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.
Summary Throughout the history of the United States, and indeed in democracies throughout the world, censuses have been key sources of information to facilitate governance and improve public understanding of nations and their communities. An evaluation of alternative census methods involves a major component of a national statistical system that produces knowledge for many purposes. The U.S. Census of Population and Housing serves two distinct functions. First, there is a constitutional mandate (Article 1, Section 2) to conduct an enu- meration of the national population every 10 years for the reapportionment of congressional seats. The population counts are also used for the redistricting of political jurisdictions at all levels of government and the allocation of federal program funds. In addition to counting people, the decennial census serves as a linchpin of the federal statistical system by collecting data on the characteristics of individu- als, households, and housing units throughout the country. The census is uniquely positioned to gather comprehensive and comparable information for communi- ties and population groups, small and large, that cannot be obtained from any other source. THE CHALLENGE FOR 2000 The broad challenge facing the U.S. census in the year 2000 is to continue to meet information requirements for the country in an environment in which it is increasingly difficult to collect good information by traditional methods. The two strongest criticisms leveled against the 1990 census are that unit
2 COUNTING PEOPLE IN THE INFORMATION AGE costs increased significantly, continuing a trend that began with the 1970 census, and that the problem of differential undercount by race persisted and possibly worsened, despite a large investment in programs that were intended to improve coverage (see, e.g., U.S. General Accounting Office, 1992~. These criticisms have contributed to the growing momentum and advocacy for fundamental charge in census operations. In constant (1990) dollars, the unit costs of counting a household have in- creased significantly in the past 30 years: less than $10 in 1960, $11 ire 1970, $20 in 1980, and $25 in 1990. These increases in expenditures have been accompa- nied by the persistent problem that the people missed by the census are not representative of the population as a whole: the Census Bureau estimates that the net undercount in the 1990 census was 1.6 percent of the U.S. population, or about 4.0 million people, and the estimated difference in the undercount rates for blacks and non-lIispanic whites was 3.9 percentage points (Hogan, 1993~. Dif- ferential coverage by certain charactenstics, such as race and geographic loca- tion, has significant implications for political representation and allocation of federal program funds. Measures of gross census error are also important indicators of census data quality. Such measures consider not only omissions (that produce undercounts) but also erroneous enumerations (that produce overcounts). The 1990 census included approximately 11 million erroneous enumerations, the largest number recorded to date (Bryant, 1993~. All undercounts and overcounts complicate the task of accurately measuring census net coverage. Undercounts and overcounts that are nonunifo~ly distributed among particular areas or types of people lead to misdistr~bution of the estimated population, even when such errors balance at larger levels of aggregation. Some advocates of census reform have also questioned the collection of detailed sociodemographic data as part of the decennial census. Since 1960, this additional information has been gathered by distributing a census "long form" to a national sample of households. One current argument is that the accuracy of the decennial population figures would be improved if long-form data collection is eliminated, reduced, or displaced in time from the effort to enumerate the popu- lation. Others have suggested that some of the data gathered in the 1990 census could be collected through alternate methods and made available for use in a more timely manner (see, e.g., Sawyer, 1993; U.S. House of Representatives, 1993~. Still others have challenged the quality of data collected on the long form, noting the high rates at which this information is gathered indirectly (relative to ~ These results were obtained by analysis of data from the 1990 Post-Enumeration Survey. The estimated undercounts for blacks and non-Hispanic whites were 4.6 and 0.7 percent, respectively. Independent estimates based on demographic analysis were similar: the net undercount was esti- mated to be 1.8 percent of the population, and the estimated undercounts for blacks and nonblacks (the latter category includes Asians and Hispanics) were 5.7 and 1.3 percent, respectively (Robinson et al., 1993).
