Decennial Census: Overview of Historical Census Issues (Chapter Report,
05/01/98, GAO/GGD-98-103).

Pursuant to a congressional request, GAO provided information on
historical census issues and reviewed the Census Bureau's plans for the
2000 census. GAO did not evaluate the potential for sucess, or make
recommendations, regarding the 2000 census.

GAO noted that: (1) the framers of the Constitution established a
requirement for the national government to undertake the census, and
described, in general terms, how it should be accomplished; (2) while
apportionment is the most widely known use of census data, the data is
also used for congressional redistricting, managing federal agencies,
and allocating federal funds, and is disseminated to state and local
governments, academia, and the private sector as well; (3) for the 2000
Census, the Bureau is planning to rely on a Master Address File, which
is to be developed, in part, from the Bureau's 1990 Census address list
and the most recent Postal Service address list; (4) the Bureau plans to
conduct a 100-percent canvass of all census blocks in early 1999 and
will request the Postal Service to validate the city-style addresses
prior to the delivery of 2000 Census questionnaires; (5) the Bureau is
planning for enumerators to deliver the questionnaires and ask that they
be mailed back; (6) the Bureau will focus its efforts to count the
homeless in the places where many of them come for services, such as
shelters and soup kitchens, as well as targeted outdoor locations; (7)
the Bureau has decided to use a paid advertising campaign in 2000 to
complement its continuing efforts with its organizational partners; (8)
the Bureau is researching the use of questionnaires in additional
languages; (9) in order to improve mail response rate, the Bureau is
planning to use a new, multiple mail contact strategy; (10) the Bureau
plans to use two new sampling procedures in the 2000 census designed to:
(a) reduce the time required for and expense of following up on the
projected 40 million housing units that may not respond in 2000 to the
questionnaires; and (b) adjust the population counts obtained from
census questionnaires and nonresponsive follow-up procedures to
eliminate the endemic differential undercount; (11) the cost of the
census has steadily increased over 200 years, and the rate of increase
will continue to escalate with the 2000 Census; (12) the 2000 Census
will rely on computer technology to a greater extent than ever before;
and (13) several thousand full-time employees from Bureau headquarters
are expected to work on the 2000 Decennial Census.

--------------------------- Indexing Terms -----------------------------

 REPORTNUM:  GGD-98-103
     TITLE:  Decennial Census: Overview of Historical Census Issues
      DATE:  05/01/98
   SUBJECT:  Census
             Population statistics
             Data collection
             Constitutional amendments
             Constitutional law
             Statistical methods
             Public relations
             Mailing lists
             Projections
             Surveys
IDENTIFIER:  2000 Decennial Census
             Census Bureau Post Enumeration Survey
             Census Integrated Coverage Measurement
             1990 Decennial Census
             Census Bureau Topologically Integrated Geographic Encoding 
             and Referencing System
             Census Bureau Current Population Survey
             Census Bureau Master Address File
             Community Development Block Grant Program
             
******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  Delineations within the text indicating chapter **
** titles, headings, and bullets are preserved.  Major          **
** divisions and subdivisions of the text, such as Chapters,    **
** Sections, and Appendixes, are identified by double and       **
** single lines.  The numbers on the right end of these lines   **
** indicate the position of each of the subsections in the      **
** document outline.  These numbers do NOT correspond with the  **
** page numbers of the printed product.                         **
**                                                              **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced.  Tables are included, but    **
** may not resemble those in the printed version.               **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************


Cover
================================================================ COVER


General Government Division

May 1998

DECENNIAL CENSUS - OVERVIEW OF
HISTORICAL CENSUS ISSUES

GAO/GGD-98-103

Overview of Historical Census Issues

(410187)


Abbreviations
=============================================================== ABBREV

  HUD - Department of Housing and Urban Development
  ICM - integrated coverage measurement
  INS - Immigration and Naturalization Service
  LUCA - local update of census addresses
  MAF - master address file
  NRFU - nonresponse follow-up
  OMB - Office of Management and Budget
  RFD - rural free delivery
  PES - post-enumeration survey
  TIGER - Topographically Integrated Geographic Encoding and
     Referencing System

PREFACE
============================================================ Chapter 0

The purpose of this report is to furnish context, background, and a
historical perspective for Members of Congress, their staffs, and
others interested in the debate on policy and operational questions
related to the 2000 Census.  It is organized by general decennial
census concepts and discusses critical issues that emerged during
decennial censuses as far back as the first one in 1790. 

This report describes, in general terms, the Census Bureau's plans
for the 2000 Decennial Census, but it does not evaluate the potential
for success of the 2000 Census nor does it contain recommendations. 
Many other GAO reports and testimonies (listed in app.  V and
available upon request) analyze the 1990 Census and provide our
perspective on the current progress and potential risks of the 2000
Census.  In addition, Census Bureau, congressional, and other reports
on technical issues and planning for the 2000 Census are available. 

In developing this report, we researched available literature on the
census (including publications on demography in early America, the
framing of the Constitution, and social history) and interviewed
Bureau officials knowledgeable about census history.  We also
reviewed publications prepared by the National Academy of Sciences'
National Research Council on census issues and drew on available
documentation describing plans for the 2000 Decennial Census.  (See
app.  IV for a listing of non-GAO publications.) It should be noted
that, as with any effort to put current events into historical
context, alternative interpretations are possible.  Often debated are
the intent and motivation of the framers of the Constitution who
created the census.  In this report, we have quoted the Constitution
and various laws relating to the decennial census and have attempted
to place their language in an historical context.  We are not,
however, providing our own independent review or interpretation of
the constitutional and statutory issues discussed in this report,
which, unless otherwise noted, are based primarily on the analysis
contained in the various publications and documents we relied upon in
preparing this report. 

Short answers to some frequently asked questions about the decennial
census are in appendix I.  Appendix II provides information on
changes in the apportionment of the membership of the House of
Representatives between the 1920 and 1990 Decennial Censuses by
region of country, changes in the nation's population and its
undercount by race and ethnicity between the 1950 and 1990 censuses,
and a snapshot of the growth and cost of census-taking since the
first decennial census in 1790.  Major contributors to this report
are identified in appendix III. 

L.  Nye Stevens
Director, Federal Management
 and Workforce Issues


WHY TAKE THE CENSUS? 
============================================================ Chapter 1


   THE CONSTITUTION REQUIRES A
   CENSUS
---------------------------------------------------------- Chapter 1:1

The United States of America in 1790 was the first modern nation to
undertake a comprehensive and periodic count of its population as a
regular responsibility of government.  But the American decennial
census--mandated in the Constitution--was also a component of a new,
unprecedented concept of representative government. 


      THE TRADITION OF
      RECORDKEEPING
-------------------------------------------------------- Chapter 1:1.1

A decennial census was an extension of colonial habits of
recordkeeping born in the traditions of Europe.  Old World religious
institutions had long kept the vital records of their parishioners,
and as early as 1611, the London Company required the residents of
Jamestown to keep a record of local christenings, marriages, and
deaths.\1 A few years later, the Virginia Assembly passed its own law
requiring not only the recordation of these events, but an annual
quantitative report of them.  Some colonies sporadically tracked a
range of data about their populations, including occupation, gender,
and age, during the pre-Revolutionary period.  Some served a
particular purpose, such as determining the number of
military-eligible men; others reflected an English tradition of
tracking population movements that developed during the great social
upheavals of the Elizabethan era.  Although the Articles of
Confederation mandated a triennial census for taxation purposes, the
revolutionary war prevented its implementation; no general census of
the colonies as a whole was ever carried out. 


--------------------
\1 James H.  Cassedy, Demography in Early America; Beginnings of the
Statistical Mind, 1600-1800, (Cambridge, MA:  Harvard University
Press, 1969). 


      THE CONSTITUTIONAL
      IMPERATIVE
-------------------------------------------------------- Chapter 1:1.2

The Constitution of the United States established a new,
philosophically innovative, and technically complex form of
government, which in turn established a need for periodic censuses. 
A principal innovation of the new government was that it would be
representative of the population by means of elections.  One of the
principal complexities the framers faced was how to make it
representative, and particularly how to reflect the interests of the
American people both as residents of a state and as individuals.  The
Senate, therefore, was designed to represent the interests of the
states, and the House of Representatives was designed to represent
the interests of individuals. 

