The obligation for the federal government to conduct a census or decennial enumeration of the population for the purpose of apportioning the Congress is part of the Constitution. But the census provides the nation with essential data that go well beyond a simple population count. The information available from the census is widely used by governments at all levels for planning and administrative purposes, by business firms, by academic researchers, journalists, educators, writers, and community and nonprofit organizations. Although much valuable information about the U.S. population—race, income, employment, housing conditions, and other social and economic data—is collected by periodic surveys and from analysis of administrative records, only the census provides data covering a wide range of population characteristics for small geographic areas and small population groups. Census information is needed by the federal government for reapportionment of seats in the Congress, and by states for drawing congressional districts that meet the requirements of the Voting Rights Act. At the federal level, Congress has enacted laws requiring the use of census data to apportion federal funds to states, cities, school districts, and other governmental units under a wide variety of statutes, and they are used by many federal agencies to design, administer, and evaluate their programs.
The decennial censuses are planned and conducted with a high degree of professional competence. Nevertheless, some important problems have emerged over the past several decades. First, after decreasing from 1940 to 1980, the overall net undercount of the population increased in 1990 (to 1.8 percent, up from 1.2 percent in 1980). Moreover, in 1990, the difference in the net undercount between the black and nonblack populations rose to 4.4 percentage points,
the highest level since 1940 (the earliest census for which estimates of this type are available). Differential undercounts for blacks, Hispanics, and others cause concern because of their implications for such important uses of census information as legislative redistricting and fund allocation. Second, the existence of an initial census count and a subsequent and separate estimate from a postcensal evaluation survey for 1990 gave rise to divisive political arguments and legal suits over which set of numbers would be the official ones to be used for these important purposes.
Third, the costs of taking the census have escalated sharply, even after allowing for inflation and population growth. One of the reasons for that cost escalation, disturbing in its own right, has been a drop in the percentage of households who cooperate by filling out and returning the mail questionnaire. The 1970 decennial census cost $231 million. If adjustments are made to reflect the rise in prices and wages and the growth in the number of households between 1970 and 1990, with a further adjustment to take account of the increased costs associated with the decline in the mail response rate, the cost of conducting a 1970-type census in 1990 would have been about $1.3 billion. The actual cost of the 1990 census was $2.6 billion, twice as much. Although it has been argued that a major cause of the cost increase was due to costs associated with the long form (the longer questionnaire sent to 1 in 6 housing units in 1990), the panel's analysis of the relationship of content and cost led us to conclude that the content of the census did not produce the added costs. More generally, the panel has found that moderate changes in content do not have a significant effect on census costs.
In spite of considerable effort, the panel was unable to pin down, item by item, the causes of the cost increases. The panel believes, however, that a major part of the cost increase was driven by the response of the Census Bureau to several outside pressures. Most important, there was an increased and politically powerful demand for accurate population counts for very detailed geographical subdivisions and in hard-to-enumerate areas. Simultaneously, public cooperation with the census process, as measured by mail response rates, declined and was lowest in precisely those areas for which the pressures for an accurate count were greatest. The Census Bureau responded by pouring on resources in highly labor-intensive efforts to attempt to count every last person.
OVERALL CONCLUSIONS AND RECOMMENDATION
The panel's first task was to investigate whether and to what extent various types of essential data can best be collected by the decennial census or by other means. Our second task was to consider and recommend the most cost-effective methods of conducting the census and otherwise collecting census-type data. We evaluated a wide range of methods for meeting the requirements of the decennial census, including radical proposals that would sharply alter the way
the data are collected, substantial changes in the context of the traditional census, and incremental changes in the census.
The basic recommendation of this report follows from four central conclusions:
It is fruitless to continue trying to count every last person with traditional census methods of physical enumeration. Simply providing additional funds to enable the Census Bureau to carry out the 2000 census using traditional methods, as it has in previous censuses, will not lead to improved coverage or data quality.
It is possible to improve the accuracy of the census count with respect to its most important attributes by supplementing a reduced intensity of traditional enumeration with statistical estimates of the number and characteristics of those not directly enumerated.
Once a decision is made to use statistical estimation for completing the count, a thorough review and reengineering of census procedures and operations could achieve substantial cost savings in the next census, even as accuracy is being improved.
