The Role of the Census
The United States will conduct its next population census in 2000—the twenty-second such census since the nation was founded. The Constitution of the United States, article I, section 2, specifically requires a decennial census to provide information for reapportionment of seats in the U.S. House of Representatives, which, by extension, also means providing information to redraw congressional district boundaries:
Representatives … shall be apportioned among the several States which may be included within this Union, according to their respective numbers.… The actual Enumeration shall be made within three Years after the first Meeting of the Congress of the United States, and within every subsequent Term of ten Years, in such Manner as they shall by Law direct.
Censuses in the nineteenth century were ad hoc operations, carried out initially by U.S. marshals and then by temporarily employed enumerators under the supervision of a census office that was set up anew and disbanded each decade. In 1902 an act of Congress established a permanent Bureau of the Census, now lodged within the Department of Commerce, that assumed responsibility for planning and supervising the census on a professional basis (see Anderson, 1988).
The modern census in the United States is a complex series of operations, involving a long lead time for planning and a multitude of steps in execution. The basic methodology for recent censuses includes the following steps (variations in the process occur in some instances: for example, census enumerators deliver unaddressed questionnaires to rural households—see Appendix B):
the advance preparation of a master list of addresses;
the mailout of questionnaires to all addresses on the list with the request that each household complete the questionnaire and return it by mail;
the follow-up by enumerators for questionnaires that are not returned and to obtain answers to missing items (some of this follow-up is done by telephone);
a series of checks to strive for more complete coverage of the population;
computerized processing of the questionnaires and tabulation of the results; and
a program to evaluate the quality of the enumeration.
With regard to content, the U.S. census has almost from the very beginning obtained added information about the population beyond the absolute minimum required for reapportionment and redistricting. Since 1940, the census has also obtained information on the nation's housing stock. In the most recent census in 1990, about 104 million questionnaires were mailed out (or delivered by census enumerators).1 Five of every six of those questionnaires ("short forms") asked households to respond to 6 questions for each person in the household and 7 questions on the housing unit. The remaining one-sixth of the questionnaires ("long forms") asked households to respond to an additional 35 questions on the household members and 23 questions on the housing unit (see Appendix A).
In other words, the 1990 census, like every census since 1940, included a sample survey as part of the complete enumeration. The 1940 census was the first to have enumerators ask questions of a sample of the population; the 1960 census was the first to have the Postal Service deliver (unaddressed) questionnaires and initiated the use of separate, printed short and long forms (the latter containing additional items asked of a sample of households).
In many respects, the 1990 census was a very successful operation, particularly considering the difficulties inherent in attempting to enumerate a large and diverse population. It was, however, expensive—in both total dollars ($2.6 billion) and costs per housing unit ($25). Moreover, these costs exceeded the costs of previous censuses, even allowing for inflation, decline in mail response rates, and growth in the number of households. Like previous censuses, the 1990 census also had errors. Most prominent and politically sensitive—because of the implications for the distribution of legislative seats and funding for government programs—were differential coverage errors, in which some population groups and geographic areas were counted less thoroughly than others. Despite the increased amount of money spent on the 1990 census, there was a somewhat higher net undercount of the total population than in 1980, as measured by the technique of demographic analysis (1.8 percent compared with 1.2 percent), and a somewhat larger difference between the net undercount rates for minorities and whites.
Following the 1990 census there was growing concern that census costs
might continue to climb with no likelihood of gains in the completeness of coverage. Fueling this concern was the evidence of declining public cooperation with the census—in 1990, only 74 percent of households mailed back their questionnaire compared with a return rate of 81 percent in 1980, thereby adding costs to send out enumerators to follow up the nonresponding households. (The mail return rates cited refer to the percent of occupied households that sent back a questionnaire; the mailback rates for all housing units, including occupied and vacant, were lower.) To obtain an independent assessment of how best to address these concerns, Congress asked the Committee on National Statistics to establish the Panel on Census Requirements in the Year 2000 and Beyond. Our panel was asked to evaluate the need for the data in the census, to consider whether needed data are better collected in the census itself or by some other data collection system, and, in general, to identify changes to the census that could ameliorate the twin problems of coverage errors and high costs.
