant undercounts at both the Sacramento and Menominee dress rehearsal sites.
As shown in Exhibit 3, the total undercount in Sacramento was 6.3 percent, and there were significant differentials between the undercounts for the city's racial and ethnic minority populations. African Americans had the highest undercount at 8.7 percent. Hispanic city residents were undercounted by 8.3 percent and 6 percent of Asian Americans were missed by the traditional count. Meanwhile, the undercount rate for non-Hispanic Whites was 4.7 percent, significantly lower.
Similar results were seen in Menominee, which had an overall undercount of 3 percent. American Indians were undercounted by 4.1 percent. (See Exhibit 4)
Recall that the national undercount in 1990 -- which was considered unacceptably large -- was 1.8 percent.
For the Columbia Dress Rehearsal, the Bureau conducted a PES like that used in 1990. However, the results have not yet been finalized.
In sum, there is both good news and bad news from the dress rehearsals. The good news is that the Bureau has proven itself capable of implementing a complicated operational plan on schedule and on budget. The bad news is that the probable undercount in 2000 will be bigger than anticipated, despite the Bureau's best efforts.
As GAO, NAS and other independent bodies have concluded, with a population of 271 million people, 43 million of whom move every 12 months, America is no longer well-served by traditional census techniques. American households receive more mass mailings than ever; people have less free time to fill out forms; many Americans are mistrustful of government queries about their lifestyles; and more and more U.S. residents speak languages other than English.
Moreover, there have been dramatic demographic changes over the past 10 years. For example, the number of traditionally undercounted minorities has grown relative to the population as a whole. As a result, the differential undercount of minorities and children is now projected to be 1.9%, even larger than in 1990. According to this estimate, a net total of 5,218,046 people would be missed by Census 2000, an increase of almost 30% over the size of the 1990 undercount.
However, even if the 1990 undercount were to be repeated, a substantial differential undercount among the states would be seen. Exhibit 5 shows the number of individuals that would be missed state-by-state if the 2000 undercount equals that seen in 1990.
III. Master Address File and Local Review
The Constitutional requirement to carry out a decennial census involves a process of identifying households -- the actual addresses -- and then counting the individuals who live there. Indeed, the primary uses of the census, such as apportioning the House of Representatives, redrawing legislative district lines, distributing Federal funds, and providing reliable data to business for investment and planning, all depend on a census that accurately locates and counts the American people as of April 1, 2000.
Since the decennial census must count people at their residences, a well-constructed list of the nation's residential addresses is indispensable to carrying out the census and to controlling costs. Thus, the Bureau's Master Address File ("MAF") is key. It determines where census forms are mailed or hand delivered, and where census enumerators visit to seek out nonrespondents.
The job of identifying all the residential addresses in the nation is not an easy one, particularly when they must be distinguished from business addresses. Today's single-family home is tomorrow's law office. Residences identified as vacant today may be occupied tomorrow. In addition, the recent economic boom has resulted in substantial residential development and the creation of new addresses that did not exist a short time ago.
With the expectation that the number of households in the United States would grow from 102 million in 1990 to roughly 118 million in the year 2000, several key changes to the Bureau's address list development process were made in preparation for the 2000 Census.
First, the Bureau discontinued the program it used in 1990 to purchase mailing lists from commercial mailing firms and direct marketers. The Bureau had determined that these sources did not provide sufficient address information for minority communities and lower income areas.(1)
Next, two significant changes to the address listing process resulted from Congressional passage of the Census Address List Improvement Act of 1994.(2) The Act established two key efforts by the Bureau to enhance address listing operations:
- For the first time, it authorized a partnership between the Bureau and the U.S. Postal Service for the ongoing exchange of address information. The Bureau continuously updates its MAF from the 1990 census using addresses provided by the Postal Service's Delivery Sequence File (DSF).
(1) U.S. General Accounting Office, Decennial Census: Overview of Historical Census Issues, 1998, p. 27.
(2) P.L. 103-430.
- Second, it also authorized the Bureau to share address information with state and local governments to afford them the opportunity to correct any errors prior to the census. This provision of the Act resulted in the Bureau's creation of the Local Update of Census Addresses ("LUCA") program.
The concept of the LUCA program is a straightforward one, and represents a significant advance over the 1990 Post-Census Local Review ("PCLR") program it replaced. Rather than being limited to sharing block-level housing unit and population counts, LUCA allows the Bureau to share full address lists with participating local governments. Local governments then work to refine the lists, adding new or missing housing units and deleting units demolished or converted to non-residential use.