With regard to proposals to drop the long form in the next decennial census and substitute a continuous monthly survey to obtain relevant data, substantial further research and preparatory work are required to thoroughly evaluate the likely effect and costs of these proposals. Continuous measurement deserves serious consideration as a means of providing more frequent small-area data; however, the necessary research and evaluation cannot be completed in time for the 2000 census. Therefore, the 2000 census should include the long form.
Our analysis leads to one overarching recommendation for a substantially redesigned census in order to contain costs, reduce error in the population count, and improve data quality. We believe that significant changes in the census, which the Census Bureau in large part is planning to test, can achieve significant improvements in the data and at the same time make it possible to realize significant reductions in costs.
A REDESIGNED CENSUS
The panel concludes that physical enumeration or pure ''counting" has been pushed well beyond the point at which it adds to the overall accuracy of the census. Moreover, such traditional census methods still result in a substantial undercount of minority populations. Techniques of statistical estimation can be
used, in combination with the mail questionnaire and a reduced scale of follow-up of nonrespondents, to produce a better census at reduced costs.
Efforts to follow up individually those who fail to return the mail questionnaire should be simplified and truncated after a reasonable effort based on several criteria (see Chapter 5), and statistical sampling should be used to estimate the number and characteristics of the nonrespondent households that remain. In addition, evaluation surveys should be undertaken to improve the overall count and reduce the differential undercount.
Statistical estimation techniques have long been used in the census for a number of purposes, including close-out procedures, determining vacant units, and imputation of missing census responses. The use of statistical methods for sampling nonrespondents and surveys to complete the count is formal recognition that modern statistical procedures can improve the process, reduce costs, and produce better data for the country as a whole and for large areas and population groups by reducing the differential undercount. (It is also true, however, that the use of statistical techniques would increase the variability for small areas somewhat.) In our judgment, so long as a good-faith effort at physical enumeration is made, completing the estimates by statistical techniques would meet constitutional requirements.
The Census Bureau is planning large-scale tests in 1995 to help design the appropriate combination of enumeration and statistical estimation. The 2000 census should integrate enumeration and estimation to produce the best single estimate of the population and of the characteristics of the population, rather than two separate estimates—an initial count and a subsequent adjustment.
Recommendation 5.1 The panel recommends that the Census Bureau make a good-faith effort to count everyone, but then truncate physical enumeration after a reasonable effort to reach nonrespondents. The number and characteristics of the remaining nonrespondents should be estimated through sampling.
Recommendation 5.2 To improve the census results, and especially to reduce the differential undercount, the panel recommends that the estimates achieved through physical enumeration and sampling for nonresponse be further improved and completed through survey techniques. The census should be designed as an integrated whole, to produce the best single-number count for the resources available. The Census Bureau should publish the procedures used to produce its final counts, as well as an assessment of their accuracy.
Essential to the success of a redesigned census are efforts to improve the
response to the mail questionnaire and also the completeness of the address list to which the questionnaires are sent. The Census Bureau should continue to stress the mandatory nature of the census. An expectation that the census is not mandatory would affect adversely efforts to increase the mail response rate and would undercut the improvements otherwise obtainable through a redesigned census. The Census Bureau has already conducted a number of field tests of improved procedures, such as the use of respondent-friendly census forms, designed to improve mail response rate, lower costs, and raise accuracy; the results to date have been promising.
Recommendation 5.3 The panel recommends that the Census Bureau incorporate successfully tested procedures to increase the initial response rate in the 2000 census, including the use of respondent-friendly questionnaires and expanded efforts to publicize the mandatory nature of the census.
To improve the master address file and other components of census operations, the Census Bureau should develop partnerships with state and local governments and the U.S. Postal Service.
Recommendation 5.4 The panel recommends that the Census Bureau develop cooperative arrangements with states and local governments to develop an improved master address file and that Congress amend Title 13 of the U.S. Code to permit sharing of this file for statistical purposes with state and local governments and the federal statistical system.
Recommendation 5.5 The panel recommends that the U.S. Postal Service and the Census Bureau continue to work together to improve the decennial census. We endorse an expanded role for the U.S. Postal Service in the 2000 census in several areas: (a) development, maintenance, and improvement of an accurate address file for the nation's residential housing units, (b) checking the address list prior to the census to improve accuracy, (c) delivery of the mailed forms, and (d) ascertainment of the vacancy status of housing units during the census.