SUMMARY - the decennial short form), either by imputation or from someone outside the household, particularly for minority populations (see, e.g., Ericksen et al., 1991~. The panel's work emphasizes those aspects of census methodology that have the greatest potential effect on two primary objectives of census redesign: reduc- ing differential undercount and controlling costs. In particular, we focus on processes for the collection of data, the quality of coverage and response that these processes engender, and the use of sampling (and subsequent estimation) in the collection process. Our field of examination is not restricted to the 2000 census: a significant number of our findings and recommendations look beyond 2000 to future censuses, relate to other Census Bureau demographic programs (current population estimates and sample surveys), and discuss the collection of small-area data from administrative files. Redesigning the nation's census should not be carried out without due consideration of other components of the Census Bureau's demographic program and the larger federal statistical system. The census data collection processes involve four key steps: (1) the con- struction of an address frame, (2) an initial process to obtain responses that can be linked to the address frame, (3) a follow-up process to obtain responses from those not covered in the initial process, and (4) a coverage assessment process that estimates the size of the population not covered through the initial and follow-up processes. In the 1990 and earlier censuses, the first three steps led to the official census estimates; whether or not to incorporate the estimates from the fourth step into the official census estimates became the "adjustment issue." For the 2000 census, the Census Bureau is proposing a fundamentally different ap- proach, called a "one-number census," in which this fourth step is an integral part of the census process that leads to the official estimates. The design of a census data collection process in essence amounts to decid- ing which methods of identification, enumeration, response, and coverage im- provement should be applied at each of the steps; whether sampling methods (and the corresponding estimation methods) should be used at any of the four steps; and, if sampling methods are used, which methods and at which steps. These decisions should be based on information about the effectiveness and costs of the various alternative methods; the 1995 census test should be a prime source of such information. In subsequent sections of this summary, we list key recommendations, num- bered in order of appearance within the report chapters. A complete list of recommendations appears at the end of the report. MAJOR INNOVATIONS FOR THE 2000 CENSUS The panel found much progress in census research and development since the 1990 census, and this progress reflects new and creative thinking at the Census Bureau. The two main areas in which innovation in census design is taking place are response and coverage improvement and expanded use of sam
4 COUNTING PEOPLE IN THE INFORMATION AGE pling and estimation. Research on response and coverage improvement has led to potentially important changes in questionnaire design and implementation. Sam- pling and statistical estimation methods are being explored to close the remaining differentials in census coverage while controlling, or even reducing, overall cost. The Census Bureau will test a variety of innovative design features in the 1995 census test. Collection of reliable information in the 1995 census test about the costs and effectiveness of census design components will be essential for their proper evaluation in particular, to inform decisions about allocating resources between efforts to improve primary response and efforts to use sampling and estimation methods to correct the counting operation. The operational constraints on the 1995 census test underscore the impor- tance of learning as much as possible from other research. For example, simula- tion studies using 1990 census data can investigate the effects of truncating nonresponse follow-up operations at different points in time, using different rates of sampling nonrespondents for follow-up, and applying different coverage mea- surement methods. Similarly, not all methods need to be tested in large-scale field settings. To ease experimental complexity, certain methods might be ex- cluded from large-scale field testing in 1995 when such an exclusion would not disrupt the research and development program or if smaller experiments (e.g., questionnaire research) conducted simultaneously with the 1995 census test will provide useful information. Sampling, Estimation, and the One-Number Census A key panel finding concerns the validity of the use of sampling and estima- tion in census-taking. The panel concludes that sampling and associated statisti- cal estimation constitute an established scientific methodology that must play a greater role in future censuses in order to obtain a more accurate picture of the population than is provided by current methods. For this reason, the panel en- dorses the Census Bureau's stated goal of achieving a one-number census in 2000 that incorporates the results from coverage measurement programs, including programs that involve sampling and statistical estimation, into the official census population totals. The panel also recommends continued methodological re- search, development, and testing in pursuit of this goal. Recommendation 4.2: Differential undercount cannot be reduced to acceptable levels at acceptable costs without the use of integrated cover- age measurement and the statistical methods associated with it. We endorse the use of integrated coverage measurement as an essential part of census-taking in the 2000 census. One proposed method for integrated coverage measurement, CensusPlus, involves intensive enumeration methods and highly trained interviewers with the objective of obtaining a complete enumeration of the population in a sample of