Local and tribal governments participating in LUCA are given substantially more time to review the data than they were given in the PCLR program in 1990. On November 6, 1998, the Board heard from Shoreh Elhami, Director of the Geographic Imaging System ("GIS") project of Delaware County, Ohio, concerning her participation in the LUCA program.
Ms. Elhami praised the cooperative relationship the Bureau staff in the Detroit office have established with her. She also noted that the LUCA process was far preferable to PCLR, because it allowed more time for local governments to participate:
"What happened in 1990 . . . I happened to be the principal planner for the regional planning commission . . . so I was given 15 days to look at all of those maps and get back with the Bureau. Of course, we didn't have any of these tools available at the time so it was a moot point, we couldn't really do anything . . . What has changed for 2000 . . . we were given the opportunity to participate in LUCA. And the three months that we were given is a lot better than 15 days."(1)
Should a LUCA participant and the Bureau disagree on the disposition of submitted address changes, the participant can request a Bureau review. If the conflict is not resolved, a final appeals process to be conducted by the Office of Management and Budget is available.
(1) Testimony of Shoreh Elhami, Hearing of the U.S. Census Monitoring Board, 11/6/98, Washington, D.C., p.110-111.
While the LUCA program holds great long-term potential, and promises to provide substantially greater benefits than PCLR, it became clear to the Bureau by late 1997 that there were barriers to full implementation of the program. First, some local governments did not have the wherewithal to fully participate in the program.(1) Second, while the Postal Service's DSF is an excellent source of addresses, adapting it for census use as mandated by Congress presented more challenges than were initially anticipated. Regional variations in the speed with which the DSF is updated were identified, and differences in the ways addresses are used by the Bureau caused some incompatibilities.(2) These differences require substantial field verification to resolve.
It became clear during the dress rehearsals that these factors would prevent LUCA from realizing its full potential for developing the most accurate MAF the Bureau needs to conduct the 2000 Census. (However, it should be pointed out that the whole point of conducting the dress "rehearsals" was to identify what is working and what needs to be fixed.) In September of 1997, the Bureau announced a re-engineering of its address listing operation to reinstitute the nationwide address listing operation it used in 1990. LUCA remains as an opportunity for local governments to share information with the Bureau, but an accurate MAF should be generated for non-participating communities with the support of the Bureau's other address listing operations.
Local governments that may not have the resources for full participation in LUCA are being urged to target their workload by concentrating on counts of addresses and housing units, rather than attempting to review every address.
The Board heard extensive testimony on the LUCA program during its field hearings, and the program's operations for the dress rehearsal and for Census 2000 were the subject of extensive discussion during field investigations and staff briefings. The overall impression was that the process holds great potential, but is far from perfect. At the Board's field hearing in Sacramento, the Board heard testimony from Kelly Grieve of the Sacramento Area Council of Governments ("SACOG"). Mr. Grieve had done extensive work with LUCA leading up to the Sacramento Dress Rehearsal and was involved at the time in the LUCA process for Census 2000. While noting that there was a great advantage to seeing the Bureau's address information prior to the census, an opportunity that did not exist in 1980 and 1990, he expressed frustration over a variety of problems that occurred during the dress rehearsal. These included Bureau challenges to address data submitted by SACOG, slow processes for submitting address changes and difficulties with maps provided by the Bureau.(3)
(1) Testimony of Christopher Mihm, General Accounting Office, before the U.S. Senate Committee on Governmental Affairs, July 30, 1998, p. 7.
(2) Preparing for the 2000 Census, Interim Report II of the Panel to Evaluate Alternative Census Methodologies, National Academy Press, 1997, pp. 16-17
(3) Testimony of Kelly R. Grieve, Hearing of the U.S. Census Monitoring Board, 12/16/98, Sacramento California, pp. 88-89, 102.
At the Board's field hearing in Columbia, South Carolina, testimony was heard from Jack Maguire of the State of South Carolina's Office of Research and Statistics ("ORS"). ORS had been approached by several governments in the South Carolina Dress Rehearsal area who felt overwhelmed by LUCA and asked for assistance. In the end, ORS signed agreements with several county and city governments in the Dress Rehearsal area to carry out the LUCA work on their behalf.(1)
Despite these challenges, the benefits of the LUCA process are borne out by the participation rates in the LUCA program for the 2000 Census. The Census Bureau estimates that 18,000 local government jurisdictions will participate in LUCA.(2) Moreover, the LUCA program, which is underway now, allows much more time for local and tribal governments to consult with the Bureau for Census 2000 than was available in 1990.