On the basis of preliminary estimates provided by the Census Bureau and our own analysis, we judge that the use of the specific cost-savings techniques that we recommend—principally, truncation and sampling for nonresponse follow-up—would
have reduced the costs of the 1990 census by $300 to $400 million. But, as noted earlier, $1.3 billion of the cost increase in the census between 1970 and 1990 is not explained by rising prices, increased numbers of households to be counted, and the effects of the decline in the mail response rate. We believe that these cost increases stemmed principally from efforts to meet strong pressures for highly accurate counts of detailed area and population group by means of highly intensive operations seeking to make a physical count of every last person. This view leads us to believe it is possible to achieve additional cost savings—beyond the $300 to $400 million—by means of a thorough-going "reengineering" of the entire census process in the context of the new census design in which statistical estimation is integrated with direct enumeration into a single-number census.
If censuses in 2000 and beyond rely on statistical methods to improve population coverage, then all procedures for enumerating the population need to be reconsidered in this new context. Such a fundamental rethinking of the census has the greatest potential for cost reductions. A thorough reexamination should consider issues of data collection, data processing, and data dissemination, such as: How many district offices are needed? How long do they need to be open? What staffing and supervisory personnel are needed? Has automation improved the census process commensurate with cost increases? How much of the product line can shift from print to electronic media? Given the use of statistical estimation to complete the count, which coverage improvement operations can be scaled back or eliminated?
As one means of undertaking a serious and successful reengineering effort, we recommend that the Census Bureau develop a plan for conducting the new "integrated single-number" census in 2000 that would eliminate a substantial fraction of the $1.3 billion increased costs between 1970 and 1990, with a target for cost savings much larger than the $300 to $400 million already identified. The Census Bureau should identify what would have to be done to meet a target for substantial cost savings and spell out the consequences both for its operations and for the quality of the resulting census.
Since we have not been able to pinpoint the specific components of operations and procedures that led to the cost increase from 1970 to 1990, we cannot predict the full magnitude of the cost reduction that would be feasible while still producing a high-quality census. But we are convinced that a good-faith effort to develop such a plan would identify important cost savings that could be made without a significant sacrifice of quality. These savings could result from reengineering all census operations in the context of a single-number census in which enumeration and statistical estimation are fully integrated.
Recommendation 5.6 We recommend that the Census Bureau undertake a thorough reexamination of the basic structure, organization, and processes by which the decennial census is conducted to
obtain the full cost-saving potential of the proposed redesigned census. As one part of its reexamination, the Census Bureau should develop a plan for the 2000 census that eliminates a substantial fraction of the $1.3 billion cost increase (in 1990 dollars) from 1970 to 1990 that is not accounted for by the growth in housing units and the decline in the mail response rate. The target for this plan should be much more than the $300 to $400 million we have already identified.
Support for the New Design
The Census Bureau should settle expeditiously on a new census design that contains a reasonable rate of direct enumeration combined with greater reliance on statistical estimation. Such a design can produce not only lower costs, but also improved accuracy for the single most important variable: the count of population, particularly by race and ethnicity. Because people not familiar with statistical sampling and survey techniques may be alarmed by some of the characteristics of a census that relies more heavily on sampling and statistical estimation, the Census Bureau should undertake a sustained program of public education about this plan well in advance of the 2000 census.
Recommendation 5.7 The panel recommends that the Census Bureau publicize, as soon as possible, its plan for implementing a redesigned census. It should then move to obtain early commitment for the new plan from both Congress and data users.
CENSUS CONTENT AND THE LONG FORM
Historically, the census has collected additional content beyond the minimum set of items needed from all households for the constitutional purposes of reapportionment and redistricting. Since 1960, most of the additional data have been collected on a separate long form sent to a fraction of households (most households have received a short form). These additional data are widely used and serve many important public purposes. The nation needs the breadth of information for small areas and small population groups that these census data now provide.
Conclusion 6.1 The panel concludes that, in addition to data to satisfy constitutional requirements, there are essential public needs for small-area data and data on small population groups of the type and breadth now collected in the decennial census.
The process for determining the content of the census (both the short and the long forms) has involved bringing together federal agencies and balancing their data needs against considerations of questionnaire length and feasibility. Views of other data users have also been sought. This process has worked well and should be maintained and strengthened.