SUMMARY census blocks. Tentative plans call for the CensusPlus operation to be Gamed out concurrently with mail and follow-up operations, and data from this coverage measurement survey would be used in producing the final, official population totals. Because it is a fundamentally new approach to measuring census cover- age, the CensusPlus operation will require thorough testing alla evaluation prior to approval of its use in a decennial census. Estimation methods that might be used in integrated coverage measurement will require further study using simula- tions of 1990 census data and field data from the 1995 census test. Recommendation 4.3: The Census Bureau should investigate during the 1995 census test whether the CensusPlus field operation can attain ex- cellent coverage in CensusPlus blocks without contaminating the regu- lar enumeration in those blocks. If substantial problems are identified, CensusPlus should not be selected as the field methodology for inte- grated coverage measurement in the 2000 census unless clearly effective corrective measures can be implemented within the research and devel- opment schedule. The panel's position on this issue is similar to the view expressed by the National Research Council's Panel on Decennial Census Methodology, convened prior to the 1990 census, which argued for balance between efforts to achieve a complete enumeration and efforts to improve the accuracy of census results through cover- age measurement methods (see Citro and Cohen, 1985~. The follow-up of households that do not return the mail questionnaire is one of the most costly and labor-intensive operations in the traditional census. Sam- pling during nonresponse follow-up offers the potential for saving hundreds of millions of dollars, but it also would increase the variability of population esti- mates, especially for small geographic areas. The panel recommends research, development, and testing of nonresponse follow-up sampling in the 1995 census test. On the basis of results of the 1995 census test, the Census Bureau should make a careful and thorough determination of the data quality and cost implica- tions of this promising approach to the long-standing problem that not everyone responds to the mail questionnaire. Recommendation 4.1: Sampling for nonresponse follow-up could pro- duce major cost savings in 2000. The Census Bureau should test nonresponse follow-up sampling in 1995 and collect data that allows evaluation of (1) follow-up of all nonrespondents during a truncated period of time, combined with the use of sampling during a subsequent period of follow-up of the remaining nonrespondents, and (2) the use of administrative records to improve estimates for nonsampled housing units.
6 COUNTING PEOPLE IN THE INFORMATION AGE Response and Coverage Improvement A second major panel finding is that problems of cost and differential cover- age should also be addressed directly at the primary response stage of the census. Ongoing research at the Census Bureau is focused on improving coverage within households by changing the census form. The panel strongly supports the con- tinuation of this research through and beyond the 1995 census test because of its potential for yielding a cost-effective means of improving the initial count and possibly reducing differentials in coverage. Recommendation 3.1: A program of research extending beyond the 1995 census test should aim to reduce coverage errors within households by reducing response errors (e.g., by using an extended roster form). This research should also evaluate the impact of these new approaches on gross and net coverage errors, as well as assess the effects on coverage of obtaining enumerations using different instrument modalities (erg., paper and computer-assisted) and different interview modes (e.g., paper instrument completed by household respondent and by enumerator). Recently tested improvements in census questionnaire format and imple- mentation procedures could increase mail response rates, thereby saving money on the follow-up of nonrespondents. Making greater use of alternate technolo- gies (e.g., telephones) and offering people multiple ways to respond (e.g., by distributing census questionnaires at public places) may yield additional cost savings. Early research suggests that changes in the fostering method could improve within-household coverage for certain groups within the population. Measures of gross census error will be important in evaluating the effective- ness of proposed methods for increasing census response, such as using new ways to develop household rosters, distributing unaddressed questionnaires, of- fering the option of telephone response, and using special "tool-kit" enumeration methods in certain small geographic areas. We believe that census methodology should strive to minimize not only omissions (that produce undercounts) but also erroneous enumerations (that produce overcounts). Aggressive research is needed to develop techniques to prevent erroneous or duplicate enumerations during a census with multiple response modes and new fostering procedures. Other census design components for example, continuous updating of the master address file and the use of administrative records-might also contribute to reducing the differential undercount. In spite of these actual and potential improvements, however, the panel also concludes that sampling and estima- tion in particular, as part of an integrated coverage measurement program- will be needed to further reduce coverage differentials to insignificant levels.
