1
Introduction

In forming a new nation after the American Revolution, the U.S. Constitution mandated the conduct of a census of population to permit peaceful reallocation of political power among the states every 10 years. From the first census in 1790, which counted 3.9 million people, the content and operation of the census, and often the use of the results, have been a source of debate (see Anderson, 1988; Anderson and Fienberg, 2001b). The 2000 census, which counted 281.4 million people, is no exception.

The dominant point of contention regarding the 2000 census has been which set of numbers to use as the official population counts for states or small areas: the totals from the census enumeration process or the results of a procedure that uses sample-based statistical techniques to adjust those totals for the estimated net undercount. Throughout the 1990s, the issue of sample-based adjustment was debated by researchers and policy makers and contested in courtrooms (see Brown et al., 1999; Freedman, 1991; Freedman and Navidi, 1992). Those debates shaped the planning of the 2000 census and set the stage for crucial decisions on the use of adjusted census data in three major applications.

The first of these major application decisions was whether sampling could be used to adjust the state counts used to fulfill the primary constitutional mandate of the census: the reapportionment of the U.S. House of Representatives. That decision was made by the U.S. Supreme Court in January 1999, when the court ruled that existing law (in Title 13 of the U.S. Code) prohibits the use of sampling in producing population counts for reapportionment. Accordingly, the U.S. Census Bureau revised its plan for the 2000 census—which had included the possibility of that use—and the apportionment totals delivered to the president on December 28, 2000, were derived from the basic enumeration.1

The Supreme Court’s ruling did not preclude use of sampling to adjust census data for other purposes; consequently, the Bureau carried out its planned

1  

The population counts for purposes of congressional reapportionment include residents of the 50 states and the District of Columbia as of April 1, 2000, comprising citizens, legal aliens (except for visitors or staff of foreign embassies), and undocumented aliens. U.S. merchant personnel on ships and federal and civilian employees overseas (who are assigned to their state of residence) are included in reapportionment counts but not in census data for other uses.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 23
The 2000 Census: Interim Assessment 1 Introduction In forming a new nation after the American Revolution, the U.S. Constitution mandated the conduct of a census of population to permit peaceful reallocation of political power among the states every 10 years. From the first census in 1790, which counted 3.9 million people, the content and operation of the census, and often the use of the results, have been a source of debate (see Anderson, 1988; Anderson and Fienberg, 2001b). The 2000 census, which counted 281.4 million people, is no exception. The dominant point of contention regarding the 2000 census has been which set of numbers to use as the official population counts for states or small areas: the totals from the census enumeration process or the results of a procedure that uses sample-based statistical techniques to adjust those totals for the estimated net undercount. Throughout the 1990s, the issue of sample-based adjustment was debated by researchers and policy makers and contested in courtrooms (see Brown et al., 1999; Freedman, 1991; Freedman and Navidi, 1992). Those debates shaped the planning of the 2000 census and set the stage for crucial decisions on the use of adjusted census data in three major applications. The first of these major application decisions was whether sampling could be used to adjust the state counts used to fulfill the primary constitutional mandate of the census: the reapportionment of the U.S. House of Representatives. That decision was made by the U.S. Supreme Court in January 1999, when the court ruled that existing law (in Title 13 of the U.S. Code) prohibits the use of sampling in producing population counts for reapportionment. Accordingly, the U.S. Census Bureau revised its plan for the 2000 census—which had included the possibility of that use—and the apportionment totals delivered to the president on December 28, 2000, were derived from the basic enumeration.1 The Supreme Court’s ruling did not preclude use of sampling to adjust census data for other purposes; consequently, the Bureau carried out its planned 1   The population counts for purposes of congressional reapportionment include residents of the 50 states and the District of Columbia as of April 1, 2000, comprising citizens, legal aliens (except for visitors or staff of foreign embassies), and undocumented aliens. U.S. merchant personnel on ships and federal and civilian employees overseas (who are assigned to their state of residence) are included in reapportionment counts but not in census data for other uses.