In section 2 of Article 1 of the Constitution, which concerns the
composition of the House of Representatives, the framers wrote: 
"Representatives and direct Taxes shall be apportioned among the
several States which may be included within this Union, according to
their respective Numbers...." If the Members of the House were to be
apportioned among the states ï¿½according to their respective Numbers,ï¿½
then the populations of the states had to be determined.  The
framers, aware that the states had already demonstrated different
ideas about how to count their populations for apportioning delegates
to the Continental Congress, stipulated the number of representatives
for each state until a census could be taken.  Furthermore, they
established a requirement for the national government to undertake
the census; and described, in general terms, how it would be
accomplished:  "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."


      APPORTIONMENT AND ITS IMPACT
-------------------------------------------------------- Chapter 1:1.3

The first census was duly taken in 1790 in a manner directed by
Congress.  The census was to have taken 9 months, but actually took
18 months.  President George Washington believed that the count of
3.9 million people was too low.  Congress, however, accepted the data
and proceeded to apportion the number of representatives in accord
with the census data.  Immediately, the debate about how exactly to
implement the apportionment started, and this controversy has
continued in one form or another for over 200 years. 

The Constitution did not specify in precise detail how to apportion
the Representatives; it only specified that there would be a minimum
ratio of 1 representative to every 30,000 of the population.  It also
did not fix the number of seats there should be in the House, and it
was silent on how the states were to elect their representatives. 
So, questions arose:  How many people should one Representative
represent?  When the population of a state divided by 30,000, or
whatever divisor was ultimately selected, had a remainder, should
that remainder be dropped or rounded?  Should the size of the House
be fixed first and then apportioned, or should an apportionment fix
the size of the House?  The questions of apportionment method were
more than academic; depending on the answers, states could gain or
lose seats in the House of Representatives. 

The debates over the years about methods of apportionment focused on
mathematics, but the crux of the matter was political power.  Not
only did various apportionment methods affect individual states'
power, but they also influenced the outcome of national political
debates and, over time, the balance of power between large and small
states, northern and southern states, and the urban east and the
agricultural/extractive west. 

The context of the debate was always the rapidly growing population
revealed by each successive census.  From the beginning, the U.S. 
population grew at a phenomenal rate:  over 30 percent between
decennial censuses until the Civil War and 15-25 percent through
1930.\2 This growth and the simultaneous shifts in geographical
concentration of the population resulted in dramatic reapportionments
among states.  To illustrate, in the 1920s, 91 representatives were
apportioned to the Middle Atlantic states of New York, New Jersey and
Pennsylvania, but by 1990, that number had declined to 65.  In the
1920s, 19 representatives were apportioned for the Pacific states of
Washington, Oregon, and California, but by 1990, that number had
increased to 66.  (Changes in the results of the apportionment of the
House of Representatives between 1920 and 1990 are shown in app.  II,
table II.1.)

In 1911, when Congress fixed the number of representatives at 435--1
per state with the rest apportioned--the census results had even
greater significance.  Before this decision, a state's loss of
population, and therefore of representation, was mitigated by
continuing increases in the total number of representatives.  Before
the change, states whose population declined relative to other states
did not often lose representatives, although their representatives
were relatively less powerful as members of a now larger House.  But
after 1911, a gain of representation for any one state came only with
a loss of representation for another state. 

Congress failed to reapportion following the 1920 Census.  The
failure was in part the result of a difference of opinion over the
method of dividing political power.  Throughout the 1920s, Congress
debated which of two mathematical models for reapportionment--whose
outcomes for distribution of House seats differed--would be used.  In
1929, one mathematical method was selected for the reapportionment,
but it was not applied until after the 1930 Census.  Furthermore, the
debate about apportionment methods was not over.  In 1941, a
different model was chosen called "the method of equal proportions.\3
" It is still in use today. 

The failure to reapportion in 1920 was also a reflection of regional
power dynamics.  The results of the 1920 Census revealed a major and
continuing shift in population from rural to urban areas, which meant
that many representatives elected from rural districts resisted
reapportionment.  Also, the growing number of immigrants entering
this country had some impact on population shifts.  Delay followed
delay as rural interests tried to come up with mechanisms that would
reduce the impact of the population shift.  Congressmen from rural
areas that would lose seats to more urbanized areas simply blocked
passage of reapportionment legislation for 9 years. 

During the congressional debates on Pubic Law 71-13 that was enacted
in 1929, language requiring that districts be composed of contiguous,
compact territory and contain the same number of individuals was
deleted.  Therefore, the reapportionment law that finally passed in
1929 was silent on the subject of rules for how the states were to
establish districts to elect their representatives.  As a result,
some states simply stopped redistricting, despite major changes in
the internal distribution of their populations over time from rural
to urban to suburban.  A process of malapportionment--meaning
establishment of districts containing unequal population
sizes--continued unchecked for decades. 

The difference in population size among congressional districts
increased, setting the stage for the debate that started in the 1960s
and continues today:  What are the standards for population size and
shape of congressional districts?  The federal courts, which had
declined for 40 years to rule on malapportionment cases, revisited
the issues in the early 1960s, and ultimately the Supreme Court
accepted the argument that ï¿½one-man, one-voteï¿½ meant that districts
had to be of equal size.\4 More recently, issues concerning the
ethnic composition of those districts and their physical shape have
arisen.  Interpretations of the Voting Rights Act of 1965, as
amended, led to the creation by 1990 of over 50 congressional
districts configured in order to achieve a majority of minority
population.  For a time, there was no limit to the peculiarity of the
shape of a district created in the process of meeting the goals of
the act.  Now, the courts are reconsidering limits to the
eccentricity of the shape of a district. 

Census data have been essential to the late twentieth century debates
on the division of representational power, just as they were to the
first apportionment debate in 1791.  Block-by-block census data were
essential to the 1960s goal of creating districts of equal population
size.  The efforts in the 1990s to achieve majority minority
districts relied on census data on the ethnicity of the people in
specific blocks.  The debate over how best to meet the constitutional
objective of representative government continues today with the
census still at the core of the debate. 


--------------------
\2 Statistical Abstract of the United States--The National Data Book,
Bureau of Census, Department of Commerce, October 1996. 

\3 The method of equal proportions is based on complex mathematical
calculations whereby each state is initially assigned one seat while
the other seats are assigned to the states based on the "priority
numbers" given to the states on the basis of their relative
population numbers.  "Congressional and State Reapportionment and
Redistricting:  A Legal Analysis," CRS Report for Congress, American
Law Division, Congressional Research Service, The Library of
Congress, 96-732A, September 4, 1996. 

\4 Baker v.  Carr, 369 U.S.C.  186 (1962). 


      DIRECT TAXATION RARELY USED
-------------------------------------------------------- Chapter 1:1.4

The census, according to Article I, section 2, of the Constitution,
was also to be used to apportion any direct taxes levied by the
federal government.  The founding fathers purposefully linked the
two.  Their thinking was that any incentive for a state to boost
population in order to gain additional representation would be offset
by the disincentive of raising its tax burden.  Direct taxation,
however, was enacted only twice--once in 1798 to try to diversify the
federal government's reliance on tariffs and customs duties and once
to finance the War of 1812.  Both taxes were based on the value of
land, houses, and slaves, and both were difficult to assess and
collect.  While this authority has never been repealed, direct
taxation based on a decennial census never became practical. 


   CENSUS DATA HAVE OTHER
   IMPORTANT USES
---------------------------------------------------------- Chapter 1:2

While apportionment is the most widely known use of census data, the
data are also used for congressional redistricting, managing federal
agencies, and allocating federal funds, and are disseminated to state
and local governments, academia, and the private sector as well. 
Data from a decennial census\5 provide official, uniform information
gathered over the decades on this country's people and their social,
demographic, and economic characteristics.  They provide the
baselines for countless other surveys and are used to develop
sampling frames for a number of other federal data collections, such
as the Current Population Survey, which is used to measure
participation in the labor market and unemployment rates. 


--------------------
\5 Decennial census data are gathered from short and long form
questionnaires.  Questions on the short form are asked of the entire
population, and questions on the long form are asked of only a
portion of the population for projection of national information. 


      CONGRESSIONAL REDISTRICTING
-------------------------------------------------------- Chapter 1:2.1

The Constitution does not require that states use the census data to
redraw the boundaries of congressional districts following a change
in the apportionment of representatives, but most states have always
used the census data for this purpose.  The general perception of the
impartiality of the Bureau and the great cost and administrative
effort required to take a census have been strong arguments in favor
of using the Bureau's data.  In addition, the ready availability of
census data is important because redistricting generally has been
required shortly after the census data are made available.  In recent
years, immediate and detailed population data have become especially
critical because some states have court-ordered deadlines to complete
redistricting. 