SUMMARY 7 OTHER KEY CENSUS DESIGN COMPONENTS Address List Development Virtually all fundamental design changes contemplated for the 2000 census depend on the existence of an accurate list of residential addresses. A geographic database that is fully integrated with a master address file is a basic requirement for the 2000 census, regardless of the final census design. Recommendation 2.1: The Census Bureau should continue aggressive development of the TIGER (topologically integrated geographic encod- ing and referencing) system, the Master Address File (MAF), and inte- gration of these two systems. MAF/TIGER updating activities for the 1995 census test sites should be completed in time to permit the use and evaluation of the MAF/TIGER system as part of the 1995 census test. The duplication of effort, cost, and complexity involved both within the Census Bureau to compile address lists for consecutive censuses and across other federal agencies and state and local governments including the U.S. Postal Service-suggests the value of creating and maintaining a master list of ad- dresses over the decade. A continuously updated master address file could serve as a national utility for the federal statistical system. Recommendation 2.4: The Statistical Policy Office of the Office of Man- agement and Budget should develop a structure to permit the sharing of address lists among federal agencies and state and local governments- including the Census Bureau and the Postal Servic~for approved uses under appropriate conditions. We note the distinction between address information and information that identi- fies or characterizes individuals or households associated with addresses; the use of the latter type of information appropriately requires stronger provisions to ensure privacy and confidentiality. Use of Administrative Records Administrative records are already an important source of data for many statistical programs, and they can play a much greater role in the future. They can be used to provide more frequent and timely small-area data at a relatively low cost, and existing files could prove especially useful if augmented with informa- tion about race and ethnicity. Legitimate statistical uses of administrative records within the federal statistical system should be facilitated, rather than hampered, by legislation and administrative rules. In particular, legislation governing the use of health care records should permit use of basic health care enrollment data
8 COUNTING PEOPLE IN THE INFORMATION AGE by the Census Bureau for the decennial censuses and for current population estimates and surveys. Recommendation 5.1: Legislation that requires or authorizes the cre- ation of individual record systems for administrative purposes should not create unnecessary barriers to legitimate statistical uses of the records, including important uses not directly related to the programs that the records were developed to serve. Preferably, such legislation should explicitly allow for such uses, subject to strong protection of the confidentiality of individual information. The panel urges Congress, in considering legislation relevant to health care reform, not to foreclose possible uses of health care enrollment records for the decennial cen- suses and other basic demographic statistical programs. The Census Bureau should pursue a proactive policy to develop expanded uses of administrative records in future censuses, surveys, and population esti- mate programs. Effective pursuit of such a policy will require establishment of a suitable organizational structure and adequate resources for research and devel- opment activities not tied directly to ongoing census and survey programs. It will also require that the Census Bureau work closely with program agencies in the development of new administrative record systems and modification of existing ones to improve their utility for statistical uses. The panel urges Congress and the Statistical Policy Office of the U.S. Office of Management and Budget to support such interagency cooperation. Recommendation 5.11: The panel urges the Census Bureau to adopt a proactive policy to expand its uses of administrative records, and it urges other executive branch agencies and Congress to give their sup- port to such a policy. Although the panel concurs with the judgment of the Census Bureau that an administrative records census i.e., a census that relies exclusively or primarily on records from administrative data systems to produce population totals- is not a feasible option for the 2000 census, we believe the possibility should be care- fully explored for the 2010 census. This program of research should begin immediately to permit, in conjunction with the 2000 census, a meaningful com- parison of the administrative records census approach to the traditional approach under full census conditions. Recommendation 5.7: During the 2000 census the Census Bureau should test one or more designs for an administrative records census in selected areas. Planning for this testing should begin immediately. A program of sustained research and development, with the cooperation of program agencies, is needed for a thorough evaluation of the administrative records census option, taking into account the current and potential content of