OCR for page 23
The 2000 Census: Interim Assessment Accuracy and Coverage Evaluation (A.C.E.) Program in summer 2000. The A.C.E. matched the results of a survey of people residing in a set of randomly selected blocks to the set of census enumerations from those blocks and used the statistical method of dual-systems estimation to produce population estimates that could be used to adjust the census data.2 Redistricting—the redrawing of boundaries of congressional and other legislative districts to reflect population shifts—was the focus of the second major application decision regarding sample-based adjustment. The Census Bureau faced a legally mandated deadline of April 1, 2001, to release population counts at the block level for redistricting purposes. On March 1, the Census Bureau recommended to the U.S. Secretary of Commerce that unadjusted counts from the enumeration process should be the official data for redistricting; the secretary adopted the Bureau’s recommendation on March 6. The Bureau’s recommendation not to adjust the block-level counts was driven by the lack of time to resolve its concerns over the accuracy of disparate population estimates derived from the census, the A.C.E., and demographic analysis (a technique that constructs a national population estimate based on birth, death, and Medicare records, and estimates of net immigration). The third major application decision on adjustment of the 2000 census is expected to occur on or about October 15, 2001, when the Bureau issues its recommendation on whether to adjust census data that are used to allocate state and federal funds and for other purposes. The Bureau has been conducting additional evaluations since March of the census, the A.C.E., and demographic analysis. It is at this juncture, between two critical decision points in the release of 2000 census data, that the Panel to Review the 2000 Census offers preliminary assessments of the census and A.C.E. processes. CHARGE TO THE PANEL Given the importance of the many uses of census data and the need to have an independent assessment of the quality of the 2000 census operations and results, the Census Bureau in 1998 asked the Committee on National Statistics (CNSTAT) of the National Research Council to convene a Panel to Review the 2000 Census. The panel has a broad charge to review the statistical methods of the 2000 census, particularly the use of the A.C.E. and dual-systems estimation, and other census procedures that may affect the completeness and quality of the data. A sister CNSTAT Panel on Research on Future Census Methods was convened in 1999 to begin consideration of the Census Bureau’s planning process for the 2010 census (see National Research Council, 2000a). The panel has conducted several activities to date to carry out its charge. The panel held three open workshops on topics related to the A.C.E. and the 2   See the Glossary for definitions of such technical terms as dual-systems estimation.

OCR for page 23
The 2000 Census: Interim Assessment possible adjustment of census counts, in October 1999, February 2000, and October 2000 (National Research Council, 2001a, 2001b, 2001c). Individual panel members and staff also discussed aspects of the census, the A.C.E., and demographic analysis with other experts in the field. Panel members and staff made observation visits in 1999 and 2000 to selected census data capture centers, regional census offices, regional A.C.E. offices, and local census offices around the country. (The panel chair and staff previously observed operations in the 1998 Columbia, South Carolina, and Sacramento, California, dress rehearsal sites.) The purpose of the trips was to familiarize the panel with census and A.C.E. operations. Because of the importance of the Master Address File (MAF) to a complete census and because of new procedures used for input to the 2000 MAF by localities, the panel commissioned a review of the 2000 Local Update of Census Addresses (LUCA) Program by a working group of six representatives of governments that participated in the program. The group conducted a survey of 101 governments that participated in LUCA, completed over a dozen in-depth case studies of LUCA participation, and analyzed data on LUCA participation and the MAF provided by the Census Bureau (LUCA Working Group, 2001). Panel members and staff reviewed the extensive documentation and evaluation results made available by the Census Bureau in support of its March 1, 2001, recommendation not to adjust the data for redistricting.3 The panel also conducted extensive analyses of microdata that the Bureau made available to the panel, the 2000 Census Monitoring Board, and the U.S. House of Representatives Subcommittee on the Census (see “Census Oversight,” below). These data not only informed the panel about the operations of the A.C.E., but also provided a window (from the census enumerations in the A.C.E. sample blocks) into operational aspects of the census itself. This is the panel’s third report. A letter report, released in May 1999, commented on aspects of the proposed sample and post-stratification design for the A.C.E. (National Research Council, 1999a). A second letter report, released in November 2000, commented on the process and evaluations that the Census Bureau planned to follow for its March 2001 decision on whether to provide adjusted or unadjusted data for legislative redistricting (National Research Council, 2000b). We expect to issue a third letter report after the Bureau makes a recommendation on adjustment of census data for such purposes as fund allocation (see Part I). The panel is charged to issue a final report in September 2002. 3   These documents, archived at http://www.census.gov/dmd/www/EscapRep.html, include a report from the Census Bureau’s Executive Steering Committee for A.C.E. Policy (2001a) and 19 memoranda in the “B” series, many of which are cited in this report.