Finally, while there have been difficulties with the accessibility of LUCA for local governments, it is important to note that LUCA has drawn almost twice the number of participant governments as PCLR did in 1990. See Exhibit 6.
(1) Testimony of Jack Maguire, Hearing of the U.S. Census Monitoring Board, 8/5/98, Columbia, South Carolina, p. 201.
(2) Additional Steps Needed to Improve Local Update of Census Addresses for the 2000 Decennial Census, Final Report of the Office of the Inspector General, U.S. Department of Commerce, September, 1998, p. 7.
Important changes have been made to LUCA as a result of the Bureau's experience in the Dress Rehearsals. Additional staff have been added to provide technical assistance to local governments requesting it. Addresses in the Dress Rehearsals were not placed on maps, or geocoded, which caused some confusion. The addresses are now geocoded for easier review by local governments.
While LUCA remains a valuable opportunity for local governments to improve the census results for their areas, and is a vital part of the Bureau's local partnership efforts, the program highlights key concerns that should be borne in mind as future changes to address listing programs are considered.
As an example, the Bureau's preliminary calculations of the rate at which questionnaires were returned as "undeliverable as addressed," were 12.9% in the South Carolina Dress Rehearsal area and 9.5% in the Sacramento Dress Rehearsal.(1) In the Census Bureau's 1995 tests of the address list in Oakland, California and Paterson, New Jersey, between 4.7% and 7.7% of the addresses were missing from the address file.(2)
Elimination of these errors currently requires extensive independent verification, such as the nationwide address canvassing operations the Bureau has undertaken as part of its re-engineering effort. A long-term program of consultation between local governments, the Postal Service and the Bureau will be required to reduce or eliminate the error rate.
Some have suggested that PCLR should be reinstated. While we certainly support the concept and practice of local review, we believe that the Bureau's current program of "pre-census" local review is more effective, more efficient and more practical than PCLR. For starters, the value of PCLR in 1990 in terms of adding people who were missed to the census count was small in relation to the work and cost required. Out of the 6.5 million census blocks nationwide, only 4.2% were challenged by local and tribal governments during PCLR (a total of 270,650 blocks). The evaluation criteria for a recanvass were met for 62% of the challenged blocks (168,255) and a recanvass operation was conducted.(3) Yet, the recanvass of 168,255 census blocks added only 124,900 people to the count -- less than 1/10th of 1% of the total 1990 population count. PCLR resulted in a net decrease of housing units in the 1990 count: 80,929 housing units were added to the count, but 102,000 housing units were deleted.(4)
As a "quality check" on the census results, as some are proposing, PCLR is also lacking. A far superior quality check would be a fully funded Post-Enumeration Survey, which by definition would be free of any possible geographic or other bias, unlike PCLR.
(1) Letter from Anthony Black, Bureau of the Census, to The Hon. J. Kenneth Blackwell, October 14, 1998.
(2) Testimony of Christopher Mihm, General Accounting Office, before the U.S. Senate Committee on Govenmental Affairs, July 30, 1998, p. 7.
(4) Id., at p. 5.
Finally, not only is LUCA far superior to PCLR, but the fact that it takes place before and during the census -- not after -- means that it will not delay the delivery of final results to the states. Given the Supreme Court's ruling, the traditional census will take several weeks longer to complete because sampled estimates are no longer allowable. PCLR would add at least six weeks to the process, with minimal benefit. Under existing law, the Bureau must deliver reapportionment numbers by December 31, 2000 and adjusted numbers for non-apportionment purposes starting early in 2001. PCLR would throw a very large wrench into that timetable.
However, we do recommend, as several of our witnesses have suggested, that Congress agree to fund grants to state and local governments to allow them to modernize and standardize their own address list and mapping technologies. As it stands today, the nation's 39,000 local governments have widely varying levels of expertise in these areas. Future censuses would be greatly aided if Bureau-friendly data capture systems were developed throughout the country.