Conclusion 6.2 The panel concludes that the process of determining the census content by involving federal agencies and eliciting the views of other users has worked well in the past and should be continued. In our judgment, the process would be strengthened by increasing the oversight and coordination role of the chief statistician in the Office of Management and Budget.
In 1990, the long form—sent to approximately 1 in 6 households, or about 15 million of the nation's 92 million households—had 58 items in addition to the 13 on the short form. Its length and complexity have led to some suggestions that dropping the long form would significantly improve the census mail response rate and substantially lower costs. It has been variously suggested that much of the lost data could be replaced by the use of administrative records or by a large-scale and continuing sample survey of households.
Our work does not support these suggestions. In our judgment:
The long form has not been responsible for the decline in response rates or the increase in costs in the last several censuses: the number of questions has been roughly the same for a number of decades, and the sampling rate has declined as a percentage of households.
Available evidence indicates that the overall effect on response rates of dropping the long form would not be large (about a 1 percentage point increase) and, in any event, would apply only to the 1 in 6 households that currently receive the long form. The evidence also suggests that the long form does not increase the undercount or adversely affect the differential undercount by race to any significant extent.
Given the necessity of conducting a decennial census, the inclusion of a long form questionnaire for a large sample of households is a cost-effective way of obtaining highly valuable information.
Conclusion 6.3 The panel concludes that the marginal cost of incremental data on the decennial census is low. In particular, we conclude that the extra cost of the census long form, once the census has been designed to collect limited data for every resident, is relatively low.
One proposal to reduce the burden of the census long form on each household that receives it is to spread out the long-form content over several, somewhat shorter forms, in an approach called matrix sampling.
Recommendation 6.1 The panel recommends that the Census Bureau evaluate the merits of a matrix sampling approach that uses several intermediate forms in place of a single long form to reduce respondent burden. The Census Bureau should examine the effects on:
satisfying data users' needs;
mail return rates;
sampling and nonsampling errors (including item nonresponse rates);
operational problems; and
data processing and estimation problems that could affect the usefulness of the information, particularly for multivariate analysis.
Another approach to reducing respondent burden, which the Census Bureau has recently been investigating, is to drop the long form from the census and substitute a continuous measurement survey—that is, a large monthly survey of perhaps 200,000 to 500,000 households. By averaging the results of the monthly surveys over a period of 3 to 5 years, more timely long-form-type data, accurate enough for use in relatively small geographic areas, could be produced. At the state level and for large metropolitan areas, estimates could be made by averaging over much shorter periods. In its preliminary work, the Census Bureau has speculated that the costs of the new continuous measurement survey over a decade could be roughly offset by the cost savings from dropping the long form from the census and by other cost reductions that might be achieved in intercensal operations.
The panel has followed the ongoing plans of the Census Bureau with great care and interest. Although we believe that the proposed continuous measurement system deserves serious evaluation, we conclude that much work remains to develop credible estimates of its net costs and to answer many other fundamental questions about data quality, the use of small-area estimates based on cumulated data, how continuous measurement could be integrated with existing household surveys, and its advantages compared with other means of providing more frequent small-area estimates. In our judgment, it will not be possible to complete this work in time to consider the use of continuous measurement in place of the long form for the 2000 census.
Conclusion 6.4 The panel concludes that the work to date on continuous measurement has overestimated the savings from dropping the long form, understated the cost of a continuous measurement system, and not sufficiently examined feasible alternatives for meeting the nation's needs for more timely long-form-type data at reasonable overall cost. We conclude that it will not be possible to complete the needed research in time to make the critical decisions regarding the format of the 2000 census. We therefore do not recommend substituting continuous measurement for the long form in the 2000 census.
For the longer term, the Census Bureau's program of research on continuous measurement should be expanded and redirected in a framework that considers alternative cost-effective ways to meet the needs of data users for intercensal estimates and how efforts to that end can be integrated with the overall federal statistical system.
Recommendation 6.2 The panel recommends that the Census Bureau broaden its research on alternatives for more frequent small-area data to encompass a wider range than continuous measurement, as currently envisaged. In that context, the Census Bureau should examine the cost-effectiveness of alternatives, the ways in which they meet user needs, and the manner in which continuous measurement or other alternatives could be integrated into the nation's system of household surveys. The research program should be carried out in cooperation with the federal statistical agencies that sponsor household surveys and should include evaluation of the quality of important data elements, the frequency and modes of data collection, and the manner in which results would be presented, as well as methods for introducing change over time.
