The Bicentennial Census
NEW DIRECTIONS FOR
METHODOLOGY IN 1990
30th Anniversary Edition
Constance F. Citro and
Michael L. Cohen, Editors
Committee on National Statistics
Division of Behavioral and Social Sciences and Education
THE NATIONAL ACADEMIES PRESS
Washington, DC
www.nap.edu
THE NATIONAL ACADEMIES PRESS 500 Fifth Street, NW Washington, DC 20001
This 30th Anniversary Edition was supported by a contract with the U.S. Census Bureau (contract number YA01323-14-CN-0033). Support for the Committee on National Statistics is provided by a consortium of federal agencies through a grant from the National Science Foundation (award number SES-1024012). Any opinions, findings, conclusions, or recommendations expressed in this publication do not necessarily reflect the views of any organization or agency that provided support for the project.
International Standard Book Number-13: 978-0-309-37297-8
International Standard Book Number-10: 0-309-37297-6
DOI: 10.17226/21728
Additional copies of this report are available for sale from the National Academies Press, 500 Fifth Street, NW, Keck 360, Washington, DC 20001; (800) 624-6242 or (202) 334-3313; http://www.nap.edu.
Copyright 2015 by the National Academy of Sciences. All rights reserved.
Printed in the United States of America
Suggested citation: The National Academies of Sciences, Engineering, and Medicine. (2015). The Bicentennial Census: New Directions for Methodology in 1990. 30th Anniversary Edition. C.F. Citro and M.L. Cohen, Eds. Committee on National Statistics, Division of Behavioral and Social Sciences and Education. Washington, DC: The National Academies Press.
The National Academy of Sciences was established in 1863 by an Act of Congress, signed by President Lincoln, as a private, nongovernmental institution to advise the nation on issues related to science and technology. Members are elected by their peers for outstanding contributions to research. Dr. Ralph J. Cicerone is president.
The National Academy of Engineering was established in 1964 under the charter of the National Academy of Sciences to bring the practices of engineering to advising the nation. Members are elected by their peers for extraordinary contributions to engineering. Dr. C. D. Mote, Jr., is president.
The National Academy of Medicine (formerly the Institute of Medicine) was established in 1970 under the charter of the National Academy of Sciences to advise the nation on medical and health issues. Members are elected by their peers for distinguished contributions to medicine and health. Dr. Victor J. Dzau is president.
The three Academies work together as the National Academies of Sciences, Engineering, and Medicine to provide independent, objective analysis and advice to the nation and conduct other activities to solve complex problems and inform public policy decisions. The Academies also encourage education and research, recognize outstanding contributions to knowledge, and increase public understanding in matters of science, engineering, and medicine.
Learn more about the National Academies of Sciences, Engineering, and Medicine at www.national-academies.org.
This page intentionally left blank.
PANEL ON DECENNIAL CENSUS METHODOLOGY
JOHN W. PRATT (Chair), Graduate School of Business, Harvard University
PASTORA SAN JUAN CAFFERTY, School of Social Service Administration, University of Chicago
ANSLEY J. COALE, Office of Population Research, Princeton University
DONALD DESKINS, Department of Sociology, University of Michigan
IVAN P. FELLEGI, Deputy Chief Statistician, Statistics Canada, Ottawa
WAYNE A. FULLER, Department of Statistics, Carnegie-Mellon University
BENJAMIN KING, Research Statistics Group, Educational Testing Service, Princeton, NJ
ALBERT MADANSKY, Graduate School of Business, University of Chicago
ALBERTO PALLONI, Center for Demography and Ecology, University of Wisconsin
JOHN ROLPH, Rand Corporation, Santa Monica, CA
COURTENAY M. SLATER, CEC Associates, Washington, DC
JOSEPH WAKSBERG, Westat, Inc., Rockville, MD
CONSTANCE F. CITRO, Study Director
MICHAEL L. COHEN, Research Associate
NOTE: Information as of 1985.