      MANAGING FEDERAL AGENCIES
-------------------------------------------------------- Chapter 1:2.2

The decennial census is a cost-effective method of providing baseline
and trend data for use by federal agencies and various other census
stakeholders, compared to the alternative of multiple data
collections by other federal agencies for their own purposes. 
Decennial census data and data from other Bureau surveys assist
federal agencies in managing their unique mission responsibilities. 
Federal agencies can use Bureau data to assist in evaluating
established programs, identifying the particular geographic area of
the county where success or problems are occurring, planning
corrective actions, and later determining if their corrective actions
were effective.  For example, Bureau data can assist federal agencies
in managing programs under the Government Performance and Results
Act.  Under this Act, agencies must measure their performance against
the goals they have set and report to Congress and the public on how
well they are doing. 

Federal agencies often turn to census data in managing their programs
because it is mandated by legislation or regulation.  The use of
census data is a legal requirement in some federal programs.  For
example, the Department of Housing and Urban Development (HUD) is
required to use Census data as the basis for allocating funds for the
Community Development Block Grant Program (42USC5302).  Without these
data HUD would be unable to meet legislatively mandated requirements
because there is no other source of data at the geographic level
needed. 


      ALLOCATING FEDERAL AID
-------------------------------------------------------- Chapter 1:2.3

The distribution of federal revenues in order to meet national
socioeconomic objectives started in the late nineteenth century with
an appropriation to each state to establish agricultural experiment
stations at land grant colleges.  In the early decades of this
century, Congress gradually expanded its provision of federal
assistance.  During the mid-1930s as New Deal programs, including
social security, expanded to account for roughly one-third of the
federal budget, the need for greater detail and higher quality census
data increased.  To this day, census data remain an important element
in the allocation of federal aid to state and local governments, and
with billions of dollars at stake, the data are scrutinized intensely
for accuracy. 

For fiscal year 1998, funding estimates indicate states should
receive about $170 billion in aid through 20 federal programs that
used census data, in whole or in part, to allocate that aid.  The
largest of these programs is Medicaid, which plans to distribute
about $104.4 billion in fiscal year 1998, followed by the Federal Aid
Highway Program at $20 billion, and $7.5 billion under Title I grants
to local education agencies.\6 Census information is important to the
distribution of these federal funds, though generally it is not the
sole factor in allocation formulas. 


--------------------
\6 1997 Catalog of Federal Domestic Assistance, Executive Office of
the President, Office of Management and Budget, Washington D.C.,
1997. 


      HELPING LOCAL GOVERNMENTS
-------------------------------------------------------- Chapter 1:2.4

The decennial census produces data that states use not only to
determine boundaries for congressional districts, but also to
establish boundaries for smaller jurisdictional divisions.  The
census is also a rich source of data to help county and city
governments plan for and provide services.  The data help them answer
questions such as the following: 

  -- Will the population of preschoolers in the various school
     districts warrant building additional elementary schools? 

  -- Are local transit systems reaching the people likely to use
     public transport, and

  -- Where and when should the next senior citizen facility be built? 

Without federal census data, state and local governments would have
to undertake their own censuses, a costly alternative given the
federal government's experience and economies of scale. 


      HELPING BUSINESSES
-------------------------------------------------------- Chapter 1:2.5

Businesses use the aggregated census data\7 available to them to plan
for and provide their services and goods.  Census data about
population trends help businesses succeed--and provide jobs in the
process--by alerting them to opportunities to provide new services
and products and tailor existing ones to demographic changes.  Census
data also help businesses efficiently target their advertising
dollars.  A free sample, for example, of a magazine focused on the
interests of Hispanic readers can be distributed based on information
at the census block level.  Companies also use population data to
locate new stores where they expect likely consumers to be, as well
as to locate production facilities where they can expect to find a
suitable labor force. 


--------------------
\7 Some Bureau data are free, while special data requests are filled
for a fee. 


TAKING THE CENSUS
============================================================ Chapter 2


   DECIDING HOW TO COUNT
---------------------------------------------------------- Chapter 2:1

ï¿½The actual Enumeration shall be made,ï¿½ according to the
Constitution, under Article 1, section 2, ï¿½...  in such Manner as
they [Congress] shall by law direct.ï¿½ In effect, this has enabled
Congress to adjust decennial census procedures allowing for changes
in American society unforeseen in 1787.  Congress responded by
delegating the census-taking to executive branch agencies while
maintaining overall responsibility and periodically enacting
legislation affecting census-taking methodology.  While changes to
census-taking methodologies have occurred, one constant--the focus on
identifying households and enumerating people within them--has stayed
the same. 


      CHANGING CONDITIONS IN
      AMERICAN SOCIETY
-------------------------------------------------------- Chapter 2:1.1

Since the 1790 Census, American society has constantly changed,
thereby necessitating changes in the methodology of enumeration in
the decennial census.  Among the most significant societal changes
have been: 

1.  Increased population mobility:  Although westward-bound frontier
pioneers were difficult to count in the late-eighteenth and
nineteenth centuries, the number of mobile Americans today is much
greater, increasing problems for census-taking.  Short-term renters,
"snowbird" retirees, students splitting their residence between home
and college, and young urbanites rotating temporary residences are a
few of the modern phenomena that have created a population mobility
unimagined in 1787.  During the period 1990 to 1994, 17 percent of
the American population on average changed residences each year.\8 \8

2.  Varied domestic arrangements:  Households have always been the
major focus of census enumeration.  In eighteenth century America,
nearly all citizens identified themselves with a household whose
members were almost always related by blood, marriage, or through
regular employment, and therefore included servants, slaves,
apprentices, and resident farmworkers.  Most people lived in a
family-occupied dwelling that was headed by a male readily able to
provide a count of and characterize the members of his household. 
Today divorce, cohabitation without marriage, and group housing,
among other domestic arrangements rarely heard of in 1787, make the
determination of whom to count and where to count them increasingly
complex.  From 1970 to 1990 alone, the number of American households
grew 47 percent, while average household size shrank from 3.1 to 2.6
people and nonfamily households grew by 128 percent.\9

3.  People of varied linguistic backgrounds:  The heads of household
to whom the census questions were posed in the late eighteenth
century came overwhelmingly from Western European cultural traditions
and spoke a limited number of Western European languages.  Today, the
U.S.  population includes people from a great variety of countries,
and language barriers pose significant challenges in taking the
census.  To deal with this diversity, in 1990 the Bureau had
questionnaire guides available in over 32 languages and had
enumerators able to speak about 50 languages. 

4.  Increased concerns about privacy:  As a result of changing
attitudes toward government in general, concerns that census
information will be passed to other government agencies, and fears of
further loss of privacy in the computer age, the rate at which the
population voluntarily responds to requests for census information
has declined.  For example, mail response (considered to be the most
reliable and cost efficient means of obtaining census information)
declined from 78 percent in 1970 to 65 percent in 1990.  (Chapter 3
of this report discusses the Bureau's efforts at mitigating these
privacy concerns.)


--------------------
\8

\8 Statistical Abstract of the United States--The National Data Book,
Bureau of Census, Department of Commerce, October 1996. 

\9 Statistical Abstract of the United States--The National Data Book. 


      DECENNIAL DECISIONMAKING
-------------------------------------------------------- Chapter 2:1.2

For the first census, Congress delegated the 17 U.S.  marshals and
their 650 assistants to undertake the census, and gave them 9 months
to do so and report the results by district to the President.  For
each of the succeeding five censuses, Congress passed a new piece of
legislation.  These censuses were similar to the first, except that
the questions to be asked grew in number with each decade.  During
this 50-year period, Congress directed that some refinements to
census-taking be made:  the tallies were passed to the Secretary of
State starting in 1800, and enumerators used printed schedules for
the first time in 1830.  Congress also continued to authorize a small
clerical staff in Washington whose function was simply to check for
clerical errors in the work and compile the tabulations. 

In 1850, Congress created a new management structure for
administering the decennial census, which was becoming an
increasingly complex undertaking as more sophisticated questions were
being asked of a growing population spread over a wide geographical
area.  Congress created a Census Office and authorized a
superintendent of the census at a salary of $2,500.  Congress also
determined that the 1850 law would govern future censuses should
Congress fail to pass authorizing legislation.  This was done to
avoid the potential for a disruption in the census-taking schedule
and possible congressional deadlocks over particular issues. 

In the last decades of the nineteenth century, Congress began to
delegate more responsibility to the Census Office, which moved beyond
clerical functions and gained authority to control the field
administration of the census and appoint or approve the appointment
of supervisors and enumerators.  Until this time, the appointment of
those staff had been a matter of political patronage.  Furthermore,
despite the fact that census activities took almost 7 years to
complete in the late nineteenth century, the census offices that
Congress authorized every 10 years closed when the work of each
successive census was done.  In 1902 Congress established the Bureau
of the Census under the Department of the Interior as a permanent
agency, which, for the first time, would not disband between
censuses.  The Bureau was transferred to the newly created Department
of Commerce and Labor in 1903. 

