Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
l T ~ ~ t. EVERY 10 YEARS, THE U.S. CENSUS BUREAU faces a job of almost bewildering complexity: counting an ever-increasing, ever-moving, ever-cliversifying population and accurately tab- ulating it by location so that election districts anc! other mechanisms of democratic government can be recalibrated to better reflect their constituencies. American citizens' most recent involvement with the census was simply returning a form in 2000, en c! their next expected involvement is doing the same 7 years from now in 2010. For the average citizen, then, it may seem strange to react now about plans for the seemingly far-off 2010 census. In truth, the complexity of the task clemancis early long-term plan- n~ng. Any project to execute fundamental changes in the way a census is conducted demands even earlier planning still. The 2000 census is not yet over by some metrics data products from 2000 are still be- ing issucc! on a flow basis, en c! the Census Bureau's program of inter- nal evaluations regarding 2000 census operations is far from complete. However, plans for the 2010 census are well uncler way. IncleecI, the years 2002 anct 2003 are not an unusual time to be discussing the 2010 census; it is closer to the truth to say that they are the deadline for effecting real change in the way the 2010 census is aclministerecl. 11
12 PLANNING THE 2010 CENSUS In 1999, the Census Bureau requested that the Committee on National Statistics (CNSTAT) convene a Panel on Research on Future Census Methods to assist the early planning efforts for the 2010 census. This report is the panel's third published comment on the emerging plans for the 2010 census. In the remainder of this section, we describe the scope of this report. To clarify the discussion, we briefly outline the major parameters of the developing plans for the 2010 census, as envisioned by the Census Bureau, with particular attention to the contrast between the 2000 census anct the developing plans for 2010. We then describe the panel anc! its charge, en c! provide an overview of the remainder of the report. THE "THREE-LEGGED STOOL" APPROACH TO THE 2010 CENSUS By many accounts, the planning process for the 2000 census was fraught with risk anc! ultimately chaotic. As summarized in National Research Council (2001a), the Census Bureau initially clevelopect a plan in 1996 that would have used statistical sampling during the pro- cess of following up with households that failed to file a mail return. Sample-basec! methods were also to be usec! to adjust final population counts for all purposes including congressional reapportionment anc! redistricting to reflect census unclercount. This proposer! plan touched off conflict that ultimately resulted in a January 1999 ruling by the U.S. Supreme Court forbidding the use of sampling to generate numbers for congressional reapportionment. The decision forced the bureau to completely overhaul the census plan little more than a year from April 1, 2000, the census target ciate. Unanticipated difficulties also impacted parts of the census process. Between January anct May 1999, when census fielc! staff concluctec! an extensive canvass of the entire address list, concern arose that the list contained coverage gaps. During the actual conduct of the census, further evaluation of the acictress list suggested that the list tract sizable levels of duplicate hous- ing unit aciciresses, leacling to an act hoc operation to screen potential duplicates for further examination anct possible reinstatement into the census (Nash, 2000~. ~DepartmentofCommercev. U.S. House ofRepresentatives, 525 U.S. 316 (1999~.