OCR for page 23
The 2000 Census: Interim Assessment CENSUS OVERSIGHT: A BRIEF REVIEW Our panel is not the only group charged with reviewing the 2000 census. Indeed, this census has been conducted in an atmosphere of unprecedented attention and oversight, not only during the actual enumeration, but also throughout the decade leading up to 2000. 1991–1996 Several oversight mechanisms were established early in the decade for reviewing 2000 census plans in response to concerns about the 1990 census: its high cost and yet its failure to reduce the net undercount or to narrow the difference between net undercount rates for minorities and the white population from what was measured in the 1980 census. The net undercount of the population, measured by demographic analysis, increased from 1.2 percent (2.8 million people) in 1980 to 1.9 percent (4.7 million people) in 1990. Net undercount rates for blacks increased from 4.5 percent in 1980 to 5.7 percent in 1990; net undercount rates for nonblacks increased from 0.8 percent in 1980 to 1.3 percent in 1990 (National Research Council, 1995:Table 2–1). The Secretary of Commerce in late 1991 established a 2000 Census Advisory Committee, consisting of over 30 representatives from a wide range of associations representing business, labor, minority groups, data users, scientific professions, state and local governments, and others. This committee met several times a year over the decade (and continues to meet on planning for the 2010 census). Also meeting regularly during the 1990s on 2000 census issues were the Bureau’s long-established Advisory Committee of Professional Associations and advisory committees for minority groups (Citro, 2000a). Two Committee on National Statistics panels were established in 1992 to address 2000 census planning. One panel was convened at the behest of Congress to consider data requirements and alternative designs for 2000 (National Research Council, 1995); the other panel was convened at the Census Bureau’s request to consider detailed methodology (National Research Council, 1994). Subsequently, a third CNSTAT panel was organized in 1996 to comment periodically on the Bureau’s maturing plans for 2000; this panel issued its final report in 1999 (National Research Council, 1999b). The Bureau reviewed the results of its research and testing with all of these groups and others, such as the U.S. General Accounting Office and the Department of Commerce Inspector General’s Office. The Bureau’s research covered a wide array of topics, including alternative methods to improve the response rate of households to mail questionnaires, simplify the short and long forms, and measure coverage of the population. By February 1996 the Bureau had decided on a plan for the 2000 census that maximized the use of statistical sample-based techniques to conduct the

OCR for page 23
The 2000 Census: Interim Assessment enumeration and adjust the population data for measured net undercount. To reduce the costs and time needed to complete the enumeration and the adjustment, the Bureau intended to follow up households that did not mail back a questionnaire on a sample basis instead of revisiting every nonresponding household. (This procedure was termed sampling for nonresponse follow-up, or SNRFU.) To be able to produce one set of adjusted population counts by the December 31, 2000, deadline for providing state population totals for congressional reapportionment, the Bureau planned an Integrated Coverage Measurement (ICM) Program. Under this “one-number census” plan, the Bureau would conduct a large survey of 750,000 households, match the survey responses with census records, and have the results ready to produce adjusted estimates by the mandated deadline. The plan using SNRFU and ICM was first tested in the field in the 1995 test census conducted at three sites (Wright, 2000). 1997–2001 Views on the merits of the Bureau’s 2000 census plan in Congress and the executive branch divided along partisan lines: the Clinton Administration and House Democrats generally supported the use of sampling while House Republicans generally opposed its use (see McMillen, 1998). Conflict between the two sides resulted in a delay of funds with which to conduct the 1998 dress rehearsal. In fall 1997 compromise legislation stipulated that unadjusted data would be released from the 2000 census in addition to any adjusted data. It also provided for expedited judicial review of the legality of the use of sampling for the census and established a 2000 Census Monitoring Board to consist of four members appointed by House and Senate Republican leaders and four members appointed by President Clinton in consultation with House and Senate Democratic leaders. The Monitoring Board was proposed as a way to address the concerns of some in Congress that the administration might manipulate the census data for political gain.4 With this mechanism in place, there was agreement that the dress rehearsal and other necessary census planning activities would be fully funded and that the dress rehearsal in one of the planned sites would be conducted without the use of either SNRFU or ICM. The House of Representatives also established a Subcommittee on the Census in the Committee on Government Reform and Oversight, which began census oversight activities in early 1998.5 4   The congressional and presidential appointees each have their own budgets and staffs and are charged to report periodically to Congress through September 2001 (see U.S. Census Monitoring Board Presidential Members, 2001a, 2001b; U.S. Census Monitoring Board Congressional Members, 2001). 5   The subcommittee is slated to go out of existence at the end of 2001.