SUMMARY 9 administrative records, the coverage of the population, and the potential for achieving significant cost reductions in censuses. It should not be assumed that a census based primarily on administrative records must duplicate all major design features of a traditional version of the census. Rather, it should be assumed that changes are possible in: (1) census content, definitions, and reference dates; (2) the census statute (Title 13, U.S. Code) and other laws and regulations governing the conduct of the decennial census; (3) laws and regulations governing access to federal and state administrative records; and (4) the content of administrative records systems. Matching and Elimination of Duplicate Records Record linkage is the identification of records belonging to the same unit (i.e., a person, household, or housing unit) either within a single data set or across two different data sets. In decennial census applications, records are matched either to eliminate duplication or to pool information from multiple sources. Many census operations involve matching one list of records to another. Needs for record linkage arise when address lists and other administrative records are used, when people are given multiple opportunities to respond to the census, and when dual-system estimation is used as part of a coverage measurement program. Historically, an initial match has been performed by a computer algo- rithm, followed by clerical verification and resolution. Many of the innovative methods being examined in the 1995 census test would place greater demands on matching technology. Thus, improvements in the accuracy or efficiency of auto- mated record linkage will support the 2000 census design by increasing the capability to produce reliable results within time and budget constraints. The development and updating of an integrated MAF1IGER system will require automated address matching and geocoding at various stages. The distri- bution of unaddressed questionnaires, the opportunity to respond by mail or telephone, and the application of other special methods are likely to increase the potential for duplication in the census enumeration; matching is needed to deter- mine whether persons and housing units were enumerated more than once. The Census Bureau's recent research on fostering may lead to new approaches to ascertaining residency, which would present new complications in assigning people correctly to geographic areas. Matching may also be needed to obtain telephone numbers during follow-up of nonresponding households. Record linkage technology will also support the development of an adminis- trative record database for the 1995 census test sites. Procedures for integrated coverage measurement could involve the automated comparison of records from the census enumeration and independent operations. The Census Bureau will need solid capability for computer matching and elimination of duplicate records in order to perform all the above tasks in an accurate, timely, and cost-effective manner.
10 COUNTING PEOPLE IN THE INFORMATION AGE Recommendation 2.2: The Census Bureau should continue its research program on record linkage in support of the 1995 census test and the 2000 census. Efforts should include studies of the effectiveness of differ- ent matching keys (e.g., name, address, date of birth, and Social Security number) and the establishment of requirements for such components as address standardization, parsing, and string comparators. Existing record linkage technology should be tested and evaluated in the 1995 census test. Limits on the ability to eliminate duplicate records may prove to be the con~ol- ling factor with regard to the feasibility of many of the innovations under consid- eration for the 2000 census design. Methods for Hard-to-Enumerate Populations The legitimacy of the census depends in part on public perception that it fairly treats all geographical areas and demographic groups in the county. Fair treatment can be defined in either of two ways: by applying the same methods and effort to every area or by attaining the same population coverage in every area so that estimates of relative populations of different areas are accurate. The objective of the census is to measure population accurately above all, to calcu- late accurate population shares in order to apportion representation properly. Therefore, obtaining equal coverage clearly takes priority over using the same methods in every area. In fact, since experience shows that treating every geo- graphical area and demographic group in the sable way leads to differential coverage, the Census Bureau has a positive duty to use methods designed to close the coverage gap. The term tool kit refers to the collection of special methods-for example, team enumeration, "blitz" tactics, local facilitators with community knowledge and ties, and bilingual enumerators for hard-to-enumerate areas. The panel supports the continued research, development, and testing of methods for differ- ential treatment of subpopulations with the goal of reducing the differentials in census outcomes across these subpopulations. The panel recognizes that experi- mentation with tool-kit methods in the 1995 census test will be constrained by pressures to limit operational complexity. Nevertheless, the panel believes that efforts should be made to plan vanations in tool-kit application across sites or areas within sites to permit comparative assessments. Recommendation 3.6: In the 1995 census test, the Census Bureau should include a larger repertoire of foreign-language materials than those cur- rently available in Spanish (both written and audio). In addition, the Census Bureau should conduct more aggressive hiring of community- based enumerators (with due consideration of local concerns about the confidentiality of census responses) and should accommodate greater