INTROD UCTION 13 The 2000 census was ultimately successful in meeting its statutory cleacilines for providing data for reapportionment anc! redistricting, but the process by which it clevelopect leaves considerable room for improvement. The final design for the 2000 census was only put into place an inadvisably short time before the census tract to go into the fielcI. Looking ahead to 2010, both the bureau anc! outside observers hope to avoict the risks anct bruising consequences of late-formect plans, while at the same time keeping the escalating costs of conducting a census of the complex U.S. population in check. In the early planning stages, the Census Bureau iclentifiec! four basic goals for the 2010 census (Waite, 2002~: 1. increase the relevance anct timeliness of census long-form clata; 2. recluce operational risk; 3. increase coverage, accuracy, anc! quality of census ciata; anc! . 4. contain costs. Basect on those goals, the Census Bureau clevelopect a general strat- egy for the 2010 census even as 2000 census returns were still being pro- cessect. As first clescribect to the panel at its December 2000 meeting, the Census Bureau's general strategy for 2010 was likened to a "three- leggecl stool." Specifically, the general strategy is predicated on three . . . . . major ~n~t~at~ves: . Modernization of the Census Bureau's geographic resources. Specifi- cally, the bureau's Master Acictress File (MAF) anct its geographic database (Topologically Integrated Geographic Encocling and Referencing System, or TIGER) will be upclatect so that they will be consistent with coordinates clerivect using global positioning systems (GPS). The intent is to save fielc! time anc! costs. · Implementation of the American Community Survey (ACS). This proposed sample survey will collect data on the same social, eco- nomic, anc! demographic variables incluclec! in the current cen- sus long form, but will do so on a rolling continuous-time ba- sis. Like earlier censuses, the census long form was aciministerect to a sample of households in 2000 (1-in-6) while most house- holcls received a short form. However, full ACS implementation
14 PLANNING THE 2010 CENSUS will permit the 2010 census to be concluctec! using only the short form. Accordingly, it is hopec! that this change will facilitate eas- ier collection of information through the Internet anct simplify data capture from census forms returnee! by mail. · Early integrated/planning. To the extent possible, census plans will be finalizes! as early as possible to facilitate effective testing in the years leading up to the census. It is hoped that this early planning will make the pre-2010 census tests more useful anc! informative as well as forestall a costly encl-of-the-clecacle crunch in finalizing census operational plans. An immediate adjunct to this three-prongecl strategy is the incor- poration of new technology in the census process. In particular, the Census Bureau's emerging 2010 census plans take advantage of a short- form-only census by including the following acictitional components: · Multiple response modes. Simplifying to a short form would make completion of the census form easier anc! quicker anc! more tractable for administering to respondents electronically. Hence, it is anticipated that the mailout-mailback component of the census would be heavily augmented with enumeration through use of the Internet anc! possibly interactive voice response via the telephone. · Mobile computing devices (MCDs) .2 Nonresponse fielc! work will make use of hanct-helct computing crevices for communication of assignments, computer-assistec! interviewing, anc! data capture. Making use of an enhanced MAF/TIGER clatabase, the Census Bureau also anticipates that MCDs equipped with GPS receivers 2The Census Bureau uses the terminology "mobile computing devices" or, more frequently, the acronym "MCD" to describe the small computers planned for 2010 field data collection. However, the choice of MCD as a label is confusing given the acronym's long-standing meaning to data users accustomed to census geography. In that context, it stands for "minor civil division," the subcounty (township) divisions that are functioning governmental units in several midwestern and northeastern states. Though current plans and goals seem to favor a hand-held computing device on the or- der of current Palm Pilots, the alternative terminology "hand-held computing device" (HCD) may be too prescriptive. Hence, "portable computing device" (PCD) would likely be a better term, but we adhere to the MCD terminology for consistency with Census Bureau usage.
INTROD UCTION 15 will allow interviewers conducting nonresponse follow-up to pin- point the location of their assigned housing units ancl, possibly, . . . r to optimize to ae1r navigation trom one assignment to ants per. Final specifications anc! cletailec! plans for the above design remain to be clevelopect; so too clo complete plans for acictressing other op- erational challenges anc! traditionally vexing problems in the coming census. These include: · collection of census data from group quarters; · enumeration of various harct-to-enumerate populations; · use of administrative records either for nonresponse follow-up or acictress list improvement; · forms of data dissemination; · question worcling, specifically with regard to the nature and con- tent of questions on race and Hispanic origin; and · plans for coverage measurement and evaluation. Planning Milestones of the 2010 Census The Census Bureau plans to conduct at least four major census tests prior to 2010 in orcler to try out new procedures anct finalize program plans. The chronology of these tests anc! other milestones in the plan- ning process for 2010 are shown in Table 1-1. In 2003, a national sample will be askoc! to participate in a test of possible response mocles (e.g., mail, Internet, anc! telephone) anc! of rewordings of the questions on race anct Hispanic origin. The 2003 test will be aciministerect only by mail anc! will not involve an active fielc! deployment of enumerators to conduct follow-up questioning. The 2004 Census Field Test will cover a wicler range of census operations, including fielc! follow-up, in pre- cleterminect sites in Georgia, New York, anct Illinois.3 The 2004 fielct 3 Under budget totals consistent with the Bush administration's requests for fiscal year 2004, the Census Bureau would scale back the 2004 test to omit the Illinois site (Lowenthal, 2003b>. The 2004 and other census tests will be discussed in greater detail in a later chapter.