OCR for page 23
The 2000 Census: Interim Assessment The Census Bureau consequently adjusted its planning to allow for two possible census designs—one based on SNRFU and ICM, the other implementing follow-up on a 100 percent basis and using a postenumeration survey to evaluate coverage. The Bureau implemented each design in a dress rehearsal in spring 1998: the rehearsal in Sacramento, California, used SNRFU and ICM; the one in Columbia, South Carolina, and surrounding counties, used traditional census methods with a postenumeration survey. A third dress rehearsal in Menominee County, Wisconsin (which includes the Menominee Indian Reservation), was a hybrid using 100 percent nonresponse follow-up and ICM. Following the January 1999 Supreme Court decision that existing law did not permit sampling for congressional reapportionment but did not rule out its use for other purposes, the Census Bureau again revised its plans. The final design for the 2000 enumeration was announced by Director Kenneth Prewitt at a press conference in February 1999, little more than a year before Census Day on April 1, 2000. That design was to follow up nonresponding households on a 100 percent basis and to deliver the enumerated counts—unadjusted for coverage errors—to the President by December 31, 2000, for use in reapportioning congressional seats. The design also included the Accuracy and Coverage Evaluation (A.C.E.) Program, in which 300,000 households would be interviewed in about 11,000 block clusters and their records matched with census records in those block clusters. Adjusted estimates, constructed by applying dual-systems estimation to the matched data, would be developed on a schedule that could permit releasing adjusted block data to all states by the statutory deadline of April 1,2001, for use in congressional redistricting. After intense debate, Congress approved this design and provided the full amount of funding requested by the Census Bureau. The amount—over $7 billion—exceeded that spent in any previous census even after adjustment for inflation and the increased U.S. population. The 2000 census design was carried out largely as planned; spending totaled about $300 million less than the amount budgeted. AVAILABILITY OF INFORMATION Throughout the 2000 census planning process and particularly in the period 1998–2001, the Census Bureau has shared its plans, research results, and other information with the unusually large array of oversight groups described above and with the broader scientific and stakeholder communities. The Bureau provided extensive documentation of the design and operational procedures planned for the A.C.E. at the first two open workshops sponsored by our panel (see “Charge to the Panel,” above). At a third open workshop of our panel, the Bureau provided a full explication of the decision process and