SUMMARY 11 flexibility in the timing of enumeration by personal visit (i.e., permitting contact during evenings and weekends). Attention must be given to the yield from tool-kit methods in terms of numbers of people who are accurately counted- not simply to the cost of imple- menting these methods. Potential problems of erroneous enumeration must be assessed during the testing and evaluation of tool-kit methods and other coverage improvement programs, and decisions about inclusion in the 2000 census should consider the relative marginal costs and benefits of these programs. A key question concerning the ability to systematically apply differential treatments to different subpopulations relates to the effective role of the planning database being developed by the Census Bureau for use in the 1995 census test. The successful use of tool-kit methods in census operations will depend on know- ing where they should be applied and on being able to apply them without creating erroneous enumerations. To the extent that the country can be stratified into easy-to-enumerate and hard-to-enumerate areas before the census, this knowl- edge can be incorporated into the methods used to produce a one-number census. The general issue of prespecification versus real-time adaptation is relevant not only for application of the planning database or a formal targeting model to improve initial census response, but also for differential treatment during non- response follow-up and integrated coverage measurement. In these latter opera- tions, there may be value in oversampling certain geographic areas that are known in advance to have coverage problems. Counting hard-to-enumerate populations may also be facilitated by means other than special enumeration methods. A current legislative proposal to change the census reference date from April 1 to the first Saturday in March could alleviate problems in counting mobile households, college students, and persons with no usual residence. A greater shift toward the middle of the month could further reduce end-of-month moving problems, although the ability to complete all phases of the census mail operation within the same calendar month is also an important consideration. Recommendation 2.3: In view of the operational advantages that are likely to result, the panel endorses the proposed change in census refer- ence date from April 1 to the first Saturday in March. Furthermore, we recommend that changing the census reference date from early in the month to midmonth (e.g., the second Saturday in March) be reconsid- ered if subsequent modifications to the mailout operation would permit all census mailings to be executed within the same calendar month using a midmonth reference date. The ethnographic research sponsored by the Census Bureau in recent years has provided knowledge about the problems of enumerating inner-city and rural low-income populations, immigrants, internal migrants, and homeless people
2 COUNTING PEOPLE IN THE INFORMATION AGE ("persons with no usual residence"~. The panel supports further comparative study of hard-to-enumerate areas and the application of existing findings to the development of cost-effective methods for counting these populations. Recommendation 3.5: The Census Bureau should conduct further com- parative studies of hard-to-enumerate areas, focusing on those parts of the country where three phenomena coincide: a shortage of affordable housing, a high proportion of undocumented immigrants, and the pres- ence of low-income neighborhoods. Studies of hard-to-enumerate areas may also inform census outreach and promotion efforts. The panel believes that more effective outreach and promo- tion efforts will require a structured research and development program and greater centralization of responsibilities for decennial census outreach and pro- motion. Use of the Telephone The telephone can and should play a much larger role in the 2000 census than it did in 1990 and in previous censuses. New technologies will allow expansion of the 800 number call-in assistance program and permit access to a wide range of automated services from any telephone in the United States. The panel supports the Census Bureau's plans to develop improved capabilities for handling incom- ing calls particularly in view of the potential negative effects on public percep- tion and response of being ill-prepared to field questions. Recommendation 3.2: The Census Bureau should use the 1995 census test and subsequent tests to inform the design of the 800 number call-in system for the 2000 census. The Census Bureau should focus on the public's response to the menu-driven