By 1913, the Census Bureau was under the authority of the Department
of Commerce and gained its role as the preeminent census, survey, and
statistical agency of the United States, which it remains to this
day.  The Bureau not only conducts the decennial census as it did in
its early history, but also about 200 other censuses and surveys. 


      TITLE 13 GOVERNS THE BASIC
      RULES OF CENSUS-TAKING TODAY
-------------------------------------------------------- Chapter 2:1.3

While legislation passed in 1850 made a new authorization for each
decennial census unnecessary, Congress continued to pass legislation
every decade for implementation of upcoming censuses.  In 1954, title
13 of the U.S.  Code was enacted to establish the basic rules for the
taking of future decennial census, including the following: 

  -- The census, as required for apportionment, must be completed and
     reported to the president within 9 months after the census date
     of April 1;

  -- the Secretary of Commerce must submit to the committees having
     legislative jurisdiction over the census, not later than 3 years
     before the next census, the subjects proposed to be included in
     the coming census and the types of information to be compiled;
     and

  -- the Secretary must submit to the committees having legislative
     jurisdiction, not later than 2 years before the next census, the
     planned questions to be included. 

Although Congress delegated responsibility in title 13 to the
Secretary of Commerce to undertake a decennial census ï¿½in such form
and content as he may determine,ï¿½ Congress has maintained authority
and responsibility under the Constitution for directing the decennial
census.  Congress exercises a continuing role in overseeing the
conduct of the census through a number of congressional committees,
including for authorization, the Senate Governmental Affairs
Committee and the Subcommittee on the Census of the House Government
Reform and Oversight Committee\10 , and for appropriations, the
Commerce, Justice, State, and the Judiciary and Related Agencies
subcommittees in the House and Senate. 

While these committees and subcommittees provide general oversight,
Congress enacts legislation from time to time that contains specific
additional direction to the Secretary.  For example, in 1994, in
order to facilitate development of accurate address lists, Congress
enacted the Census Address List Improvement Act of 1994 that allowed
the Bureau and the U.S.  Postal Service to exchange address list
information under certain conditions. 


--------------------
\10 The House Government Reform and Oversight Committee established a
Subcommittee on the Census, effective February 1998, whose primary
function is oversight of the 2000 Decennial Census. 


      ADDITIONAL INFLUENCES ON
      CENSUS-TAKING
-------------------------------------------------------- Chapter 2:1.4

The Bureau and its predecessor entities have always been responsive
to congressional direction, but they have also been influenced by the
many users of its statistics.  In the nineteenth century, state
governments, scholars, business associations, and reformers were
among those who influenced the questions contained in the schedules,
and the censuses provided them data that helped them in their various
endeavors.  Professional statisticians have been and continue to be
influential in the Secretary's determination of the form and content
of the questions, as well as in decisions concerning the presentation
of data. 

In the late twentieth century, the influence of various interest
groups has had an effect on the census.  Advocates for the homeless
spurred the Bureau's efforts in the last several censuses to count
people who live in shelters and on the streets.  In the 1970s, the
Bureau created several advisory committees of experts involved with
minority issues.  Recently, racial and ethnic groups urged the Office
of Management and Budget (OMB) to reconsider the federal standards on
race and ethnicity classifications; their efforts resulted in the
1997 changes to those standards, which will allow individuals to
choose more than one racial category when completing their census
questionnaire.  The plans for tabulating and reporting of these new
racial categories by the Bureau continue to be a much debated issue. 


   DECIDING WHO TO COUNT
---------------------------------------------------------- Chapter 2:2

The Constitution identified who should be counted in the decennial
census in Article 1, Section 2 with the following language.  The
count "shall be determined by adding to the whole Number of free
Persons, including those bound to Service for a Term of Years, and
excluding Indians not taxed, three fifths of all other Persons
[slaves]."

Although the framers were specific about how to count (or not count)
Native Americans and slaves, they were not specific about whom to
count.  Only one important criterion for eligibility was established: 
ï¿½personsï¿½ rather than ï¿½citizensï¿½ were to be counted, meaning
citizenship was not to determine who should be counted.  There was
little reason to be more specific since the population in the 1780s
was relatively homogenous, stationary, monolingual, and organized in
stable household units. 

In the years since the framing of the Constitution, however, many of
those conditions have changed, posing new philosophic and pragmatic
issues.  Among the changed conditions are the following: 

1.  Illegal aliens:  No one in 1787 was an illegal resident because
the first laws controlling immigration were not passed until 1875.\11
The Immigration and Naturalization Service (INS) estimated there were
approximately 5 million illegal aliens residing in the United States
as of October 1996, with approximately 275,000 to 300,000 illegal
aliens arriving yearly. 

2.  Temporary and seasonal workers:  Nearly everyone who made the
journey to America to work in the eighteenth century stayed for
years.  Many were bound to do so by bonds of indenture or slavery,
while the broken ties with the distant homeland and the high cost of
returning encouraged others to stay.  In l990, the INS reported about
140,000 foreign citizens working temporarily in the United States in
a variety of occupations. 

3.  Homeless people:  Paupers without family to assist them and
depending therefore on the public benefice were few in number in the
1780s, and they were generally lodged at public expense in a
household where they would be counted.  In contrast, the Bureau
counted over 230,000 homeless persons during the 1990 Census.  The
number of homeless with no fixed address, however, is a matter of
conjecture, with other estimates ranging from 800,000 upward.\12

4.  Foreign visitors and U.S.  citizens living abroad:  While
short-term travel was not unheard of in the late eighteenth century,
the number of Americans living abroad and the number of foreign
citizens visiting in the U.S.  was insignificant.  In contrast,
throughout 1990, approximately 16 million foreign citizens visited
the United States for business and/or pleasure.  Furthermore, on
Census Day in 1990, about 1 million federal civilian and military
employees were living and working abroad.  (The Bureau's residency
rules generally do not include in the population count Americans
living abroad who are not federal or military employees, or foreign
visitors to this country.)

The lack of specificity in the Constitution about who should be
counted has raised questions over time about the eligibility of
certain categories of people.  When Congress passed the 14th
amendment in 1868, which modified Article 1, section 2, to eliminate
the language concerning slaves and indentured servants, Congress
debated whether to change the definition of those to be counted from
"persons" to ï¿½citizensï¿½ or ï¿½voters,ï¿½ but decided to keep the original
language.  The effect of legislation and court decisions over the
past centuries is that the language of Article 1, section 2, is read
at its most inclusive.  All persons who are resident in the United
States on Census Day, whether here legally or illegally, are to be
counted. 


--------------------
\11 Jeffrey S.  Passel and Karen A.  Woodrow, ï¿½The Judicial Basis for
Enumeration of Undocumented Aliens in the 1980 Census and
Implications for 1990ï¿½, U.S.Bureau of the Census, 1984. 

\12 "Homelessness:  Toward Another Decade of Homelessness?  An Issue
Paper," GAO/RCED, September 15, 1995. 


   DECIDING WHAT TO ASK IN THE
   CENSUS
---------------------------------------------------------- Chapter 2:3

The decennial census never simply counted heads.  Since the earliest
days of the republic, Congress directed the Bureau or its
predecessors to gather additional information as it enumerated the
population.  In the nineteenth century, the trend to greater numbers
of questions, which peaked with an encyclopedic number\13 in 1890 on
a large variety of issues, was inspired by the curiosity of a
self-conscious young nation and by the need to form public policy. 
In the twentieth century, the census questions have been increasingly
shaped by the need to fulfill the data requirements of programs
legislated by Congress and to properly allocate the federal funds
authorized by those programs. 


--------------------
\13 The 1890 Census contained inquiries on subjects that later became
separate censuses; these inquiries included the subjects of
agriculture, crime, insurance, mines and mining, manufacturing,
transportation, etc.  This census contained a total of 13,161
questions; the majority of households probably answered little more
than the 45 questions on population. 