IV. Administrative Records
In the midst of the debate over Track I vs. Track II, some census observers proposed a possible Track III -- the use of "administrative records" as an alternative census methodology. Government agencies at the federal, state, and local levels collect extensive personal data on individuals within their jurisdictions. Medicare, Social Security, motor vehicle registrations, gun registrations, driver licenses, voter rolls, Food Stamp and other administrative lists contain valuable information about the American people.
The idea of conducting the census by adding up names on lists is theoretically possible and efficient. But, based on our research and the expert testimony we received, the use of administrative records as a major component of the 2000 operational plan -- beyond what the Bureau already has in store -- is not feasible.
We heard from experts who raised a range of issues which have already been identified, and extensively studied, regarding the many challenges of using administrative records in the decennial census.
First, research suggests that while most of the U.S. population is included on one or more existing administrative databases, no single record system contains all the information the Bureau needs.(1) For example, in addition to identifying individuals, the Bureau must match individuals in the census to an address for purposes of redistricting, and collect race and ethnic data for Voting Rights Act purposes. Even when several federal, state and local databases are combined, there is no guarantee that enough information will be gleaned from them to determine if a person has been missed in the census count.
It is believed that the Internal Revenue Service ("IRS") has the most extensive administrative records system. Roughly 90 percent of the population files returns and another six percent can be identified through "informational documents" like W2's.(2) However, even if we were to merge the information from this system with the Social Security Administration and the Medicare system's records, a huge cross section of people would be missed. And, this cross section includes many of the same people who would be missed in the census.
Second, administrative data bases don't contain all the information the Bureau is required by law to ascertain. For example, IRS files do not contain race and ethnic data at all,(3) and older Social Security files collect this information in ways that no longer match federal data collection standards.
(1) Modernizing the U.S. Census,Report by the Panel on Census Requirements in the Year 2000 and Beyond p. 64.
(2) John L. Czjaka at a Joint Statistical meeting Panel August 1996.
(3) Modernizing the U.S. Census,Report by the Panel on Census Requirements in the Year 2000 and Beyond p. 64.
While the federal government sets standards for the collection of race and ethnicity data in its own programs, there is no universal standard. As an example, there is no single standard to identify certain characteristics such as Hispanic origin. On most records, providing ethnic information is voluntary.
Based on these concerns and his extensive experience in the field, Dr. John Czajka of Mathematica Policy Research, Inc., a member of the National Academy of Sciences, told the Board that:
"The more important question for coverage improvement is whether our principal administrative records systems capture a high proportion of the persons who are actually missed on the Census. This is an area where we flat out need to know more before we can design an effective coverage improvement strategy (using administrative records)."(1)
Some point to the Bureau's experience in 1990 with parole and probation records as a model for using administrative records in 2000. Unfortunately, of the 447,757 people added to the census count through the use of these records in 1990, 57 percent were incorrectly enumerated.
John Thompson, the Bureau's Associate Director for the Decennial Census, told us:
"We concluded [in 1990] that administrative records varied in accuracy and quality, and while they did have the potential for coverage improvement, we could only get that if we did extensive amounts of field verification before we actually put them into the census."(2)
Third, the regularity with which administrative records data bases are updated is a serious and ongoing problem. It has been estimated, for example, that there are more deceased people in administrative records systems than the total number of people who were undercounted in 1990.(3) In other words, the use of administrative records at this point in time would introduce more error into the census than they would remove.
(1) Testimony of John L. Czajka, Hearing of the U.S. Census Monitoring Board, 11/23/1998, Suitland, Maryland, p. 20.
(2) Testimony of John Thompson, Hearing of the U.S. Census Monitoring Board, 11/23/1998, Suitland, Maryland, p. 107.
(3) Testimony of John L. Czajka, Hearing of the U.S. Census Monitoring Board, 11/23/1998, Suitland, Maryland, p. 23.
In addition, an accurate census requires that an individual be identified at his or her residence as of April 1, 2000. Few administrative record systems keep address information this current. For example, the majority of tax forms are updated once a year, motor vehicle records are updated even less frequently, and once a social security number has been issued there is no need to update that record again until the individual has retired or otherwise begins receiving benefits. As a result, people who move to another state but fail to change their driver's license could be improperly counted in a census if these records were among those used to enumerate the population.
Fourth, there is the problem of "matching" the administrative files to the actual census results. This is a difficult, uncertain and expensive process. As shown in Exhibit 7, evidence from the Bureau's 1995/1996 Administrative Records test shows that there is a low rate of matches between whole census households and whole administrative record households. Consider this example: a person might list his formal name, "Robert Smith," on the census questionnaire. But, the same person's voter registration could be listed as "Bobby Smith."