16 PLANNING THE 2010 CENSUS Table ]-] Planned Testing and Development Cycle for the 2010 Decennial Census, Assuming Short-Form-On~y Census Year Census Activity 2002 Begin planning and develop methods for 2004 Census Field Test 2003 Conduct 2003 National Census Test, a survey administered by mail but offering multiple response modes (mail, telephone, Internet) and rewording of race and Hispanic origin questions 2004 Conduct Census Field Test, emphasizing use of mobile computing devices, in selected sites in New York, Illinois, and Georgia; Conduct Overseas Enumeration Test in France, Kuwait, and Mexico 2005 Analyze results and refine methodology 2006 Conduct Census Test, involving prototype technical systems 2007 Analyze results and refine and integrate systems and methods 2008 Dress rehearsal 2009 Begin to implement operations 2010 Conduct census SOURCES: Waite (2002~; U.S. Census Bureau, Public Information Office (2003b). test shoulc! be the first major test of the bureau's plans for using mo- bile computing crevices. In January 2003, the Census Bureau announced an Overseas Enumeration Test to be fielclec! in 2004, a test intenclec! to gauge the response of U.S. citizens living in France, Kuwait, anct Mexico to outreach anc! marketing efforts (U.S. Census Bureau, Public Infor- mation Office, 2003b). In 2006, an as-yet unspecified test will focus on general anct reengineerect technical systems. Finally, a full-flecigect dress rehearsal will be conducted in 2008. It is hoped that avoiding a late-clecacle crush in designing census plans will make the 2008 exercise a true rehearsal rather than a late experimental test, as was the case with the 2000 census dress rehearsal.4 4In spring 1998, debate over the use of sampling methods in nonresponse follow- up led to the dress rehearsals for the 2000 census being cast as a comparison of basic designs. The original sampling-based framework was tested in Sacramento, California; a traditional census with a post-enumeration coverage survey was tested in Columbia, South Carolina; and a hybrid approach was fielded in Menominee County, Wisconsin.
INTROD UCTION 17 CHARGE AND BRIEF HISTORY OF THE PANEL The Panel on Research on Future Census Methods has a broac! charge to review the plans for acquisition, analysis, anct evaluation of research data neeclect to begin planning for the 2010 decennial census. The panel is charged to suggest improvements and preferred approaches, as well as to suggest priorities for analyzing census exper- imental anc! tracking data. Having been formec! cluring the buildup to the 2000 census, a major role of the panel was to observe the implemen- tation of the 2000 census, examining census accuracy anc! evaluating research program results, in orcler to determine appropriate lessons for the 2010 census. Because of its early formation, this panel is unusual in the experi- ence of previous National Research Council panels regarding the cle- cennial census. This early start has been crucial for giving advice on the broac! shape of the 2010 census, but it has presented unique challenges. Since its inception, the panel underwent two nearly year-long periods of relative dormancy clue to the clemancis of carrying out the 2000 cen- sus. The first hiatus arose in 2000 clue to heavy clemanct on the Census Bureau anc! its senior staff cluring the active follow-up anc! processing of the 2000 census. The seconct hiatus occurrect cluring the summer anct fall of 2001 as the Census Bureau engaged in intensive research over the question of whether to statistically adjust census data for estimated unclercount.5 However, the work of the panel continucct cluring these hiatus periods in the absence of formal panel meetings. Members of the panel joined members of its sister CNSTAT panel the Panel to Review the 2000 Census to visit census operations centers anc! local census offices cluring the conduct of the 2000 census. In acictition, both panels 5As referenced earlier, the U.S. Supreme Court ruled in January 1999 that sampling methods could not be used in deriving the tallies used to reapportion the U.S. House of Representatives. This ruling left adjustment of census data for other purposes as an open question. Confronted with inconsistencies that the Census Bureau felt it could not resolve to its satisfaction in time to meet a legally mandated deadline for delivery of data to the states, the Census Bureau decided in March 2001 that it would not adjust the data used for redistricting. After still more research, the Census Bureau decided in October 2001 that it would not adjust 2000 census data for other nonredistricting purposes. National Research Council (2001a) provides additional information on the 2000 census adjustment debate and decisions.