COMMITTEE ON NATIONAL STATISTICS
1984-1985
LINCOLN E. MOSES (Chair), Department of Statistics, Stanford University
LEO BREIMAN, Department of Statistics, University of California, Berkeley
JOEL E. COHEN, Laboratory of Populations, The Rockefeller University
WAYNE A. FULLER, Department of Statistics, Iowa State University
SEYMOUR GEISSER, School of Statistics, University of Minnesota
F. THOMAS JUSTER, Institute for Social Research, University of Michigan
JANE A. MENKEN, Office of Population Research, Princeton University
JOHN W. PRATT, Graduate School of Business, Harvard University
S. JAMES PRESS, Department of Statistics, University of California, Riverside
CHRISTOPHER A. SIMS, Department of Economics, University of Minnesota
BURTON H. SINGER, Department of Statistics, Columbia University
COURTENAY M. SLATER, CEC Associates, Washington, DC
JUDITH M. TANUR, Department of Sociology, State University of New York, Stony Brook
DAVID L. WALLACE, Department of Statistics, University of Chicago
EDWIN D. GOLDFIELD, Executive Director
MIRON L. STRAF, Research Director
MICHELE W. ZINN, Administrative Associate
NOTE: Information as of 1985.
Foreword to the
30th Anniversary Edition
The decennial census, mandated in Article 1, Section 2, of the U.S. Constitution, produces data that are widely used by the government, academic, and private sectors, as well as being the subject of much interest and scrutiny by the general public. The census has also been the subject of wide-ranging study.
In 1985, the Panel on Decennial Census Methodology, convened by the Committee on National Statistics (CNSTAT) of the National Research Council of the National Academies, issued its final report with recommendations for research and development to improve the quality of the decennial census of population. The panel’s report, The Bicentennial Census: New Directions for Methodology in 1990 (National Research Council, 1985), provided a comprehensive review of census methods in the United States over time and in other Western countries. It made recommendations for research and development to lead to improvements in the 1990 and succeeding censuses in four areas: improving the count by reducing the numbers of people who were included erroneously (e.g., duplicates) and the numbers of people who were inadvertently omitted; the uses of sampling and administrative records to improve the cost-effectiveness of the count; measuring the completeness of the count; and adjusting the count to correct for coverage errors.
Since the report’s release, many of the panel’s observations and recommendations have proved to be prescient, while some of them have proved to be infeasible to carry out within the time constraints for delivering population totals for purposes of congressional reapportionment and legislative redistricting. Yet the panel’s report remains of interest for its comprehensive
review in two areas, one being the issues that confronted the census toward the end of the 20th century in light of growing evidence of errors in the count, particularly differential coverage of demographic and socioeconomic groups, the other being advances in statistical methods that held promise to improve the count.
The report has long been out of print, and it was never available electronically. Given continuing interest in it, and, more generally, in the evolution of the U.S. decennial census, CNSTAT decided to celebrate the report’s 30th anniversary by making it available in an online edition.
In preparing this online edition, the panel’s study directors, Constance Citro (currently, CNSTAT director) and Michael Cohen (currently, CNSTAT senior program officer), carefully proofread the original text only to correct previously undetected typographical and formatting errors. The index was also carefully read to ensure that the page references continued to be correct. We are immensely grateful to Jacqui Sovde, former CNSTAT program associate, for her painstaking work to create working files from scans of the original text. We are also grateful to Eileen LeFurgy, CNSTAT program coordinator, the staff of the Reports Office of the Division of Behavioral and Social Sciences and Education, and the National Academies Press.