Fifth, there are some European countries that use administrative records to conduct their censuses, including Denmark and Sweden. However, unlike the United States, these countries have a "national ID" system. Without such an individual identifier, the task of matching would be very difficult. Does Congress want to establish a national ID for all Americans?
It also is interesting to note that many of the nations which rely on administrative record censuses also supplement their national data with statistical sampling to correct for errors in their administrative record systems.(1)
Sixth, another important question for the use of administrative records is one of verification. Previous tests have shown that, when the Bureau was unable to match its census data to an administrative record placing an individual at a specific address, it was the census data that proved accurate far more often (see Exhibit 8). Because of these inconsistencies in the files, the Bureau feels that verification would have to come from two independent sources before an individual could be confidently added to the count.(2)
(1) Modernizing the U.S. Census, Report by the Panel on Census Requirements in the Year 2000 and Beyond, National Academy Press 1994, p. 63.
(2) Dr. Cynthia Clark, Census Bureau Briefing to the Census Monitoring Board, 10/28/98.
Seventh, traditional American notions of privacy and a preference for limited government intrusion have erected a number of barriers to the widespread sharing of information on individuals between government agencies. Clearly, a comprehensive system of using administrative records for census purposes would require substantial legislative action. As Dr. Czajka told us:
"I think it very likely that a significantly broader use of administrative records for census purposes may not be possible without legislation that recognizes the legitimacy of such uses and provides both guidelines and safeguards."(1)
Despite the many practical obstacles, the Bureau is not opposed to using administrative records where feasible. In fact, the Bureau already uses a number of records to help facilitate the decennial census, primarily in the area of address list development.
As described above, the MAF is built from the Postal Service's Delivery Sequence File ("DSF".) The DSF is updated as letter carriers identify addresses for new delivery points or changes in the status of existing addresses. Even so, there are problems. Updates are delayed. Not all carrier routes provide the same quality information. Rural addresses contain imprecise geographic descriptions and some addresses do not appear on the list at all.
In addition, the Bureau uses files from the Department of Health and Human Services, the Department of Justice, the Veterans Administration and the Department of Energy to compile lists of group quarters or institutions. Files from the Department of Defense are used for enumeration of the military population, and files from colleges and universities are used to supplement the enumeration of their student populations.
The Congressional Members of the Monitoring Board have proposed targeting America's major urban areas -- where most of the undercounted population lives -- and using Medicaid and other records to enumerate individuals, especially children, who might otherwise go uncounted in the 2000 census. The concept is admirable. If it can be done, we're all for it. The problem is, the Bureau and all the outside experts we consulted agree that it cannot be done in time for 2000. The records are not current, many of the children who probably will be missed in a traditional census have not even been born yet, and there is not enough time between now and April 1, 2000 to design -- much less implement -- a system for pursuing this project without introducing substantial flaws into the census count itself. There also is the problem of asking Congress to appropriate funds to address the problem of the undercount in some areas of the country but not all.
(1) Testimony of John Czajka, Hearing of the U.S. Census Monitoring Board, 11/23/1998, Suitland, Maryland, p. 36.
But, this does not mean that we are opposed to greater use of administrative records. In particular, we support extensive research in 1999 and 2000 that will lead to much greater if not complete reliance on administrative records during the 2010 census. We urge the Bureau to follow through on its plans to conduct several tests of administrative records data bases in the upcoming census. These plans include:
- The Bureau will assemble a number of national level administrative record files. These files will be unduplicated using Social Security numbers and assigned block level geographic codes using the Master Address File/Topologically Geographic Encoding and Referencing system (MAF/TIGER). No attempt will be made to assign records to individual housing units.
- The Bureau will also assemble a number of administrative records sets that will be unduplicated using the SSN and addresses, which will then be matched to a complete list of possible residential addresses on the Master Address File. These locations will include about one million housing units and a population of about 3 million. About 50 percent of the housing units will be selected based on criteria associated with a high likelihood that a person can be captured in an administrative record census. The remaining units will be selected from those less likely to be included in an administrative records capture.