OVERVIEW OF THE REPORT
In the remainder of this introductory chapter, we summarize our findings and conclusions about users' requirements for census data and the extent to which these data serve important public policy purposes (appendices provide detailed reviews of various types of census data uses). We also outline the criteria—including implications for content, cost, quality of information, feasibility, timeliness, and public acceptability—that we used to evaluate proposed changes in census methods, including alternatives to the census itself for collecting needed content. Chapter 2 briefly reviews the history of coverage problems in the census, and Chapter 3 reviews in detail the costs structure of the 1990 census and the patterns and reasons for the marked escalation in census costs over the past few decades.
The next four chapters address issues of method and content for the 2000 census, with reference to our evaluation criteria. Chapter 4 considers the merits of radical alternatives that have been proposed to replace current census methodology (e.g., establishing a national population register). Chapter 5 considers fundamental reforms to the current census process, such as the use of sampling in nonresponse follow-up. Chapter 6 discusses issues related to the overall census content, including whether the long form, which goes to a sample of households, adversely affects the costs and coverage of the basic census and whether a system of continuing household mail surveys could replace the long form. Chapter 7 considers the needs for census data on the important content items of race and ethnicity and the problems associated with obtaining such data.
THE NEED FOR CENSUS DATA
Before we consider how best to conduct the census, or whether and how to replace the traditional census process in whole or in part with some other data collection system, we should be clear about the data requirements that a census (or a comparable alternative) must satisfy. We review below data needs for reapportionment and redistricting, which are constitutionally mandated, and other data needs, for which a rationale must be established.
Data Needs for Reapportionment and Redistricting
The bedrock purpose of the census is to provide data for legislative reapportionment and redistricting. Over the last few decades, legislation and court rulings have extended this basic constitutional mandate for the census. Court decisions implementing the "one-person, one-vote" principle have required the census to produce data that will support legislative reapportionment and redistricting for all levels of government with a very high degree of accuracy. The census, therefore, cannot merely count the population, but must determine precisely where people live. Also, the Voting Rights Act, as originally enacted in 1965 and continued and amended several times since, has required the census to produce not only population counts by age but also data on race and ethnic origin at a very fine geographic level of detail. Although the Voting Rights Act and court decisions do not specify the census as the source of such data, the census (or a comparable replacement for it) is the only practical source (see Appendix C).
The legal requirements for census data for purposes of legislative reapportionment and redistricting determine several key features that a census must have. First, the census must include an attempt at a complete enumeration of the entire country at one point in time. Second, the census must provide basic data (e.g., age and race) at the smallest possible geographic level, namely, the census block. Data at the block level are needed not so much for themselves as to permit legislative districts to be drawn by combining blocks to meet court-mandated criteria for equal populations across districts and appropriate representation of minority groups under the Voting Rights Act.
Other Data Needs
Given the constitutionally grounded requirement for a census with such basic information as age, race, and sex (these items are necessary as well to ensure that people are counted once and only once), what additional content is needed? The census is unique among the statistical activities of the federal government in that it provides data on a range of topics that are reasonably accurate for small geographic areas and also small subgroups of the population
(e.g., small ethnic and occupational groups). The federal government sponsors many surveys that provide more in-depth information on a particular subject (e.g., income, transportation, housing, educational attainment), but such surveys provide estimates only for relatively large geographic areas (e.g., regions, states, and metropolitan areas). Administrative records systems also provide useful information, but they typically pertain to a particular population group (e.g., recipients of social security and public assistance benefits) and have limited content. The census alone provides a broad range of information that can be cross-tabulated for small geographic areas (and small subgroups).
One question is whether all this information is serving important purposes. Another question is whether collecting all this information materially impairs the accuracy and increases the costs of obtaining the minimal set of items that is needed for the census's primary purpose. A third question is whether some other data collection system could better serve to obtain needed content or whether the census itself remains a cost-effective vehicle for this purpose. We address the latter two questions in Chapter 6. Here we consider the value of the additional content that is currently collected in the census.