Constance F. Citro, Director
Committee on National Statistics
Contents
The Importance of Choice of Methodology for 1990
Proposed Changes in Methodology
Independent Reviews of Decennial Census Plans
Overview of the Report and Recommendations
2 Purposes and Uses of the Decennial Census
Historical Uses of Census Data
Distinguishing Features of the Modern Census and Its Uses
Effects of Census Error on Key Uses: Review of Research
Effects of Errors in Postcensal Estimates
Appendix 2.1—State and Local Government Uses of Census Data
Local Agency Uses of Census Data
Appendix 2.2—Census Data Use in New Jersey—A Case Study
3 Census Methodology: Prior Practice and Current Test Plans
Methodology Used in Previous Censuses
Methodology Used in Other Western Countries
Census Bureau Research Plans for 1990
Assessment and General Recommendations
Appendix 3.1—An Overview of Sequential Hot-Deck Imputation
Appendix 3.2—A Description of Iterative Proportional Fitting
4 Evaluating the Decennial Census: Past Experience
Methods of Coverage Evaluation
Coverage Evaluation Prior to 1980: Micro-Level Methods
Coverage Evaluation Prior to 1980: Macro-Level Methods
The 1980 Post-Enumeration Program
Recent Use of Administrative Lists for Coverage Evaluation
Considerations for Assessing Alternative Coverage Evaluation Methods
Error Profiles for Coverage Evaluation Methods
Appendix 4.1—An Introduction to Estimation from Multiple Lists
Appendix 4.2—Operational Aspects and Modeling of Computer Matching
Operational Difficulties of Matching
A Mathematical Model for Record Linkage
5 Taking the Census I: Improving the Count
Hard-to-Count Groups in the Census: What Is Known
Coverage Improvement Programs: Past Experience
Census Bureau Plans for Testing Coverage Improvement Programs for 1990
Needed Research on Undercount and Overcount
Issues in Coverage Improvement: Questionnaire Content
Issues in Coverage Improvement: Special Enumeration Procedures
Appendix 5.1—Gross Omissions and Gross Overenumerations in the Census
Gross Overenumerations of People
6 Taking the Census II: The Uses of Sampling and Administrative Records
The Use of Sampling for Follow-up
Sampling for Coverage Improvement
The Use of Administrative Records and Sampling for Improved Accuracy of Content
Appendix 6.1—Cost Estimates for a Sample Census
Appendix 6.2—Illustrative Follow-up Scenario Using Sampling
Appendix 6.3—Improving Data on Housing Structure Items: A Suggested Method
7 Adjustment of Population Counts
Evaluating Adjustment: Loss Functions and Yardsticks
Considerations of Internal Consistency
Census Bureau Plans for Research and Testing on Adjustment
Priorities for Research and Testing on Adjustment
Appendix 7.1—A Quick Look at Loss Functions and Apportionment
Appendix 7.2—An Introduction to Hierarchical Bayesian Techniques
Appendix 7.3—Aggregation of Synthetic Estimates: A Counterintuitive Example
8 Measuring the Completeness of the 1990 Census
Recapitulation of Major Issues in Coverage Evaluation
Appraisal of Census Bureau Pretest Plans for Coverage Evaluation
The 1990 Demographic Analysis Program: Possible Improvements
The Reverse Record Check Program: Considerations for 1990
The 1990 Post-Enumeration Program: Possible Improvements
Systematic Observer Methodology: Considerations for 1990
Appendix 8.1—The Population of Illegal Aliens: Methods and Estimates
Appendix 8.2—Estimates of Variance and Cost for a Large Systematic Observer Study
Tables
Table 2.1 Uses of Census Data in Selected Federal Grant-in-Aid Programs
Table 2.3 Selected Measures of Accuracy of County Population Estimates for 1980, by Size of County
Table 2.4 Percentage of Error in Subcounty Population Estimates for 1980
Table 2.5 Selected Measures of Accuracy of Subcounty Population Estimates for 1980, by Size of Area
Table 3.1 Notation for Iterative Proportional Fitting for a Small Two-Way Table
Table 4.2 Scheme to Identify Various 1980 PEP Estimates
Table 4.4 Two-Way Table Underlying Simple Dual-System Estimation
This page intentionally left blank.