- The Bureau also will evaluate the use of administrative records for non-response follow-up activities. The study will assess the feasibility of using administrative records in place of field interviews to obtain data on nonresponding households. The project will be conducted in four phases. Phase one conducts a sample of nonresponding households in the AREX2000 sample will be selected for enumeration. During phase two, a computer match will be made between the census and the administrative records database. During the third phase, information obtained from administrative records will be reconciled with information obtained from the AREX2000 research project. During the fourth phase, a comparison and analysis of the computer match and subsequent reconciliation will be conducted.
- The Bureau also will assess the potential of administrative records systems to reduce the cost and workload associated with maintaining the Master Address File. The Bureau plans to acquire and process national files from the IRS Individual Master File (Form 1040), the IRS Information Return Master File, the Health Care Finance Administration's Medicare enrollment database, HUD's TRACS file, the Selective Service's registrants file, the Indian Health Service file and the National Telephone Data Base or Standard Statistical Establishment List. These files would be combined to produce a single administrative records data base that would be used to update the MAF in selected areas prior to the recanvass operation.
Research into the potential uses of administrative records and the census has been carried out by a wide range of expert bodies and organizations, including the 2000 Census Advisory Committee, the Census Advisory Committees on Race and Ethnicity, the Census Advisory Committee of Professional Associations, the Association of Public Data Users, the Working Group on Ancestry on the U.S. Census and the Council of Professional Associations on Federal Statistics.
In particular, the NAS convened two panels which addressed this issue. Panelists agree that while administrative records could play a significant role in decennial census activities in the areas of content evaluation, content improvement, and content collection, the use of scientific estimation will be necessary to make the process manageable and cost effective.(1)
(1) The Bicentennial Census: New Directions for Methodology in 1990, Report by the Panel on Decennial Methodology, National Academy Press 1985, p. 264.
Although the NAS panels have endorsed the concept of using administrative records to support census activities, they concluded that:
"An administrative records census is not a feasible option for 2000," but the Bureau "should intensify its efforts to use administrative records, recognizing that the expanded use of administrative records is an evolutionary process."(1)
GAO also has determined that there is no reason to believe that it is viable for the 2000 Census to attempt to rely fully or even primarily on administrative records. According to the GAO, the technical obstacles facing an administrative records census are too great and the amount of research that needs to be completed is too significant to meet challenges in time for the 2000 Census.(2)
However, as many experts have noted, the potential use of administrative records in the 2010 census is quite high. As the Board continues its investigations, we intend to consult with experts both outside and within the Bureau to identify a list of specific policy, technological and other reforms that would be needed to conduct an administrative records census in the year 2010.
(1) Modernizing the U.S. Census, Report by the Panel on Census Requiremets in the Year 2000 and Beyond, National Academy Press 1984, p. 68.
(2) U.S. General Accounting Office, Census Reform: Major Expansion in the Use of Administrative Records for 2000 is Doubtful, 1992, GAO-GGD PP. 92-54.
V. Computer Modernization
The Census Bureau has awarded several key contracts for the modernization of its computer and office automation systems in preparation for the 2000 Census. Some of these major contracts are:
Data Access and Dissemination (DADS). To design a new, web-based system for easy access to data generated by the Census Bureau, including results from the 2000 Decennial Census. The contract has been awarded to IBM.
Data Capture System 2000 (DCS 2000). To provide imaging equipment, data keying equipment and processing systems for the Data Capture Centers ("DCC") to be used in the 2000 Decennial Census. The contract was awarded to Lockheed Martin.
Data Capture Services. To provide all DCC functions not included in the DCS 2000 contract. TRW was awarded this contract, which includes DCC facilities management, office automation services, staff, office equipment and supplies.
Telephone Questionnaire Assistance (TQA). To provide a fully integrated nationwide network of call centers capable of supporting high-volume, limited duration inbound Telephone Questionnaire Assistance calls from the general public during the 2000 Census. In addition, the system provided will support outbound follow-up calls to residential telephone numbers furnished by the Census Bureau. EDS was awarded the TQA contract.
|Data Access and Dissemination||IBM||$26,000,000||4/97|
|Data Capture System||Lockheed||$73,000,000||3/97|
|Data Capture Services||TRW||$87,000,000||1/98|
|Telephone Questionnaire Assistance||EDS||$100,000,000||12/98|
Source: Bureau Briefing for the Board, "Census 2000 Issues and Updates," Nov. 6, 1998.
On October 23, 1998, the Board visited the Bureau's Data Capture Center in Jeffersonville, Indiana. The 2000 Decennial Census will mark the first time in which the Bureau is outsourcing the management of most of its data centers.