18 PLANNING THE 2010 CENSUS jointly established a working group to evaluate the Local Update of Census Aciciresses (LUCA) program, by which state, local, anc! tribal governments could review address lists or population counts for their areas anc! suggest revisions. The Working Group on LUCA completed its report to both panels in early 2001 (Working Group on LUCA, 2001). Previous and Future Reports of the Pane} This seconc! interim report is the third report issucc! by the panel. In February 2000, the panel issucct its first interim report, Designing the 2010 Census (National Research Council, 2000), based on early infor- mation gleaned from the panel's first two meetings. In December 2000, the panel heart! the Census Bureau's first presentation of its preliminary 2010 census strategy anct offered early feedback on the general strategy in a letter report to acting census director William Barron in February 2001 (National Research Council, 2001c). The Panel on Research on Future Census Methocls will complete its work by the end of 2003, at which time it will issue a fourth anc! final report. THE IMPORTANCE OF INTEGRATION Though it covers a wide range of the topics under the panel's charge, this interim report is neither intended to be an exhaustive assessment of the Census Bureau's plans for the 2010 census nor a comprehensive checklist of problem areas whose solution woulc! ensure a quality 2010 census. Insteact, the primary focus of the report is to underscore the importance of one of the three major initiatives of the 2010 census plan envisioned by the Census Bureau: early integrated planning. The panel wrote in its letter report (National Research Council, 2001c) that it "generally agrees with the Census Bureau on the goals of the two major initiatives of the current 2010 census strategy the ACS and the MAF/TIGER upgrade and strongly agrees on the importance of early planning." However, the three-prongecl strategy for 2010 "ap- pears to lack an overall framework." Put another way, the panel was concerned that the Census Bureau's planning for 2010 was commenct- ably early but sufferer! from lack of attention to integration. It was unclear how the various pieces of the proposed plan interact anct sup-
INTROD UCTION 19 port each other. For example, it was unclear whether upciatec! MAF/ TIGER extracts woulc! be available to support the sample selection anc! data collection processes of the ACS, or whether they would be avail- able in time to support effective testing of mobile computing crevices in the various census tests prior to 2010. The panel wrote that the Cen- sus Bureau's plan stressec! the importance of integration but insteac! suggested "compartmentalizecl thinking, without clue attention to how efforts across clivisions within the Census Bureau will be coorclinatec! and synthesizecl." Consequently, we urged that the bureau develop a "business plan" that is, a clear statement of objectives anc! how they will be accomplished, when various steps in the census process must be completed, how much those steps will cost (in terms of both monetary anc! nonmone- tary resources, including those that must be ctivertect from other areas to complete tasks in a timely fashion), the clegree to which the steps will interact, anct what benefits will accrue through each step. A critical feature of such a "business plan" for the 2010 census is a full enumeration of the costs anc! benefits either monetary or nonmonetary associated with each component of the census strategy. An example of a nonmonetary benefit is the improvement in quality of short-form data en c! timeliness of long-form data if the long form is replaced by the ACS. Then, as now, the panel remains convinced that integration is key to development of a strong plan for the 2010 census. A primary goal of this interim report is to build on the recommendations from the let- ter report anc! in the course of providing an initial assessment of 2010 census programs suggest areas in which greater attention to integra- tion of programs anc! efforts is neeclecI. Structure of This Report In Chapter 2, we examine one segment of 2010 census planning in which the notion of integration is solicily established anct uncler way. Since 2000, the Census Bureau has sponsored a pilot project to cloc- ument the logical architecture of the decennial census essentially, to map all activities anct their information clepenclencies involved in the