OCR for page 23
The 2000 Census: Interim Assessment kinds of evaluations it planned to complete by March 2001. These evaluations were needed for the Executive Steering Committee for A.C.E. Policy (ESCAP), comprised of senior Bureau staff, to recommend to the director—and the director in turn to recommend to the Secretary of Commerce—whether to release adjusted or unadjusted block data for redistricting.6 The Census Bureau also facilitated arrangements for staff and members of its oversight groups—including our panel, our sister Panel on Research on Future Census Methods, the full 2000 Census Monitoring Board, Republican and Democratic staffs of the House subcommittee, and the 2000 Census Advisory Committee—to observe census operations in the field during spring-summer 2000 at locations around the country. In addition, the Census Bureau director testified frequently before the House subcommittee throughout the conduct of the census, and the Bureau responded to numerous requests for information on census operations as they unfolded. Beginning January 19, 2001, the Census Bureau took the unprecedented step of providing advance access to extensive source data, including A.C.E. microdata, to its major oversight groups—our panel, the 2000 Census Monitoring Board, and the House subcommittee. The intent was to enable these groups to become familiar with the files that would form the basis of the evaluations to be reviewed by ESCAP. The data were made available under a memorandum of understanding not to make public any information that would identify individuals.7 It was also agreed that no summaries would be made public until ESCAP had announced its recommendation and released its report and supporting evaluations publicly on the Bureau’s Internet web site (http://www.census.gov/dmd/www/EscapRep.html). Overall, the thoroughness with which the Census Bureau has documented every step of its procedures and the amount of information the Bureau has shared with our panel and others are noteworthy and deserving of high praise. In its conduct of the 2000 census and the A.C.E. Program, the Census Bureau has reacted positively in its willingness to accept scientific and public review and criticism in the interests of the best possible outcome for the census. MARCH ADJUSTMENT DECISION The March 1, 2001, decision by the Census Bureau not to recommend adjustment of the census counts for purposes of legislative redistricting was surprising to many in light of the evolution of the census design over the decade 6   The Clinton Administration had issued a regulation that delegated the authority for the adjustment decision from the Secretary of Commerce to the director of the Census Bureau; the Bush Administration reversed that action, returning the authority to the secretary. 7   The microdata are only available for analysis at a secure site at Census Bureau headquarters. Persons designated by each group to have access to the microdata were sworn in as special Census employees for that purpose.

OCR for page 23
The 2000 Census: Interim Assessment of the 1990s and particularly due to the more prominent role of the A.C.E. relative to the analogous 1990 Post-Enumeration Survey (PES). Also, public statements by Census Bureau officials before the census was completed stressed the limitations on the ability to count everyone in the nation through field procedures and the likelihood that statistical adjustment would improve the estimates of the population for important purposes.8 Census Bureau’s Decision Process The Census Bureau reached its conclusion not to adjust after carefully following the decision process it had specified, which was publicly explained at the panel’s workshop in October 2000. All of the evaluations that the Bureau proposed to conduct were completed and reviewed by ESCAP. Given the time constraints, these evaluations could not be exhaustive, but included detailed assessments of A.C.E. operations, supplemented by more limited assessments of census operations and comparisons of adjusted and unadjusted census counts for different levels of geography.9 It was hoped that these assessments, which largely addressed how well operations were performed, would provide sufficient information to conclude that adjusted counts did (or did not) represent an improvement over the counts from the census process. In addition, the Census Bureau planned to take account of population estimates from demographic analysis, which have historically provided a comparison standard for the census. What, then, were the reasons for the decision not to adjust? An important reason cited by the ESCAP report was the inconsistencies between the population estimates from the census, the A.C.E., and demographic analysis; those inconsistencies could not be resolved or explained with the available evaluation data within the time available for the decision. As shown in Table 1-1, the A.C.E. estimated that the overall net undercount dropped from 1.6 percent of the population in 1990 as measured by the Post-Enumeration Survey to 1.2 percent of the population in 2000 and that the net undercount for blacks dropped from 4.4 percent in 1990 to 2.1 percent in 8   For example, Census Bureau Director Kenneth Prewitt wrote: “The Census Bureau has determined that the A.C.E. is operationally and technically feasible and expects, barring unforeseen operational difficulties that would have a significant effect on the quality of the data, that these corrected data will be more accurate than the uncorrected data for their intended purposes” (Prewitt, 2000:2). 9   The A.C.E. evaluations covered rates of noninterviews in the independent P-sample and missing data in the P-sample and the census-based E-sample; quality control of the matching process; the extent of imputation required for unresolved match and enumeration status; inconsistent assignment of sample cases to estimation groups in the two samples; and variance due to sampling and imputation error in the dual-systems estimates. The census evaluations covered mail return rates; quality assurance of enumerators’ field work; results of unduplication operations; and extent of missing data. Comparisons with 1990 census data were included when feasible. Documentation of these studies is provided in the “B” memoranda on the Census Bureau’s web site.