The Jeffersonville DCC will carry approximately 25 percent of the overall data processing workload for the decennial, but will narrow its focus to special cases and will be exclusively responsible for the nationwide adjustment survey. It also will process the address listing and code the questionnaires before they are sent out.
The other three DCCs (located in Phoenix, Arizona; Pomona, California; and Baltimore, Maryland) will be responsible for check-in mail returns, questionnaire edits, data capture and payroll and personnel administration systems management.
TRW is providing management services for the DCCs. As the National Processing Center, Jeffersonville's management will remain under the Bureau's command but will follow TRW's management plan for the other DCCs.
Important questions concerning DCC operations and contractor performance have been raised in recent months, and we will turn to a full examination of these issues in a future report.
VI. Outreach and Promotion
The Bureau's outreach, partnership and promotional efforts are critical components to increase awareness about Census 2000, to increase mail response rates and to reduce the undercount.
The Bureau will ultimately hire approximately 320 partnership specialists to coordinate local partnership and outreach during the 2000 census. Partnership specialists have a wide range of tasks, including responding to media inquiries, assisting in the recruitment of enumerators, obtaining space for recruiting and training census workers.
The four major components to the Bureau's community outreach and partnership plan are: 1) Complete Count Committees; 2) Census in the Schools Project; 3) paid advertising; and 4) Advisory Committees.
Local government officials establish the Complete Count Committees (CCC) and may include representatives from, but not limited to, community, business, media, civic, religious and educational institutions and organizations. These committees help to develop a local outreach strategy to encourage public participation in the census, particularly for hard-to-enumerate communities. CCCs provide the Bureau with local or regional knowledge, expertise and insight into reaching undercounted communities.
While CCCs are a critical component to the success of Census 2000, the Bureau should be cautious in relying too heavily upon them for outreach efforts. The committee's resources may vary for each city, town or county. Committee members are not Bureau employees; they are unpaid local consultants who may become involved as volunteers or local appointees.
The Census in the Schools project aims to increase response rates and facilitate awareness of Census 2000 by targeting students enrolled in school. The Bureau hopes that students will be effective conduits and encourage their parents or guardians to complete census questionnaires. The project includes components for students in kindergarten through grade 12 and includes components for use in classes for adult English as a Second Language, citizenship and high school equivalency degrees. The Bureau awarded a contract to Scholastic, Inc., to develop the curricula.
In 1990, the Bureau relied solely on free public service advertising to get its message out through television, radio and the print media. For the first time, the Bureau has opted to use a paid advertising campaign to encourage participation in 2000. The campaign is aimed at increasing mail return from the public at-large, targeted groups and populations that were disproportionately missed in previous censuses.
The Bureau awarded a $100 million contract to Young & Rubicam, Inc., ("Y&R"), along with four partnering advertising agencies. Y&R has developed a layered media strategy designed to increase awareness and mail response rates and to support partnership and outreach efforts to all segments of the population. The focus of the advertising campaign is on those least likely to respond to the Census Bureau's initial mailing of census forms.
Members among both the Presidential and Congressional appointees to the Monitoring Board have expressed interest in the details of the Y&R contract and the company's plans to accomplish the ambitious goals required of the Census 2000 advertising program. We will turn to an examination of this program in upcoming hearings before the design of the advertising campaign is finalized, and will address the advertising plan in subsequent reports.
VII. Field Offices and Staff Recruitment
In order to support the full range of Census 2000 activities, the Bureau will open a nationwide network of Local Census Offices ("LCOs").
The Bureau's original operational plan for the 2000 census called for the opening of 476 LCOs. However, as a result of last week's Court's decision, the Bureau will be required to hire and supervise a substantially larger number of enumerators and support personnel. These increased demands will require the Bureau to open roughly 520 LCOs.(1)
In response to the uncertainty over the methodology to be used in Census 2000, Congress required the Bureau to prepare to carry out the census whichever way the issue was resolved. This "dual track" planning requirement had raised the specter of a substantial number of LCOs being opened, and then closed as unnecessary when a final decision on methodology was made.
In order to avoid that outcome, the Bureau compared LCO placement in both plans and forced as much overlap as possible. As a result, 400 LCOs were made common to both plans. By focusing on opening these offices first, the Bureau was able to delay the opening of offices unique to one plan or the other until March 1999. The arrival of the Court's decision prior to that time will allow the Bureau to avoid opening any unnecessary LCOs.