      FROM SIMPLE TO COMPLEX
      QUESTIONNAIRE CONTENT
-------------------------------------------------------- Chapter 2:3.1

Even before the first census was taken in 1790, Congress considered
asking a range of additional questions, including one which would
determine individuals' military eligibility.  After debate, however,
Congress authorized enumerators to pose six questions:  the name of
the head of each family, the number of free white males over 16 and
under 16, the number of free white females, the number of other free
persons, and the number of slaves.  The 1800 and 1810 Censuses made
further distinctions among the ages of the free white respondents,
and the 1820 Census added distinctions for age and sex of the slave
and free black populations and also broke new ground in collecting
basic information about peoples' occupations.  The 1830 Census added
a count of the numbers of deaf, dumb, and blind household members,
and the 1840 Census added questions on literacy, schooling, and
revolutionary war pensioners.  This first period of census-taking
reflected the concerns of a new nation absorbed in its political
experiment and identity. 

In 1850, the question of what to ask in the census became highly
political as the nation debated how to handle the coming crisis
between the northern and southern states.  The focal point of the
debate was what level of detailed information to gather about slaves,
but the debate became a debate on the census itself and what was the
proper reach of the federal government.  At the same time, new
questions were asked that gathered information about schools, crime,
churches, and pauperism.  A growing national awareness about the
changing ethnic composition of the American population was reflected
in the census questions.  A question about unnaturalized foreigners
had been posed in the 1820 Census.  For the 1850 Census a question
was asked on the householders' place of birth by the identification
of the state, territory, or country where born and the birthplace of
parents. 

Immigration, and particularly immigration from southern and eastern
Europe, became a critical issue in American politics in the last
decades of the nineteenth century and the first decades of the
twentieth century, and the answers to census questions became a part
of the debate.  For the 1910 Census respondents were asked to
identify their mother tongue in a further effort to determine
individuals' ethnic background.  In 1921, such information gathered
over the decades was used when Congress enacted legislation that
ended America's historic policy of open immigration.  The law limited
immigration to 500,000 people per year and was to limit the
percentage of immigrants from any country to their proportional
representation in the 1910 Census.  That law's 1924 successor, the
National Origins Act, further cut immigration levels and returned to
the 1890 Census as the basis for immigration quotas. 

Until the 1930 Census, the details of the questions on the form were
specified minutely by Congress.  In the 1929 law authorizing the 1930
Census, Congress specified areas to be investigated but left, for the
first time, the exact questions to the Bureau. 

Unemployment was one of the areas that Congress directed the Bureau
to investigate in the 1930 Census.  As the economic crisis of the
1930s wore on, the need for more information with regard to the
population's socioeconomic condition increased as legislators and
government officials at the federal and state levels evaluated
existing programs and planned new efforts to deal with the
Depression.  The 1940 Census of Population and Housing included
questions on income, internal migration, and social security status,
as well as more refined questions on unemployment.  In addition,
Congress authorized a new set of questions about the types of
plumbing, heating, and appliances in peoples' dwellings. 

It became apparent prior to the 1940 Census that the amount of
information the Bureau was required to collect had come to exceed the
Bureau's ability to gather and tabulate it in an accurate and timely
manner.  As a result, the Bureau developed a new methodology for the
1940 Census and included supplementary questions that were asked of
only a portion of the population.  The Bureau's statisticians used
the data to extrapolate to the general population. 

For the 1950 Census, the Bureau moved toward limiting the decennial
census' primary focus to population, demographic, and housing
questions.  Many questions, such as those concerning unemployment,
moved to separate surveys and censuses often done at more frequent
intervals.  Today, the Bureau administers about 200 surveys related
to various economic and demographic issues. 

In 1960, the Bureau began to move toward the mail-out/mail-back
census questionnaires that we know today in order to eliminate
enumerator bias.  The nature of the population and housing questions
remained relatively constant from 1960 to 1990 with many
supplementary questions being asked of only a portion of the
households.  In 1960, for example, the Bureau asked 7 population and
14 housing questions on a short form questionnaire and posed an
additional 28 population and 30 housing questions on the long form
questionnaire sent to 25 percent of the households.  For the 2000
Census, an effort has been made to reduce the number of questions and
hence the burden on respondents.  The short form questionnaire is
currently designed to have 5 population and housing questions and the
long form questionnaire, which the Bureau plans to send to 17 percent
of the population, is currently designed with 45 additional
questions. 


      RACE AND ETHNICITY QUESTIONS
      TODAY
-------------------------------------------------------- Chapter 2:3.2

The census has collected data on race and ethnicity in a variety of
forms for 200 years.  Since the l960s, data on race and ethnicity
have been used extensively in civil rights monitoring and enforcement
covering such areas as employment, voting rights, housing and
mortgage lending, health care services, and educational
opportunities.  Over the last several years, however, the form of
those questions has been a topic of considerable debate within
American society. 

In the mid-1970s, OMB collaborated with other federal agencies to
standardize racial and ethnic data collected and published by the
federal government.  The result was OMB's 1977 Statistical Policy
Directive No.  15, which provided for classifications based on four
races--American Indian or Alaskan Native, Asian or Pacific Islander,
Black, and White; and one ethnicity--Hispanic Origin or Not of
Hispanic Origin.  These classifications applied to all federal
government data collection efforts and were often used by state
agencies.  The Bureau used these standard classifications too,
although in the 1980 and 1990 censuses, the questionnaire also
provided an "other response" category selection and a place where a
respondent could write in another category.  In addition, the
Bureau's long-form or sample form gathers more information on
ancestry. 

During the 1990s, these standards came under increasing criticism
from people who believed that the minimal categories set forth in
Directive 15 did not reflect the increasing diversity of the American
population resulting from growth in immigration and in interracial
marriages and who, therefore, urged changes in the standards.  Other
people, however, feared that changing the categories would decrease
the number of officially designated members of some racial and ethnic
groups, which might decrease the distribution of federal dollars
devoted to the programs designed to benefit those groups. 

An interagency group was convened in March 1994 to consider proposed
changes to the names of the groups, as well as several suggested
additions to the categories for race and ethnicity.  Another
suggested change the group considered was the addition of a
"multiracial" category.  On October 30, 1997, OMB issued revisions to
the standards of Directive 15.  The revised standards, which are to
be used by the Bureau for the 2000 census, have a minimum of five
categories:  American Indian or Alaska Native, Asian, Black or
African American, Native Hawaiian or Other Pacific Islander, and
White.  Two categories remain for ethnicity:  Hispanic or Latino and
Not Hispanic or Latino.  OMB also directed that forms, including the
2000 Census questionnaire, must tell respondents to ï¿½select one or
moreï¿½ categories to identify themselves.  By choosing multiple
categories, respondents can indicate a multiracial identity.  (For
information on the percentage of population by race and ethnicity,
see app.  II, table II.2.)


   COUNTING METHODS
---------------------------------------------------------- Chapter 2:4

The United States has always based its censuses on an enumeration of
persons residing in households as reported by one of its members. 
This self-enumeration method reflects an American commitment to a
minimally intrusive government and respect for individual privacy. 
In contrast to this method, China requires its people to report to
local government offices to register their existence, and Norway and
Denmark consolidate the records of various government agencies to
determine a population count.  In December 1997, the government of
Turkey conducted its latest quinquennial census whereby the entire
population is counted manually in one day over a 14-hour period. 
Citizens were required to stay home and be counted under threat of
punishment if found in public without special permission.  Starting
in 2000, the Turkish government plans to change to more modern
statistical procedures. 

For censuses prior to 1960, enumerators went door-to-door posing
census questions and recording the information.  While there was no
standardized recording process before 1830, thereafter enumerators
used a variety of standardized census forms to record the
respondents' answers.  With advances in technology, the compilation
of information for each new census became more sophisticated. 

After some testing in the 1960 Census, the Bureau began in earnest
during the 1970 Census to move away from the door-to-door census.  It
began mailing census questionnaires to households to be filled out
and returned.  Should households fail to mail back the census
questionnaires, enumerators follow up with telephone calls and
door-to-door visits.  As a last resort, enumerators solicited census
data from knowledgeable people such as an addressee's building
superintendent, letter carrier, or neighbors.  These last resort data
ranged in detail from just the number of people in the household to
all the information requested on the census questionnaire.  It should
be noted that nonresponse follow-up activities have become more and
more challenging as the public becomes less responsive to census
questionnaires.  It should also be noted that even though this
census-taking methodology has existed for just the last three
censuses, it is generally referred to as a traditional census. 

In addition to the nonresponse follow-up, three other particularly
challenging activities in conducting the last three censuses were (1)
identifying or obtaining correct addresses for households, (2)
enumerating people in nontraditional housing, and (3) encouraging
public participation. 


      ASSOCIATING ADDRESSES WITH
      HOUSEHOLDS
-------------------------------------------------------- Chapter 2:4.1

Finding all households and being able to geographically pinpoint
their location are important parts of the decennial census.  It is
the persons residing in those households who make up the population
counts of the United States, and it is the locations of households
that provide the population counts for smaller geographic areas, such
as states, congressional districts, counties, cities, and towns. 