Summarizing our conclusions about the content of the 2000 census, we see the long form as a cost-effective means of collecting needed data, and we conclude that there is no feasible alternative to the long form for 2000.
Recommendation 6.3 The panel recommends that the 2000 census include a large sample survey that obtains the data historically gathered through a long form.
DATA ON RACE AND ETHNICITY
Historically, the decennial census has included questions on race and ethnicity and provided classifications for respondents to check. The number and
character of the classifications have changed over time as social attitudes have changed, as changes have occurred in the racial and ethnic makeup of the population, and as the needs for race and ethnicity data by governments and society have evolved. The short-form questions on race and ethnicity in the 1990 census included 15 categories or subcategories listed in the question on race and 4 categories in the question about Spanish/Hispanic origin.
In recent decades these data have become particularly significant. The growing public concern about problems related to racial and ethnic discrimination has emphasized the need for accurate demographic data on race and ethnicity. The Voting Rights Act of 1965 (as amended), together with court interpretations, has substantially expanded the need for race and ethnicity data at the small-area level for congressional redistricting and other purposes. At the same time, the disturbing growth in the differential undercount in the 1990 census for ethnic and racial minorities has called attention to problems with the demographic data on race and ethnicity collected by the census.
At present, there are growing but conflicting public pressures for enlarging and changing the race and ethnicity classifications used in the census. The census form provides for self-identification of race and ethnicity by respondents. But the presumption that only one classification per respondent is appropriate is being challenged by the growing heterogeneity of American households as a result of rising immigration, increased interracial and interethnic marriages, greater numbers of people with multiple racial identities, and expanded recognition in recent decades of the fluidity of race and ethnicity classifications. The design of future censuses must take these changes into account.
Conclusion 7.1 The panel concludes that there are inherent ambiguities to racial and ethnic status. We recognize that reporting of this status in the census has multiple meanings for respondents and that the context of reporting affects how people self-identify.
Conclusion 7.2 The panel concludes that the census questionnaire must be developed as a well-tested means for obtaining nationally useful data on race and ethnicity to meet constitutional and legally mandated federal requirements and for other legislative and program needs and informational purposes.
Recommendation 7.1 The panel recommends that the Census Bureau expand its examination and testing of race and ethnicity questions to provide comprehensive information on: (1) public understanding of the concepts and acceptability of questions, (2) compatibility among the several census items and the utility of cross-tabulations, (3) the comparability of census data to race and ethnicity data collected in other federal surveys or obtained from
administrative records, and (4) the quality of data for small areas and specific groups. This research needs to be given a high priority so that the results may be incorporated into the review of Statistical Directive 15 currently being conducted by the Office of Management and Budget.
Recommendation 7.2 The panel recommends that the Office of Management and Budget issue a revision of Statistical Directive 15 sufficiently early to provide adequate time for planning and testing for the 2000 census and coordinated implementation of changes by all affected agencies.
SOME RADICAL ALTERNATIVES TO THE CENSUS
The panel evaluated a number of proposals for radically different ways of collecting the data now provided in the decennial census before deciding to recommend a redesigned census combining traditional enumeration with greater use of statistical estimation. (These alternatives are discussed in Chapter 4 of the report, prior to the discussion of the redesigned census in Chapter 5 and census content issues in Chapters 6 and 7. This Summary reverses the order of the discussion.)
There are three critical requirements for census data to which the panel attached great weight in assessing alternative data collection methods. First, for constitutional, legislative, and many other legitimate reasons, households, people, and their characteristics must be associated with specific geographic locations—with residential addresses. Second, and of particular importance since the enactment of the Voting Rights Act, data are required on the racial and ethnic identification of people, again associated with residential addresses. Third, there is a range of critical needs for small-area data that include information on a wide variety of population characteristics.
A National Register
A population register is a special type of administrative record, a system maintained by the mandatory continuous registration of all residents. If each individual is issued an identification number (like a Social Security number), a register can be cross-referenced to administrative records to provide a wide variety of data linked to addresses. To be useful for census-type data, people would be legally required to update their addresses at periodic intervals with the governmental unit that keeps the register.