20 PLANNING THE 2010 CENSUS census process from the initial development of aciciress lists through to final data outputs. This mapping of the logical architecture is a ma- jor step toward developing the business plan we suggested in our letter report. More importantly, establishment of the logical architecture- beginning with an "as-was" moclel of the activities anct their informa- tion clepenclencies involvec! in the 2000 census allows for the moclel to be acljustect to reflect new assumptions. Revisect "to-be" models can then be compared against each other to solidify plans for the 2010 cen- sus process. This information then becomes an important resource for specifying anc! assembling the actual physical technical infrastructure of the census anct the myriad computer information systems that must work in sync to achieve census goals. In subsequent chapters of this report, we then turn to the other two major initiatives envisioned in the Census Bureau's 2010 strategy, noting our concerns that it is not yet clear how these two broacI-stroke initiatives will achieve clesirect goals either on their own or in concert with each other. In Chapter 3, we examine the bureau's proposed MAF/TIGER Enhancements Program. The basic objectives of the MAF/TIGER Enhancements are laudable anc! worthy of praise. How- ever, in terms of furthering the Census Bureau's overall goal of accurate enumeration in 2010, the panel is worried that the bureau's current plans do not put sufficient weight on the MAF part of the program. In other worcis, the current Enhancements Program cloes not clearly document how aciciresses will be aciclec! to the Master Aciciress File anc! from which sources, en cl more importantly does not clearly tie in to efforts to ensure that the MAF is free of duplicates anc! has full coverage of acictresses in multiunit structures. The American Community Survey, which we describe in Chapter 4, is a new potential source of fine-level clata on the American populace on a much more timely basis than the once-a-clecacle snapshots currently afforclect by the decennial census. The concept of continuous sample- basecl estimation is one of great statistical import and could surely be the basis for a report in its own right; the treatment in this report is cleciclecily not intenclec! as our complete statement on the ACS. Our message on the ACS is much the same as our message regarding the MAF/TIGER Enhancements Program. The potential merit of the ACS is indisputable, anc! we support its implementation. But a compelling case has not been made for how the ACS fits into the broader goals
INTROD UCTION 21 of the census. The clear objective of the ACS is to supplant a long- form component of the 2010 decennial census, but it remains to be fully clemonstratect that the survey can actually meet all the neects that census long-form data currently satisfy. Does the increased currency of ACS data offset increases in variability inherent in the data as those figures are usec! by census stakeholclers? We briefly outline challenges that remain in building a case for ACS estimation anct look forward to continucc! work on the topic. Incluclect within the early integrated planning component of the Census Bureau's 2010 strategy is the goal of maximizing the benefit of early, comprehensive testing of revisect census processes over the course of the clecacle. Major census tests are plannec! in 2003, 2004, . anc! 2006, with a ciress rehearsal to be helc! in 2008. In Chapter 5, we comment on the shape of the 2003 test as it was presented to the panel. As yet, the plans for the 2004 anc! 2006 tests are sufficiently unclear to the panel that our ability to provide precise comment is limited; in the chapter, we sketch some clesirec! features for the upcoming tests. We close this discussion in Chapter 6 by suggesting areas of strong interest to the panel that cut across clivisions of the Census Bureau anc! deserve attention in the coming months. We find much in the extant planning for the 2010 census that is encouraging, and we are confident that with continuing efforts to describe how all the various pieces fit and work together the Census Bureau's particularly early start on lay- ing the foundation for the 2010 census will ultimately serve it well.