In 1970, the Bureau changed its primary census-taking methodology
from door-to-door enumeration to mail-out/mail-back.  With this
change, the association of mailing addresses with households'
locations became more important.  The Bureau's strategy of
pinpointing the physical location of households, or geocoding,
continues to be important for 2000 for such procedures as nonresponse
follow-up. 

For prior censuses, the Bureau has constructed a new address list
from scratch.  For urban areas, the Bureau started with address lists
purchased from commercial vendors; for suburban and rural areas,
Bureau employees made a physical reconnaissance.  Many aspects of
American life make accurate identification of housing units
difficult, including rapid suburban and rural housing construction,
urban demolition and conversion, and the mobility of some housing
units, such as mobile homes and recreational vehicles.  Locating the
housing units of such diverse types in a country with an anticipated
118 million households in 2000 will be a labor-intensive and
expensive task. 

For the purpose of the census, every housing unit address is geocoded
to a census block whose size varies but which generally contains 30
to 85 people.  Through 1980, the Bureau relied on maps that had
hand-plotted census block boundaries and had new streets and features
drawn in by temporary employees or enumerators.  The resulting maps
could be rough and hard to read.  By the 1990 Census, a new
computer-generated mapping system--called the Topographically
Integrated Geographic Encoding and Referencing System (TIGER) was in
place.  TIGER is designed to locate every housing unit on one of the
seven million TIGER maps representing each census block.  The TIGER
maps can be easy to update and can be printed off the database by
many users other than the Bureau. 

In the past, the Bureau had not used U.S.  Postal Service address
lists to develop its own list for several reasons.  There was concern
about protecting individuals' privacy, and the Postal Service was
prohibited under title 39 of the U.S.  Code from sharing its lists. 
In addition, the Postal Service lists may not conform to the Bureau's
specialized needs.  For example, the Postal Service's addresses are
for mail delivery points and may not differentiate between more than
one household at an address, whereas the Bureau needs this household
differentiation at all addresses.  Furthermore, Postal Service post
office boxes or RFD addresses (which may not indicate the actual
location of a residence) cannot be used by the Bureau because
questionnaires must be delivered to actual household addresses in the
event followup becomes necessary. 

Nonetheless, Congress and the Bureau have recognized that cooperation
with the Postal Service can alleviate some of the cost and labor
burden in preparing for a census.  In 1994, Congress passed the
Census Address List Improvement Act, which allows the Postal Service
to share information with the Bureau.  The Postal Service now
notifies the Census Bureau of new and newly-obsolete addresses.  The
Census Bureau also provides the local governments with a list of
addresses in particular locales, so that they can point out
discrepancies with their own information.  To protect privacy,
however, the act specifies that only officials designated as census
liaisons can handle the Bureau's copy of their jurisdiction's address
list, which does not contain names of residents at the addresses, and
that the liaisons are prohibited from disclosing address list
information or using it for local purposes, such as identification of
illegal housing. 

For the 2000 Census, the Bureau is planning to rely on a Master
Address File (MAF), which is to be developed, in part, from the
Bureau's 1990 Census address list and the most recent Postal Service
address list (referred to as the delivery sequence files).  Under a
reengineering plan approved in September 1997, the Bureau also plans
to conduct a 100- percent canvass of all census blocks in early 1999
and will request the Postal Service to validate the city-style
addresses prior to the delivery of 2000 Census questionnaires.  For
the 2000 Census the Bureau is not going to rely on the Postal Service
to deliver questionnaires to non-city style addresses as they did in
1990.  Instead, the Bureau is planning for enumerators to deliver the
questionnaires and ask that they be mailed back.  Furthermore, the
Bureau has no plans to purchase addresses from commercial vendors as
it did in the prior three censuses.  Vendors' lists were found to be
less accurate in low-income areas, which are not a high priority for
companies selling goods and services. 


      LOCATING NONTRADITIONAL
      HOUSING
-------------------------------------------------------- Chapter 2:4.2

Because the Bureau's basic data collection method revolves around
households, counting people who do not live in traditional households
can be especially difficult.  Such people live in group quarters,
such as shelters for battered women and the homeless; nursing homes;
college dormitories; migrant worker camps; military installations;
and remote areas.  It takes special efforts to count these
individuals. 

In the 1990 Census, the Bureau tried a Street and Shelter Night
program to count the homeless wherever they could be found on a
particular night.  In 2000, the Bureau will focus its efforts to
count the homeless on the places where many of them come for
services, such as shelters and soup kitchens, as well as targeted
outdoor locations.  The emphasis will shift from finding the homeless
on street corners to identifying them through the organizations that
assist them.  Other nontraditional procedures include cooperation
between the Bureau and the Department of Defense and the U.S.  Coast
Guard to count individuals on military installations.  Another
special operation will count highly transient individuals living at
recreational vehicle parks, commercial or public campgrounds, and
marinas.  Remote areas of Alaska will be enumerated in mid-February,
a time when the difficult travel to these areas by dogsled and
snowmobile is somewhat easier rather than on April 1. 

Another way that the Bureau plans to count individuals in
nontraditional households is by making census questionnaires
available at public places, such as post offices and community
centers.  In this way, people who did not receive a mailed
questionnaire will have a greater chance to be counted.  This new
approach does introduce a higher risk, which the Bureau continues to
assess, of multiple responses for a given household or person. 
"Unduplication" formerly required a massive clerical operation, but
now the Bureau expects that advances in computer storage, retrieval,
and matching, along with image capture and recognition, will give the
Bureau a much greater ability to eliminate duplicative responses. 

Several lawsuits alleging undercounts of the homeless were filed
against the Bureau following the 1990 Census.  Despite the efforts
planned for the 2000 Census, the count of the homeless and other
people living in nontraditional households is likely to be less
accurate than for those living in housing units that can be plotted
on a TIGER map.  Issues of possible undercounts of people living in
nontraditional households will likely surface again in the 2000
Census. 


      ENCOURAGING PUBLIC
      PARTICIPATION
-------------------------------------------------------- Chapter 2:4.3

In 1970, 78 percent of the households who received a mailed
questionnaire fill it out and returned it; in 1990, that percentage
dropped to 65 percent.  Based on the response rate for other surveys
in the meantime, the mail return rate for the 2000 Census could be
even lower.  This decline in the mail response rate poses not only an
enumeration challenge to the Bureau, but also a major financial
problem.  The cost of eliciting responses from the 34 million
households that failed to return their questionnaires in 1990 was
$730 million.  Nonresponse follow-up was one of the most costly
operations of the 1990 Census.  Encouraging voluntary public
participation, therefore, is a major objective of the Bureau. 

Lack of awareness of the census was not a major problem in the 1990
census.  Apathy, concerns over loss of privacy, and fears that census
information might be shared with other government agencies, however,
were major impediments to achieving high rates of returned
questionnaires.  To encourage the public to mail back the
questionnaires, the Bureau spent approximately $75 million on
promotion and outreach in 1990, and received pro bono promotional
services valued at $65 million from the Advertising Council--a
nonprofit organization that administers public service advertising
campaigns.  The Bureau reached the public through the media and
through coordinated efforts with local and state governments,
national and community organizations, and business and religious
entities.  However, because the Bureau had little control over when
or where the Advertising Council disseminated the Bureau's message,
it has decided to use a paid advertising campaign in 2000 to
complement its continuing efforts with its organizational partners. 
The Bureau estimates the cost of all outreach and promotion
activities will be about $230 million for the 2000 Census. 

In order to improve the mail response rate in 2000, the Bureau is
planning to use a new, multiple mail contact strategy.  The Bureau
plans to increase the number of its mail contacts with households by
sending out a letter notifying households of the coming
questionnaire, an initial questionnaire, a thank you or reminder
card, and possibly a replacement questionnaire.  Both the initial and
any possible use of a replacement questionnaires will be barcoded to
minimize counting duplicate submissions.  In areas lacking city-style
addresses, either the Bureau or the Postal Service will implement
segments of this strategy.  The multiple mail contact strategy was
used in the 1995 Test Census and showed a potential for increasing
the mail response rate.  Multiple mail contact will also be tested
during the 1998 dress rehearsal. 

Language barriers can be an obstacle to gathering a full count of the
population.  During the 1970 Census, despite the fact that 9.2
million U.S.  residents spoke Spanish in their homes, the census
questionnaire was not printed in Spanish.  Since then, the Bureau has
tried to remove that obstacle by printing the questionnaires in both
English and Spanish, hiring enumerators with foreign language skills,
and providing toll-free telephone assistance in languages other than
English.  In 1990, census questionnaire guides were available in 32
languages.  For 2000, the Bureau is researching the use of
questionnaires in additional languages. 