OCR for page 23
The 2000 Census: Interim Assessment TABLE 1-1 Alternative Estimates of Percentage Net Undercount of the Population, April 2000 and 1990   2000 1990   Demographic Analysis   Demographic Analysis   Category Base Alternate A.C.E. PES Total population –0.65 0.32 1.15 1.85 1.58 Black 2.80 3.51 2.06 5.68 4.43 Male 5.10 5.81 2.37 8.49 4.90 Female 0.63 1.32 1.78 3.01 4.01 Nonblack –1.19 –0.17 1.02 1.29 1.18 Male –0.93 0.17 1.40 1.97 1.52 Female –1.44 –0.50 0.64 0.63 0.85 Difference, black-nonblack 3.99 3.67 1.04 4.39 3.25 NOTES: All estimates include the household and group quarters population. Net undercount rates are calculated as the estimate (from demographic analysis, A.C.E., or PES) minus the census count divided by the estimate. The census count of the population in 2000 was 281.4 million; the census count in 1990 was 248.7 million. The census count by race in 2000 varies depending on the treatment of people who reported more than one race. The net undercount estimates shown for 2000 are an average of estimates calculated using two different tabulations of the census by race (see Chapter 5). Minus sign (–) indicates a net overcount of the population. For 2000 demographic analysis net undercount rates, “base” uses the original estimate that includes an allowance for 6 million undocumented immigrants; “alternate” uses an estimate that arbitrarily doubles the flow of undocumented immigrants between 1990 and 2000, allowing for 8.7 million undocumented immigrants total (see Chapter 5). SOURCE: Robinson (2001a:Table 6). 2000. The A.C.E. also estimated marked reductions in net undercount rates from 1990 for children under age 18, people who rent their homes, Hispanics, Asians, and American Indians on reservations (see Chapter 6). The reductions in net undercount are heartening, although there is a puzzling aspect to them that the ESCAP report did not discuss: because estimates for key components of the dual-systems estimation formula were similar in the A.C.E. and the 1990 Post-Enumeration Survey, the net undercount should have been similar in both years, other things equal. However, the Census Bureau’s initial estimate from demographic analysis indicated that the 2000 census resulted in a slight (0.7%) net overcount of the population and that the A.C.E. overstated the population by even more. Even when the Bureau adjusted the demographic analysis estimate upward to allow for a larger number of undocumented immigrants than were part of the base estimate, the net undercount in 2000 was only 0.3 percent of the population. Demographic analysis agreed with the A.C.E. in estimating a reduction in net undercount for children. Demographic analysis also estimated that black and nonblack net undercount rates differed by about 3.7–4.0 percentage points, compared with only 1.0 percentage point as measured by the A.C.E. (see Chapter 5).

OCR for page 23
The 2000 Census: Interim Assessment The ESCAP report did cite several areas of concern about A.C.E. operations that might have affected dual-systems estimation. It questioned the level of “balancing error” that may have occurred in a procedure called targeted extended search. (Balancing error is when different criteria, such as different areas of search, are used in processing the independent survey and the sample of census enumerations—see Chapter 6.) It also questioned the level of “synthetic error” that may have occurred for dual-systems estimates of population groups called post-strata. (Synthetic error occurs when the people included in a post-stratum—who are intended to have the same likelihood of being included in the census or the A.C.E. —are in fact not sufficiently similar in this respect.) The report also considered late additions to the census and cases of people for whom minimal information was obtained in the field (who required imputation to complete their census records). Neither of these groups could be included in the A.C.E. There were substantially more such people in 2000 than in 1990, but the report concluded that they did not likely affect the dual-systems estimates.10 The Census Bureau had always planned a longer term evaluation program, in addition to the short-term evaluations that were feasible to carry out before March 1, 2001. Given its conclusion that the accuracy of the census, the A.C.E., and demographic analysis population estimates could not be definitively established, the Bureau has expedited several evaluations on the longer term agenda and is carrying out additional evaluations to help reach its planned decision by October 15 on whether to recommend adjustment of census population estimates for such purposes as fund allocation. Release of these evaluations is expected to accompany the decision. Panel’s Assessment The panel, in its letter report issued November 2000 (National Research Council, 2000b:2), commented on the Census Bureau’s evaluation process for the March 2001 decision as follows: The planned analyses appear to cover all of the evaluations that can reasonably be expected to be completed within the time available. Furthermore, they appear to be sufficiently comprehensive that they will likely provide support for a reasonably confident decision on adjustment in March. However, since the numbers themselves, which are, of course, critical to the evaluation process, are not yet available, it is not possible at this time to comment on what the adjustment decision should be nor to conclude definitively that the planned short-term evaluations will be adequate to support the decision. 10   A lawsuit brought by the state of Utah is challenging the use of some kinds of imputed census records for purposes of congressional reapportionment (see http://www.attorneygeneral.utah.gov).