A national register has the advantage of timely data availability (more frequently than once every 10 years). If the cost of operating a population register
is absorbed by other administrative uses, then the additional cost of using the data for statistical purposes may be relatively low; however, the United States has no such system in place. Moreover, proposals to create a register would raise privacy and civil rights concerns almost surely sufficient to make it unacceptable in the American culture, and, without a high degree of public cooperation, a national register would produce low-quality data.
Conclusion 4.1 The panel concludes that a national register for the United States is not a feasible replacement for the decennial census.
An Administrative Records Census
There are three potential ways in which administrative records—such as records of the Social Security Administration, the Internal Revenue Service, and the vital statistics systems of state and local governments—can be used to supplant or supplement the decennial census: (1) administrative records could substitute for the entire census; (2) a very simple census could be taken with only a few basic questions and other necessary census-type information obtained from administrative records; and (3) administrative records could be used in the taking of the regular census—for example, by helping to find hard-to-reach population groups—and to improve the quality of intercensal data. Only the first two options represent radical alternatives to the conventional census because administrative records are already used to assist in the census.
The possible use of administrative records to take the place of all or a major part of the census is hampered by several factors. No single set of administrative records is complete enough for the data needed, so that large-scale linkage of records from different systems would be likely to be required, leading to increased costs and privacy problems. Many sets of records (e.g., the bulk of Social Security files) lack up-to-date addresses. There would be political, social, and statistical reliability problems inherent in requiring that ethnicity and race data be supplied as part of administrative records. It would be very difficult to ''unduplicate" administrative records when data from several different record systems are combined without the availability of a comprehensive and uniform personal identification number. Nevertheless, the use of administrative records in the census can be expanded, and we suggest ways in which this can be done.
Conclusion 4.2 The panel concludes that an administrative records census is not a feasible option for 2000.
A Census Conducted by the U.S. Postal Service
Since the U.S. Postal Service regularly visits the vast majority of residences in this country, some have suggested that it be given the task of collecting decennial
census data by delivering a questionnaire to every residence and conducting the basic census enumeration, with the Census Bureau responsible for designing the questionnaire and tabulating the results. This suggestion does not deal with the fact that the most difficult and costly task of the census is collecting information from those who have failed to return the mail questionnaire, either through direct follow-up visits or by survey techniques. For neither of these tasks is the Postal Service a suitable or cost-effective vehicle. The hourly costs of letter carriers are more than three times the hourly costs of census enumerators in the 1990 census. A joint study by the Census Bureau and the U.S. Postal Service concluded, for a variety of reasons, that it would not be practical for letter carriers to replace census enumerators.
Conclusion 4.3 The panel concludes that the decennial census cannot be conducted by the U.S. Postal Service in any way that would be cost-effective or an improvement over the conventional census.
The U.S. Postal Service and the Census Bureau can greatly assist each other in census-related activities, and we make specific recommendations to foster that cooperation.
A Sample or Rolling Census
Two sample survey approaches are possible for completely replacing the conventional census. One is the use of a large sample survey to produce an estimate of the total U.S. population and its characteristics at a single point in time. Some have referred to such a large survey for estimating the population as a sample census. The other approach is a rolling census, which entails the complete enumeration of the population in such a manner that the sample enumerated at any given time is a random sample of the whole population. One option for a rolling census is an annual 1-in-10 sample survey designed to enumerate the entire population over a 10-year period.
The key challenge to a sample census is that the Constitution calls for a complete count of the population at one point in time. Also, by all the available evidence, household surveys experience higher net undercoverage rates and more severe differential undercoverage than does the complete-count census. Furthermore, a sample census would require the development of an accurate sampling frame (e.g., a complete list of all dwellings in the United States) from which to draw the sample. For a rolling census, the cost is likely to exceed the cost of a conventional census. Finally, public acceptability of either a sample census or a rolling census is likely to be low because both lack the high-profile, attendant publicity, and historic image of the decennial census.
Conclusion 4.4 The panel concludes that a rolling or sample census is not an acceptable replacement for the 2000 decennial census.