Federal Agency Needs for Census Data
We begin by looking at the legal requirements for census content in federal statutes and agency regulations. It is dangerous to push too hard for a requirement that each and every item on the census must be legally mandated. Some items lacking legal mandates may be needed for such important purposes as improving census coverage. Also, the determination of the length and content of the census questionnaire is not well served by requiring agencies to seek statutory approval for each item. Such a process could result in an increase in census content and, in any event, should not substitute for a decision process that makes deliberate trade-offs among agency needs, considered all together, and that takes into account the changing needs for data to inform current and emerging policy concerns as well as those from the past.
Nonetheless, it is important to determine whether the census content is required to serve important federal policy and program purposes. We reviewed materials compiled by the Census Bureau and the Office of Management and Budget on the federal mandates for items in the 1990 census (see Appendix M). The results are quite clear: almost all census items are required by federal government agencies to meet specific mandates. For some items, legislation requires the item and specifies that it be obtained from the census. For other items, the legislation requiring the item does not specify the census as the source, although the census (or a comparable replacement for it) is the only practical source. For a few items, there is no legislation but a requirement generated by agency program operations.
Listed below are just a few of the many examples of legislatively mandated items (see Appendix M for others):
The Elementary and Secondary Education Act requires data on the poverty status of school-age children by school district to allocate federal compensatory education funds. School districts vary in size from entire cities or counties to small towns or even smaller districts. At the present time, only the census long-form sample provides the necessary small-area data on income for determining the proportion of poor school-age children in each district.
The 1975 and later amendments to the Voting Rights Act require the director of the Census Bureau to determine political jurisdictions (counties, municipalities, and townships) that must implement procedures for bilingual voting to protect the rights of language minorities. These determinations are currently made from long-form census data on mother tongue, citizenship, educational attainment, and English-language ability, together with information on age, race, and ethnicity.
Title 42, Section 1786, of the U.S. Code requires the Department of Agriculture Food and Nutrition Service to obtain census data by state and county to determine the number of women, infants, and children whose families have incomes below the maximum income limit for the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC).
Various sections of the U.S. Code require the Department of Agriculture Rural Development Administration to use census data on income for tracts and counties to allocate grant funds and determine loan interest rates for several assistance programs (e.g., the Emergency Community Water Assistance Program and the Water and Waste Disposal Loan and Grant Program).
Just a few of the other applications of census data by federal agencies that are based on authorizing legislation include:
The Immigration and Naturalization Service uses census data for governmental units on place of birth, citizenship, year of entry, and other characteristics to prepare congressionally required reports and for program planning and evaluation purposes.
The Equal Employment Opportunity Commission uses census labor force data by zip code to analyze statistical evidence in class action charges of employment discrimination.
The Department of Transportation uses census data on disability for traffic analysis zones to monitor compliance with the Federal Transit Act and the Americans with Disabilities Act.
In addition, the census serves important functions in the federal statistical system. Some of these functions, such as providing denominators for vital rates
(e.g., birth and death rates) and reweighting household sample surveys to more accurately reflect the distribution of the population (such surveys invariably have higher coverage errors than the census itself), involve the basic census content on age, race, and sex. Other uses of census data by federal statistical agencies involve additional content items, for example:
The census address list serves as the sampling frame for many federal surveys, saving agencies the expense of developing such a frame (with the exception of areas of new construction). Often, such census content items as income and occupation are used to make the survey sample design more cost-effective.
Census data on income and other characteristics are used by the Bureau of Economic Analysis in developing regional, state, and local-area personal income estimates. In turn, these estimates are used by federal agencies in formulas that allocate federal funds for such programs as Medicaid and Aid to Families with Dependent Children.
Census data on place of work and other characteristics are used by the Office of Management and Budget to define metropolitan statistical areas, which in turn have many applications in federal programs.