OCR for page 23
The 2000 Census: Interim Assessment As it turned out, the Bureau concluded that the evaluation studies did not provide sufficient information to decide that adjusted counts would be clearly preferable to unadjusted counts for redistricting. Although not mentioned by the Census Bureau, reaching a conclusion on this point is more difficult when the adjustments to be made for population groups are generally small11 The panel does not necessarily agree with the weight that the Bureau gave to each factor in its decision: specifically, we conclude that demographic analysis estimates are sufficiently uncertain that they should not be used as a standard for evaluation at this time (see Chapter 5). Nonetheless, we believe that the Bureau followed a reasonable process. We also believe that its decision not to recommend adjusting the census data in March was justifiable, given its conclusion that additional evaluations of the quality of the A.C.E. —and of the census itself—were needed to resolve its concerns. The fact that the Bureau did not recommend adjusting the census counts to be provided for redistricting does not carry any implications for the usefulness of statistical adjustment methods based on dual-systems estimation. In particular, the panel views it as an open question whether adjusted counts would constitute an improvement over unadjusted counts for such purposes as fund allocation. The panel will assess the Census Bureau’s October decision on the basis of the evidence available at that time (see Part I). SCOPE AND LIMITATIONS OF THIS INTERIM REPORT This interim report provides a preliminary assessment of census and A.C.E. operations, briefly reviews the Bureau’s demographic analysis, and examines the puzzle from the reduction in net undercount measured by the A.C.E. The report does not address the superiority of one or the other set of population counts (the census or the A.C.E. estimates) for particular uses of the data. The information available to the panel in preparing this report is extensive but does not make it possible to draw definite conclusions about the quality of the census operations or the quality of the resulting data. This report reviews the available information to identify areas for further research and to provide background for subsequent reports of the panel. The panel may revise its interim assessments when new information becomes available. This report has eight chapters (including this introduction), a glossary, and three appendixes. Chapter 2 considers evaluation issues, including the multiple uses of census data (which complicate the task of evaluation), sources of error in the census (from which no data collection or estimation effort can be completely free), and types of evaluation. Chapter 3 provides background information on 2000 census operations, noting important differences from the 11   A small (or zero) net undercount for the population as a whole is not a reason for or against adjustment because net undercounts can mask sizable gross errors of omissions and erroneously included enumerations. The issue is how the balance between these components of error differs among population groups and geographic areas, resulting in different net undercount rates.

OCR for page 23
The 2000 Census: Interim Assessment procedures used in 1990. (Appendix A provides a more detailed description.) Chapter 4 provides the panel’s initial assessment of census operations. (Appendix B provides detailed results from an analysis of mail return rates in 1990 and 2000.) Chapter 5 reviews demographic analysis and the issues surrounding the demographic estimates for 2000. Chapter 6 provides background information on the population coverage estimates from the A.C.E., the method of dual-systems estimation, and the A.C.E. operations, including important differences from the procedures used in the 1990 PES (Appendix C provides a more detailed description of the A.C.E. operations). Chapter 7 provides the panel’s initial assessment of the A.C.E. operations. Finally, Chapter 8 examines the characteristics of two groups of people who could not be included in the A.C.E. They are people whose census records were available too late for A.C.E. processing and people who required imputation to complete their census records. These groups, especially the second, account for the puzzling aspects in the A.C.E. and PES results noted above. The panel looks forward to continuing its evaluation of the census and the A.C.E., as additional information becomes available that can support firmer conclusions about the quality of the adjusted and unadjusted census data and their appropriate use. The panel notes that the census obtained a wide range of socioeconomic data on the long form (which was sent to about one in six households). These data are also important to assess, given the use of long-form estimates for allocation of billions of dollars of federal funds to states and localities and for other important policy purposes. Processing of the 2000 long-form data has not been completed; the panel expects to assess those data in its final report, as part of the panel’s final assessment of the 2000 census.