INTERCENSAL SMALL-AREA DATA
The availability of accurate demographic and related data for small areas—cities, counties, school districts, neighborhoods, and other small geographic areas—is most important for the efficient planning and operation of many federal, state, and local government activities, for large numbers of business firms, and for research in many areas. A wide variety of government agencies has come to place great reliance on small-area data, and the decennial census is a key source for such information. Not only is it used directly, but it can also be combined with more current data from other sources to produce additional information. The census data themselves, however, suffer from the disadvantage of progressive obsolescence as people move and economic conditions change over the 10 years between censuses. In a mobile population such as that of the United States, this obsolescence is particularly rapid.
Although the enhancement of current annual surveys might, at some additional cost, provide more timely estimates for larger geographic areas (states, metropolitan areas, and populous counties and cities), the cost of surveys large enough to supply annual data for small areas would be prohibitive. We commented above on the proposal to substitute a continuous measurement survey for the decennial long form that would produce a 3- or 5-year moving average of small-area data. So far there has been little support from small-area data users for such a move. Still, the Census Bureau, in the redirected research program that we recommend, should carefully explore the response of users to the proposal.
The most cost-effective way of improving the availability of intercensal small-area data is likely to be the expanded use of federal, state, and local administrative records to supplement decennial census data. By conducting experiments with the use of administrative records, the Census Bureau could gradually build a reservoir of knowledge about the potentials and the limitations of such records and the modifications in records that would be needed to make them useful for statistical purposes. Efforts along this line would strongly benefit from the expansion of mutually cooperative arrangements among the Census Bureau, federal agencies, and state and local governments.
Recommendation 8.1 The panel recommends that the Census Bureau work to improve the amount, quantity, and frequency of small-area intercensal data:
The Census Bureau should conduct experiments with federal administrative records for deriving more frequent small-area intercensal data estimates. At a minimum, the panel recommends that the Census Bureau geocode several large federal administrative record systems and use them to produce small-area estimates.
The Census Bureau should work with state and local governments to enhance the quantity and frequency of small-area data.
Recommendation 8.2 The panel recommends that the Census Bureau give a single unit sole responsibility to exploit administrative records and produce small-area intercensal estimates on a frequent basis. Its work on administrative records should examine geographic consistency and quality. The unit should develop methods for increasing geographic content; establishing consistency of federal, state, and local administrative data; augmenting content on national records; augmenting usefulness of the resulting information through modeling; and computerizing approaches to database management to facilitate the use of administrative data in a census. If the content of administrative records can be improved for use in preparing small-area estimates, that is desirable, but the major purpose of the unit would be to produce small-area intercensal estimates.
In support of each recent census, the Census Bureau created a master file of residential addresses and, in 1990, linked that file to a geographic reference system (called TIGER), which provides map locations for physical features including roads, streets, and address ranges. The panel has been informed that the Census Bureau and the U.S. Postal Service are well along in the development of a cooperative arrangement under which the vast bulk of the master address file for the census and the associated TIGER system would be kept continuously updated. On the basis of preliminary estimates provided by the Census Bureau, the 10-year cost of keeping these files updated is, to a first approximation, no higher than the cost incurred each decade to build them from scratch.
The updated address files would be a major tool that state and local governments could use, in combination with decennial census data and currently available information from administrative records and other sources, to improve small-area data during intercensal periods. At the same time, such local uses of the address file would greatly contribute to its completeness. An address file that is complete and accurate and geographically referenced to small areas will also support the redesigned census, with its increased reliance on sampling and statistical estimation to determine the number and characteristics of nonrespondents. To make possible an accessible continuously updated national address file, politically acceptable changes in the laws and regulations governing census information would need to be worked out in order to make the address file (without information about residents) available to cooperating state and local governments for statistical purposes only.
The Census Bureau should examine the cost and coverage of a continuously updated national address file and geographically referenced database. The examination
should evaluate its use as a sampling frame for large federal household surveys, its use for geographically referencing administrative records for small-area intercensal estimates, and possible savings that would offset its costs.
Recommendation 8.3 If the cost estimates for continually updating the master address file and associated geographic-referencing database (including costs by the Census Bureau and others) are comparable to the cost of one-time updating just prior to the census, the panel recommends that development proceed. If the cost estimates are higher, then the clear advantages of the continuously updated address system should be weighed against the additional costs. If a decision is then made to continue, the Census Bureau should proceed with the necessary steps, including the necessary accompanying safeguards, to make the master address file available for statistical purposes to the federal statistical system and to cooperating state and local governments.