Needs of Other Data Users
Reviewing federal agency uses of census data made it clear to us that the items in the 1990 census were there to serve important federal agency purposes. No item was included because it served exclusively the interest of some other community (e.g., business, academia).2 Of course, many other organizations and individuals in addition to federal agencies—state and local governments, business organizations, universities and other research institutions, nonprofit organizations, the media, students, and individual citizens—make use of census data, often in conjunction with data of their own. Indeed, since the availability of computer-readable data files beginning with the 1970 census, there has been an explosion in the use of census data by all kinds of users for all kinds of purposes (see Appendices C-H).
In some cases, these uses are a direct consequence of federal mandates—for example, states and local governments need certain items to obtain federal aid and to satisfy the requirements of the Voting Rights Act. In other instances, there is no direct link to a federal requirement, but the uses nonetheless serve federal purposes in a broad sense.
Thus, local governments make extensive use of small-area census data to identify target populations for services and to allocate facilities and resources to serve target populations most effectively. For example, they may use census tract or block group data on a variety of characteristics to locate health clinics or to determine the need for establishing programs to serve groups of new immigrants.
Often such analyses use data from other sources—for example, local governments frequently map local administrative data (e.g., on hospital admissions, clinic visits, school enrollment) to the same geographic areas for which they obtain census data, in order to determine rates of service use, indicators of unmet need, and the like. For these kinds of uses, the administrative data add an important element, but the array of census data is the linchpin that makes the analysis possible.
Businesses also make extensive use of census data. Some business uses—for example, to evaluate progress in hiring disadvantaged groups by analyzing occupation and demographic characteristics of local labor markets—directly serve public purposes established in federal legislation and court actions. Other business uses—for example, to help determine the best site for a plant or retail outlet by analyzing socioeconomic and demographic characteristics in different locations—serve public purposes indirectly by helping to make the best use of scarce investment dollars for expansion and redirection of business operations.
Research uses of census data serve a public purpose by advancing knowledge that in turn often has policy implications. For example, analysis of census data on patterns of internal migration contributes to understanding of the factors that influence regional growth and decline. Also, analysis of census data on the numbers, characteristics, and destinations of groups of new immigrants is directly relevant to current highly visible policy issues surrounding the benefits and burdens of immigration to the United States.
Finally, census data serve the important purpose of informing the public—through the media and other dissemination channels (e.g., libraries). Citizens benefit from information on the characteristics of their locality or their population group—individually and in comparison with other localities or groups—in performing many civic functions, such as voting, organizing to support legislation, and otherwise participating fully in a democratic society.
Proposals are sometimes made that data users should pay for collecting needed census information. These proposals miss the mark in several respects. First, from our review, there are no items included in the census solely to serve other users and not also to meet a federal agency requirement; hence, there are no items that could be charged off to private organizations or other nonfederal users. To the contrary, their uses of census information represent an added return on the federal investment in census data collection and one that is very cost-effective: once the data have been collected and tabulated for federal purposes, the cost of providing them to others is normally quite small.3 Moreover, having census data available to all and not just to those who pay is an important factor in maximizing the added returns from census data use (e.g., small businesses or poor localities are not disadvantaged in this regard).
Second, although federal agencies could be asked to pay for ''their" items, there are problems with such an approach. Many items, such as income and education, are needed as analytical variables by almost all agencies. Also, such a financing system would complicate planning for the census, in that the scope of the questionnaire could be uncertain until very late in the decade. Finally, given the need for agencies to have their items included in the census in order to carry out their mandates, it is unlikely that agencies would opt out of the census. Hence, to have agencies pay for items would at best amount to a rearrangement of federal budget dollars with little or no impact on overall census costs and, more likely, would add to costs given the need for added coordination.
Conclusion About Data Needs
We believe that the need for census content (beyond the minimum set of items) is clear. By this statement, we do not mean to endorse each and every item nor to suggest that the overall content could not be reduced somewhat. But we do mean to endorse the necessity and usefulness of data on a range of topics that can be cross-tabulated for small areas and small groups.