call-routing system, acceptance of the computer-administered interview, possible differential mode effects between a computer-administered interview and one administered by an interviewer, and the technical feasibility of administering interviews using voice recognition and voice recording. The Census Bureau should also develop and implement a monitoring system in these tests to collect operational and cost data on the call-in program. The use of computer-assisted outbound calling will be possible in the 2000 census because of the availability of electronic directory services that can match telephone numbers to addresses. The Census Bureau will be able to add tele- phone numbers to the master address file for a significant number of address listings, and this resource can be used to make outbound telephone calls both to prompt mail nonrespondents to retune their forms and to complete the enumera- tion by telephone. Computer-assisted telephone interviewing will also play a key
SUMMARY 13 role in the integrated coverage measurement program being tested as part of the 1995 census test. Recommendation 3.3: The Census Bureau should expand the research program involving the acquisition of telephone numbers for MAF ad- dresses by working with more companies that offer electronic directory services and developing an optimal protocol for matching addresses. If the Census Bureau is able to acquire unlisted telephone numbers for a 1995 census test site, it should carefully monitor the results obtained from calling households with unlisted numbers. Findings in the survey research literature and in technology assessments conducted for the Census Bureau suggest that these telephone applications offer the potential for considerable cost savings and improved data quality. The panel encourages the Census Bureau to continue development of telephone-based meth- ods for testing in 1995 and, if successful, for adoption as part of the 2000 census. ALTERNATIVE METHODS FOR LONG-FORM DATA COLLECTION As noted above, there is growing interest in developing capabilities to obtain information about small geographic areas and subpopulations more frequently than every 10 years. In addition, there is a perception within Congress and among other interested parties that the additional content gathered by the decen- nial census long form places a severe burden on respondents and negatively affects census cost, coverage, and quality. These issues and, in particular, the accuracy of this perception are being addressed more thoroughly by the National Research Council's Panel on Census Requirements in the Year 2000 and Beyond. Nevertheless, we have reviewed two possible alternatives to the decennial long form, both currently being inves- tigated by the Census Bureau, for collecting certain types of information beyond simple population counts: (1) the use of multiple sample forms in the decennial census (including the application of matrix sampling) and (2) a large, continuous, monthly survey (the so-called continuous measurement option). The panel's consideration of these alternative methods has not been moti- vated by the concerns about cost and differential coverage that have been promi- nent in the movement for reform of the decennial census. Rather, the panel believes that the transfer of information-gathering responsibilities from the de- cennial long form to an alternative, such as a continuous measurement survey, should be based on judgments about the cost-effectiveness of these methods in meeting the current and future information needs of census data users. Toward that end, the panel supports continued research and development of a continuous measurement program as a potential future source of sociodemographic data for
4 COUNTING PEOPLE IN THE INFORMATION AGE small areas and small populations, recognizing that significant issues must be addressed before such a program can be seriously considered for adoption. Recommendation 6.1: The panel endorses further research and evalua- tion of a continuous measurement program. In conducting this work, the Census Bureau should establish, and continually reinforce, a com- mitment to simultaneous research and development of cost estimation, data collection and processing methods, estimation procedures, and user needs. Matnx sampling refers to a technique designed to spread and reduce respon- dent burden by dividing a survey instrument into multiple instruments with par- tially overlapping contents. On the basis of its examination, the panel finds that the conditions favorable to use of matrix sampling are either unlikely to be obtained or have not been well studied in the context of the decennial census long form. The panel therefore believes that matrix sampling is unlikely to present an effective alternative to long-form data collection in 2000. Recommendation 6.5: The panel endorses the Census Bureau's plan to investigate the impact of form length and content on mail response rates in the 1995 census test. Even if the operational feasibility of multiple sample forms is confirmed in the 1995 census test, the Census Bureau should not introduce matrix sampling without undertaking further re- search. Such research should be assigned low priority relative to other decennial census research projects.