Certain racial and ethnic minorities have long been undercounted in
the census.  Language barriers, fears of deportation, and a greater
tendency to live in nontraditional households are factors that have
led to this undercount.  In the 1970s and 1980s, the Bureau
established advisory committees on the Hispanic, African-American,
Native American, and Asian and Pacific Islander populations, to help
the Bureau find ways to improve its count of these groups.  Those
advisory committees will continue to function for the 2000 Census,
but the obstacles to increasing minority participation in the census
have not been eliminated. 

There are categories of people who have incentives to avoid
participating in the census.  Individuals who are in the U.S. 
without the proper documentation or who otherwise have reason to fear
various law enforcement or regulatory government agencies are
unlikely to be convinced to be counted. 


   SAMPLING AND STATISTICAL
   ESTIMATION IN THE 2000 CENSUS
---------------------------------------------------------- Chapter 2:5

The Bureau plans to use two new sampling procedures in the 2000
Census.  One is designed to reduce the time required for and expense
of following up on the projected 40 million housing units that may
not respond in 2000 to the questionnaires.  The other, referred to as
Integrated Coverage Measurement (ICM), is designed to adjust the
population counts obtained from census questionnaires and nonresponse
follow-up procedures to eliminate the endemic differential
undercount. 


         NONRESPONSE FOLLOW-UP
         SAMPLING
------------------------------------------------------ Chapter 2:5.0.1

As in the previous three censuses, the Bureau plans to encourage
households to mail back the questionnaires that have been mailed to
them or left at their homes.  Four weeks after Census Day, the Bureau
plans to implement a procedure known as nonresponse follow-up (NRFU)
to collect information from households that have not returned their
forms.  The 2000 procedure departs from previous censuses in that it
incorporates sampling to select the housing units that the Bureau
will contact for NRFU.  A sample of housing units is to be selected
in each census tract, sorted by geography and form type (long versus
short form) to make sure that the sample is distributed evenly across
nonresponding housing units in each tract.  Each sample is to be
selected immediately after the cutoff date for mail returns, and
households are to be selected in sufficient numbers to ensure that
the number of housing units in the sample, when added to the
households that have voluntarily returned their forms, will total at
least 90 percent of households in the tract.  Data for households not
in the sample is be imputed by a systematic procedure that relies on
data collected from geographically contiguous households. 

To illustrate, in a tract where 70 percent of the households
responded to the census questionnaire, the Bureau would draw a
two-thirds sample (to reach 90 percent) of the remaining households. 
They would then use results of this follow-up enumeration to impute
characteristics of the households not selected for the sample. 

There are to be several exceptions to this procedure.  First, all
housing units in blocks that have been selected for the Integrated
Coverage Measurement survey are to be contacted (100-percent
nonresponse follow-up).  Second, nonresponse follow-up is not to be
conducted in rural households that are listed by enumerators.  Data
from these households are to be collected by enumerators listing the
housing units.  Third, late data submitted voluntarily by a household
is not to be thrown away and is to be used in preference to data
either collected by an enumerator or imputed by NRFU, if the
questionnaire is received before the completion of NRFU data
collection activities. 


         USING ICM TO ADJUST
         POPULATION COUNTS
------------------------------------------------------ Chapter 2:5.0.2

The purpose of ICM is to adjust for errors that occur in
census-taking.  (Errors in past censuses ares discussed below under
Accuracy of Past Censuses.) In general, ICM is a statistical
procedure that would be used in an effort to improve the accuracy of
the original data collected by the census by reconciling that data
with data obtained from an independent sample of 750,000 households. 
The reconciliation process, referred to as Dual System Estimation,
applies probability theory to the ICM and the census figures to
generate a third, better estimate of the true population.  ICM would
be conducted after basic data collection, including nonresponse
follow-up, had ended and would estimate the extent to which people
were correctly counted, missed, or included by error in the census. 

Since ICM would be the last step in the census process and its
results would be an integral part of the final census numbers, the
Bureau plans to release only one set of official census numbers.  In
order to accomplish this ï¿½one-number censusï¿½, the Bureau is planning
for both nonresponse follow-up and ICM to be completed quickly so
that it can announce results by December 31, in time to meet the
deadline for reporting census data for apportionment purposes.  The
use of laptop computers during the post-enumeration interviews is
planned to speed the process of reconciling differences between the
census and ICM data.  The cost of ICM, including the laptop
computers, is projected to be $325 million. 


         CONSTITUTIONAL AND LEGAL
         ISSUES ON SAMPLING
------------------------------------------------------ Chapter 2:5.0.3

Article I, section 2, of the Constitution refers to an ï¿½actual
Enumerationï¿½ of the population for the census.  It also vests
Congress with the authority to conduct censuses ï¿½in such a Manner as
they shall by Law Direct.ï¿½ Congress has, in turn, delegated this
authority to the Department of Commerce through title 13 of the U.S. 
Code.\14 The question now being debated is whether the latitude
allowed the Secretary of Commerce includes the use of the statistical
methods proposed for 2000, provided the Secretary determines that so
doing would improve the census accuracy, or whether the requirement
for an ï¿½actual Enumerationï¿½ limits that discretion.  The issue also
has a similar statutory incarnation:  Title 13 of the U.S.  Code
states that the Secretary of Commerce is to undertake a decennial
census in such form and context as he may determine, including the
use of sampling procedures, yet it excludes authority to use sampling
in the determination of population for purposes of apportionment of
representatives in Congress.\15

The question of whether sampling is statutorily and constitutionally
permissible in determining the decennial census count can only be
definitively resolved by the Supreme Court.  The Supreme Court has
not yet considered the specific issue of whether the use of sampling
violates the Constitution and, in the course of considering past
challenges to the conduct of the census, has specifically stated that
its rulings were not to be construed as either prohibiting or
allowing the methods. 

Sampling through the use of the long form questionnaire to obtain
demographic information has become an unremarkable part of late
twentieth century census-taking in America.  However, the possible
use of sampling and statistical estimation to adjust the 1990 census
population count raised fundamental constitutional and statutory
issues that continue to be debated today.  The resolution of these
issues is now essential to the completion of planning for the 2000
Census. 


--------------------
\14 13 U.S.C.  141(a). 

\15 13 U.S.C.  195. 


   ACCURACY OF PAST CENSUSES
---------------------------------------------------------- Chapter 2:6

Ever since George Washington questioned the results of the first
census in 1791, the accuracy of any given census has been in
question.  The questions have always been legitimate:  The census has
never counted 100 percent of those it should, in part because
American sensibilities would probably not tolerate more foolproof
census-taking methods, such as requiring residents to register with a
central governmental authority.  In addition, some percentage of the
populace has always chosen to evade census-takers out of fear. 
Others have gone, and will continue to go, uncounted because there is
an incongruence between the Bureau's primary means of locating
individuals and particular individuals' circumstances.  For example,
in the nineteenth century, isolated homesteaders were difficult for
enumerators to locate and count.  Today, young urban males are
especially likely to be missed in an enumeration process based on
associating people at fixed household addresses. 

Until the 1940s, there were no means to answer the question of how
inaccurate a particular census had been, or at least no means less
prone to inaccuracy than the census itself.  The Bureau began to
evaluate census coverage in the 1940s, at first based on comparisons
of birth and death certificates and other administrative data--a
procedure known as demographic analysis.  The Bureau also began to
use statistical methods based on sampling, a method that involves
using a representative part of a population to convey information
about a whole population.  Since 1940, the Bureau has quantified the
amount by which any census undercounts the population.  (See app. 
II, table II.3 for net undercount estimates.\16 )

Measures of the total undercount have been possible since 1940 with
demographic analysis, but detailed measures of the differences among
undercounts of particular ethnic, racial, or other groups have only
become available since 1980.  The statistics reveal that some
subgroups of the population are counted less completely than others. 
The availability of the data and the fact that not only
representation but also allocation of federal resources is at stake
has made the composition of the undercount a sensitive and widespread
concern. 

In the 1940 census, the Bureau instituted its first effort to gain
accurate information through sampling.  The Bureau responding to
pressure to add a multitude of questions on unemployment, housing,
and income, among others, developed a set of supplementary questions
that were asked of only 5 percent of the population.  The Bureau
statisticians, using newly-developed statistical methods, used those
answers to extrapolate to the general population.\17 This statistical
method continues in use today, although the percentage of households
receiving what is now known as the long form questionnaire is to be
17 percent. 