To date, the Bureau has signed leases for 130 LCOs. Locations for all 520 LCOs have been identified through the General Services Administration ("GSA").
In its original plan for the 2000 census, the Bureau calculated that more than 285,000 census takers and support personnel would be needed in order to enumerate the anticipated 118 million housing units and 275 million people in the United States.(2) In mid-1998, the Bureau estimated that filling those positions would require evaluating over 2.7 million applicants.(3)
Last week's Court decision will require the hiring of a substantially larger enumerator workforce at a substantial increase in cost.
Performance of the Bureau's personnel systems to date, as tested during the dress rehearsals, received generally favorable reviews. The OIG applauded the Bureau's personnel successes during the Dress Rehearsals to recruit, screen and hire competent staff and to meet their hiring goals. However, some minor procedural problems were identified, which are now being reviewed and will be discussed in future reports.(4)
(1) John Thompson, briefing to Census Monitoring Board.
(2) "How America Knows What America Needs!" Bureau of the Census Fact Sheet, D-3240 June 1998.
(3) "Census 2000 in a Flash" Bureau of the Census Fact Sheet, D-3237, June1998.
(4) Office of the Inspector General, "Bureau of the Census, Sacramento Dress Rehearsal Identifies Needed Improvements in Personnel Administration," September 1998.
VIII. Freedom from Political Bias
Congress asked the Board to determine if politics has played an undue role in the decision-making process at the Bureau. We have examined this issue closely and raised it with several of the career civil servants who testified at our hearings.
At no point have we detected any public or private concern among Bureau personnel about any inappropriate political interference on the part of the Administration or Congress. To the contrary, we are confident that the Bureau's senior managers have been and are guided only by the desire to fulfill their legislative mandate to produce the most accurate count possible. It should also be noted that only three of the Bureau's 9,500 permanent employees nationwide are political appointees of the Clinton Administration.
We would note that our view on this matter is shared by Christopher Mihm of GAO and by President Bush's Census Director, Dr. Barbara Bryant, both of whom testified on the issue last year:
"I must say, though, that based on the knowledge that I have now stretching back over a decade with the professionals working in senior levels within the Census Bureau, my impression, very strong impression, is that if any of them even got the whiff that there was political manipulation, The Whistleblower Protection Act would kick into overdrive." (Mr. Mihm) (1)
"The Plan for Census 2000 is not a scheme conceived by the present Administration. Rather it is an experienced driven plan based on evaluation of the 1990 census to improve on the methods used then. A research driven plan that in mid-decade combined two streams of research: First five years of research by Census Bureau professionals and second, recommendations of a Congressionally-mandated study by experts assembled by the National Academy of Sciences. Research began in 1991." (Dr. Bryant) (2)
While the Bureau's decision making process for Census 2000 has been free from political bias, we must mention the degree of political and media scrutiny that the Bureau has been subjected to in its planning for this decennial.
(1) Testimony by Christopher Mihm before the Senate Committee on Governmental Affairs July 30, 1998.
(2) Testimony by Dr. Barbara Bryant before the Subcommittee on the Census Committee on Government Reform and Oversight (U.S. House of Representatives) September 17, 1998.
The 2000 Census has been the subject of great deal of oversight activity: eight audits by OIG, three audits by GAO, six hearings by the Monitoring Board, and extensive requests for testimony and documents from the Bureau's oversight committees in the Congress. Each of these entities has generated numerous requests for materials from, meetings with and briefings by Bureau management and staff.
The Bureau's ability to maintain its objectivity and professionalism in the midst of this extraordinary level of scrutiny and questioning is an achievement for which it should be recognized.
In sum, the major findings and recommendations of the Presidential Members of the Board as of Feburary 1, 1999 are:
More than anything else, we urge Congress and the Administration to resolve the remaining operational issues as soon as possible. Census Day is exactly 14 months away. That's not much time. The nonpartisan professionals at the Bureau need definitive direction now.
Co-Chair, Presidential Members
DR. EVERETT EHRLICH
cc: The Honorable John Podesta, Chief of Staff, The White House
The Honorable William Daley, Secretary of Commerce
Members of the 106th Congress
Questions or comments regarding this
service? Contact the GPO Access User Support Team
by Internet e-mail at firstname.lastname@example.org;
by telephone at 1-202-512-1530 or toll free at 1-888-293-6498;
by fax at 1-202-512-1262.