This page intentionally left blank.
In 1982 the American Statistical Association Technical Panel on the Census Undercount recommended “that the Bureau of the Census sponsor an outside technical advisory group on undercount estimation and related problems” (American Statistical Association, 1984:256). Partly on the basis of that recommendation, the Census Bureau requested the Committee on National Statistics to establish a panel (1) to suggest research and experiments, (2) to recommend improved methods, and (3) to guide the Census Bureau on technical problems in appraising contending methods with regard to the conduct of the decennial census.
In response to that request, the Panel on Decennial Census Methodology was established and charged with investigating three major issues from a technical viewpoint, setting aside legal considerations:
- Adjustment of census counts and characteristics, including exploration of formal criteria to evaluate measures of undercount and alternative adjustment procedures;
- Uses of sampling in the decennial census, specifically investigation of whether sampling for coverage improvement and of nonrespondents for follow-up can improve accuracy at a given cost; and
- Uses of administrative records, including investigation of the possible utility of various types of records for improving the accuracy of census counts and the efficiency of census operations.
At our first meeting, in January 1984, we took a broad view of the charge and identified additional topic areas beyond those listed for pos-
sible investigation. We decided that in order to reach sensible conclusions regarding a choice of methodology for the decennial census it was critical to examine uses of census data and the degree of accuracy needed to satisfy each use. We also decided that it was essential to conduct a thorough review of procedures for improving census coverage and of methods of evaluating the completeness of coverage achieved in the census. Well-designed and well-executed coverage improvement programs can importantly reduce errors in the census. Well-designed and well-executed coverage evaluation programs inform users about the quality of the census results and are the source of input data necessary for any type of adjustment of the census counts to reduce errors further.
The panel produced an interim report that focused on recommendations for improvements in census methodology that warranted early investigation and testing. That report, Planning the 1990 Census: Priorities for Research and Testing (National Research Council, 1984), addressed three topic areas that were central to the original charge: (1) uses of sampling for the census count, (2) methodologies for evaluating completeness of coverage of the census, and (3) issues related to the adjustment or modification of census counts and characteristics. In addition, the report reviewed the Census Bureau’s plans for the 1985 pretest of a two-stage methodology for conducting the census.
This report updates and expands our ideas and conclusions about decennial census methodology. In it we endeavor to assess the merits of investigating proposed changes in the decennial census that represent important departures from past practice and, specifically, to recommend concepts and procedures that we believe the Census Bureau should place high on its list of priority objectives for research and testing directed toward the nation’s bicentennial census in 1990.
Our report, following an introduction and overview that presents in collected form the recommendations of the panel, includes three background chapters, on purposes and uses of the decennial census, on methodology of prior censuses and current 1990 census testing plans, and on past experience with coverage evaluation. The report offers general and specific planning recommendations in five areas: (1) overall strategy for planning the 1990 census, (2) procedures for coverage improvement as part of the census, (3) uses of sampling and administrative records in taking the census, (4) adjustment of census counts and characteristics, and (5) measuring the completeness of the 1990 census.
With regard to the key issue of adjustment of census counts, we argue for balance between efforts to achieve a complete enumeration and efforts to improve the accuracy of census figures through adjustment procedures. We believe that adjustment cannot be viewed as an alternative to obtaining as complete a count as possible through cost-effective means. However, the
evidence is overwhelming that no counting process will in fact enumerate everyone. Given our belief that the ultimate goal of the census should be the accuracy of the figures, we recommend that the Census Bureau pursue a vigorous program of research on coverage evaluation and adjustment methods that, if successful, would permit adjustment of the 1990 census counts.