Hence, the issue, in our minds, comes down to whether there is some other mechanism that can obtain equivalent data, either at lower cost then adding content to the basic census or at about the same cost but with other benefits (e.g., greater frequency of data). We address this issue in Chapter 6 in our discussion of the cost and coverage effects of the long form and possible alternatives to it. In Chapter 8, we discuss more generally alternatives for obtaining more frequent small-area data between censuses.
CRITERIA FOR EVALUATING APPROACHES TO THE CENSUS
There are some major national policy choices involved in considering requirements and techniques for the decennial census. The panel sees its role as one of identifying those choices and providing analysis to make informed decisions possible. Some major options for conducting future censuses illustrate the dilemma in thinking about what a future census should look like. First, should the census be a highly intensive physical enumeration with a complete physical follow-up of all nonresponses, or should physical enumeration efforts be reduced through the use of sample-based follow-up? Should the census include an intensive effort to physically count every person, or should it rely on various statistical techniques for adjustment? The trade-offs, in this case, involve costs, accuracy, and the acceptability of statistical adjustment. A second choice involves how to implement the basic requirements of court decisions mandating equality in population size of voting districts. Both the court decisions and the Voting Rights Act—grounded firmly in the Constitution, including the Fourteenth and Fifteenth amendments—mandate census data requirements. Third,
what level of information beyond basic demographic detail must be provided every 10 years for small areas or small population groups? A large number of programs and funding allocation are based on decennial census data. Data from surveys or administrative records could be used to provide more timely estimates, albeit with less geographic detail than the census. Administrative records can provide estimates for small areas with selected content that is designed and reported for administrative purposes. To what extent can alternative sources meet reasonable needs, and at what cost?
Our conclusions on the merits of proposed changes in the census, including radical alternatives, fundamental reforms, and minor alterations, are based on several criteria. Most important among them are the following:
Information content and statistical quality. Under this rubric, we asked such questions as: Does the proposed alternative provide a range of information for small geographic areas and population groups? Specifically, does it provide data needed for constitutional purposes and for other required uses by federal government agencies? Does the alternative reduce or increase errors in the data, such as coverage errors, sampling errors, and other biases?
Operational feasibility and costs. Under this heading, we asked: What are the costs of the alternative, including developmental costs and likely steady-state operating costs? What are the implications for work flow and organization of the census process? Would legislation be required to implement the alternative?
Dissemination. Can the alternative provide data more frequently than every 10 years? How rapidly after data collection can output be provided? Can output be provided for locally defined and changing geographic areas? In addition, we asked whether various aspects of the census data collection process could be used, perhaps in combination with other sources, to provide better intercensal information.
Public acceptability. Will the alternative be acceptable to the public, or will it suffer by comparison to the high profile and historic image of the traditional census? Will a reasonably large proportion of households respond with accurate information? Will the data produced be politically acceptable for constitutional and legislative purposes?
Using these criteria, we considered ways for the U.S. census in 2000 to provide a rich array of needed data for small areas and small population groups at reduced costs and with acceptable quality. Most of the focus of this report is on planning for the 2000 census. It may not be possible, however, to fully implement all of the reforms advocated in this report in 2000, and thus the 2000 census may be a transition to a fully redesigned census in 2010 and beyond. We therefore have considered ways to improve estimates throughout the decade following
2000 and to achieve possibly even greater reform of the methodology for censuses in the year 2010 and beyond.
As it turned out, about 92 million questionnaires went to occupied households, and the remainder were sent to units that were determined to be vacant or no longer residential.
The items on appliances that were included in the 1960 and 1970 censuses (e.g., whether the household had a washing machine or dryer—see Appendix A) are often cited as examples of content provided to serve exclusively business marketing interests. However, these items were asked primarily because they represented important indicators for federal agencies of standard-of-living differences across the country; their use by the private sector was secondary. They may have been retained on the questionnaire for one too many censuses, but they were requested and justified by federal agencies.
People do pay some minimal amount for census data at present. The sale of census data normally includes the reproduction and publication costs.