With regard to issues related to adjustment and other topics considered by the panel, the emphasis of our report is on the extensive research and testing needed during the next few years to support sound decisions regarding the choice of a particular methodology for the 1990 and subsequent censuses. Given the limited number of testing opportunities available compared with the range of ideas that appear attractive to try out, we believe it is imperative for the Census Bureau to choose among research objectives. We have therefore endeavored to provide timely advice regarding what we believe are the most promising avenues to pursue.
John W. Pratt, Chair
Panel on Decennial Census Methodology
This page intentionally left blank.
The Panel on Decennial Census Methodology wishes to thank the many people who contributed to the preparation of this report.
The staff of the Bureau of the Census has been extremely helpful, and we would like to thank particularly certain individuals for their assistance. Peter Bounpane, assistant director for demographic censuses, and Barbara Bailar, associate director for statistical standards and methodology, have been very generous of their time in providing assistance to the panel. Other Census Bureau staff who contributed valuable information include: Frederick Bohme, Charles Cowan, Gregg Diffendal, Richard Engels, Robert Fay, III, Penelope Harvison, Roger Herriot, Catherine Hines, Howard Hogan, Matt Jaro, Bruce Johnson, Charles Jones, Eli Marks, Nampeo McKenney, Susan Miskura, Jeffrey Passel, Paula Schneider, John Thompson, David Whitford, Kirk Wolter, and Arthur Young.
A number of members of the broader statistical community gave very helpful presentations at meetings of the panel: Eugene Ericksen of Temple University and Mathematica Policy Research, Monroe Sirken of the National Center for Health Statistics, John Tukey of Princeton University and AT&T Bell Laboratories, and Kenneth Hill and Robert Warren of the Committee on National Statistics. We would also like to thank Harold Nisselson of Westat, Inc., who prepared background material on coverage evaluation programs, Margaret Boone, anthropologist, who made a helpful presentation on ethnographic research, and Judith Rowe of the Princeton-Rutgers Census Data Project for very useful comments on Chapter 2 of the report.
The members and staff of the Committee on National Statistics were valuable resources to the panel in the course of our work. Stephen Fienberg,
as chair of the committee, gave us the benefit of his enthusiastic support, direction, and encouragement in the early stages of the project. Committee member David Wallace read the draft report with great care and made detailed, incisive comments that improved the report in important ways. Staff members Edwin Goldfield, Thomas Jabine, Daniel Levine, Margaret Martin, and Miron Straf provided useful guidance and assistance. Kenneth Hill made very helpful comments on Chapters 4 and 8 of the report. Deserving of special thanks is Julie Kramen, who assisted in the production of many of the tables included in this report. The panel is grateful to Christine McShane, editor, for the Commission on Behavioral and Social Sciences and Education, for her fine technical editorial work, which contributed greatly to the organization and readability of this report. We would also like to thank members of the Commission on Behavioral and Social Sciences and Education and the Committee on National Statistics who reviewed the report and offered cogent comments.
I would like to express special thanks to the panel’s own staff. Constance Citro, as study director, stayed on top of everything, attended to myriad administrative matters, chased down elusive details and people, and arranged and ran efficient, productive meetings. Her ability at the same time to write rapidly, accurately, and clearly about anything and everything was continually amazing. Without sacrificing quality, she met schedules no one else thought possible. The buck could safely be passed to her any time, and often was—yet her effort was unstinting, her temper never ruffled. She was aided nobly throughout by Michael Cohen, research associate. He bravely did battle with topics none of the panel would touch, bearing with remarkable stoicism the slings and arrows inevitably ensuing. Together they were responsible for drafting most sections of the report and rewriting patiently to accommodate comments from all directions.
Finally, I wish to thank the panel members themselves for their generous contributions of time and expert knowledge. Several of them prepared background materials for discussion and drafted sections of chapters. Most provided detailed comments on drafts of the report. Of course, no individual member of the panel should or would want to be held responsible for every word or idea expressed, but the report does represent their collective thinking on the issues addressed. They were an unusual pleasure to work with and learn from; the chief reward of a panel chair was amply paid.
John W. Pratt