The Bureau has evaluated the magnitude and characteristics of census
errors and undercounts for decades, but it has never used the
findings of these evaluations to actually correct coverage errors. 
In 1990, a survey called the Post Enumeration Survey (PES) was used
to determine the error in the 1990 census.  After the census was
taken, PES enumerators interviewed a sample of 5,000 census block
clusters containing 150,000 households, and matched by name the
people counted in the PES with those counted in the census.  The
extent to which housing units and people were correctly enumerated,
missed, or counted in error was used to estimate error for the entire
census.  Rates of error were then determined for 1,392 various types
of people or post strata in the population and applied to every
person counted in the census.  The post strata were based on such
characteristics as age, sex, race, ethnicity, location, and status as
renter versus owner of housing.  Thus, for example, if Asian and
Pacific Islander females between the ages of 18 and 29 were found to
be undercounted by 1 percent, an adjusted census would have counted
each person in that post stratum as 1.01 persons. 

Matching had been tried in the post-enumeration effort of 1980, but
the computer technology was not sufficiently sophisticated to base an
adjustment on the effort.  The quality of the data improved in 1990,
but the Secretary of Commerce determined that the evidence to support
an adjustment was inconclusive and decided not to adjust the 1990
census results.  The decision whether to adjust the census with the
results of PES was complicated by the fact that the 1990 census
figures had already been released when the PES figures became
available in the spring of 1991.  The Secretary of Commerce expressed
concern that having two sets of numbers could create confusion and
might allow political considerations to play a part in choosing
between sets of numbers when the outcome of the choices, such as
differences in apportionment of seats in Congress, can be known in
advance of a decision. 


--------------------
\16 The ï¿½net undercountï¿½ reported does not reveal the even larger
errors in the census.  The net undercount is the difference between
those missed and those counted twice.  In 1990, for example, the net
undercount was 4 million.  The number, however, of persons missed was
10 million, and the number of persons counted twice was 6 million,
making the gross error 16 million. 

\17 Statistical analysis was critical to the 1950 Census in a
different way when statistical analysis by the Bureau revealed that
enumerators were introducing errors into the census results.  This
realization was one of the factors that led the Census Bureau to
introduce a 1960 test of the method of self-enumeration via mailed
questionnaires that has become the standard today. 


PROTECTING PRIVACY AND CONTROLLING
COSTS ARE PERSISTENT CENSUS
CONCERNS
============================================================ Chapter 3


   PROTECTING PRIVACY
---------------------------------------------------------- Chapter 3:1

Title 13 of the U.S.  Code prohibits the Bureau and its employees
from releasing or allowing anyone other than Department of Commerce
employees to examine individual census records.  Penalties of up to
$5,000 and 5 years in prison for violating the provisions of the
Title apply.  Despite the Bureau's strict policies, stringent
penalties, and i ts modern record of conscientious defense of the
confidentiality of its records against a number of agencies and
groups that have sought to obtain certain records, some portion of
the population fails to respond to the census, or responds
reluctantly, out of fear that their personal information will find
its way into the public domain. 


      THE EVOLUTION FROM PUBLIC
      POSTING TO STRICT PRIVACY
-------------------------------------------------------- Chapter 3:1.1

From 1790 through 1840, the censuses were entirely public.\18 In
fact, during this period, the census results by household were posted
"in two of the most important places" in the enumeration districts by
Congress' express direction.  The purpose of the posting was to allow
omissions and errors to be caught by districts' residents. 

After the 1840 Census, census results were no longer publicly posted,
but there was no law formally safeguarding the confidentiality of the
information.  Bureau policy, however, as enunciated by the Secretary
of the Interior in 1850, was that the returns were to be used
"exclusively for the use of the government, and not to be used in any
way to the gratification of curiosity, the exposure of any man's
business or pursuits, or for the private emolument of the marshals or
assistants." However, because the originals of the census were given
to local officials, the security of the returns could not always be
ensured. 

The 1880 Census Act included major changes with regard to privacy. 
Enumerators were required to swear an oath not to disclose any
information to anyone except their supervisors, and census returns
were no longer given to local officials, but were filed instead with
the Department of the Interior.  Business information was protected,
but information related to individuals was not.  That information was
available at the discretion of the Director of the Census for a fee. 

In the early 1900s, the Bureau focused on a different threat to
confidentiality, which was the potential that businesses might, by
analyzing aggregate pieces of information provided at the local
level, deduce the identity of their competitors and information about
them.  The 1910 Census Act prohibited the Bureau from publishing data
from which a business might be identified. 

The discretion given the Director of the Census to release
information related to individuals allowed Civil War veterans to
obtain information that helped them prove their age and status for
pension purposes at a time when census records might be the best or
only source of official information.  World War I era men received
information from the Bureau to prove that they were too young to be
eligible for the draft.  Exercising the same discretion, the Director
agreed in 1917 to supply federal officials with the names and ages of
individuals potentially eligible for the draft, and in 1921, the
Director approved the provision of information to private
institutions promoting literacy that wanted to use Bureau records to
identify illiterate people in the nation. 

Later in the 1920s, the Bureau, following the guidance of the Justice
Department, began to narrow the circumstances under which information
could be released.  In 1930, it denied access to a federal agency
called the Women's Bureau, which wanted the names, addresses, and
occupations of some women.  In 1942, the Bureau turned down the War
Department's request for the names and addresses of people of
Japanese descent living in the West--although the Bureau did identify
geographic concentrations of Japanese. 

The new practice of thoroughly restricting access by private or
public entities to census records was codified in title 13 of the
U.S.  Code in 1954.  The Supreme Court, citing title 13, ruled in
1982 that the Bureau could not even release its address list without
names to the City of New York so that city officials could compare
their lists with the Bureau's.  Subsequently in 1994, Congress passed
the Census Address List Improvement Act which allows the Bureau to
share address list information with local governments as part of its
decennial address list development procedures. 

Title 13 assures complete confidentiality for all records in the
Bureau's custody.  Once the records are passed to the custody of the
National Archives, the Archives can then release them for public use
when the records are 72 years old.  For records not yet in the
Archives, individuals can, for a fee, get a copy of their own record
or their minor child's record, but for anyone else they must have a
signed authorization.  For a deceased person, a death certificate or
similar evidence must be submitted, as well as proof that the
applicant is either a direct blood-line descendant or an heir of the
person requesting the information. 


--------------------
\18 "Privacy and Confidentiality in the U.S.  Census--A History",
Bohme, Frederick G.  and David M.  Pemberton, History Staff, Bureau
of Census, Department of Commerce, 1991. 


      PRIVACY ISSUES TODAY
-------------------------------------------------------- Chapter 3:1.2

Title 13 provides for strict confidentiality and substantial
penalties for deliberate release of individuals' information.  The
Privacy Act of 1974 does not apply to census records.  The Bureau has
long been concerned about inadvertent release of information about
individuals via published data that could be analyzed in a way to
reveal a particular respondent's data.  The Bureau has procedures to
prevent the possible identification of a particular household's data,
especially when it is cross-tabulated with other information. 

In order to prevent such an accidental release of economic
information, in earlier days, the Bureau would visually inspect the
data before release and, when necessary, would collapse it into broad
categories or delete information from certain cells in tables. 
Today, the Bureau uses computer programs to ensure that information
cannot be analyzed to reveal individuals' information.  Additional
techniques, such as random rounding or exchanging household
statistics among census blocks, are being studied to avoid potential
problems. 

Nonetheless, in an age when many people feel anxious about the reach
of marketers, poll-takers, and others who come armed with
computer-based data about individuals, the concern over privacy and
confidentiality will be hard to vanquish, and its effect on
census-taking will not easily be mitigated.  Some part of the
declining response rate is a function of people's anxiety about what
will become of the data they provide to the Bureau. 


   RISING CENSUS COSTS
---------------------------------------------------------- Chapter 3:2

The cost of the census has steadily increased over the course of 200
years.  The rate of increase has been dramatic in the twentieth
century:  the 1960 Census cost $523 million,\19

and the 1990 Census cost $2.6 billion--an increase of 400 percent
after adjusting for inflation.  The rate of increase will continue to
escalate with the 2000 Census, which is projected to cost $4 billion. 
Three major factors are involved in the soaring costs over the last
40 years:  an increase in the number of housing units to be
enumerated, an increasing use of expensive technology, and the number
of staff needed to take a decennial census. 


--------------------
\19 \19 The actual cost of the 1960 Census was $120 million, but
using 1990 constant dollars the cost is $523 million.  In addition,
the 1960 Census was not conducted using primarily mail-out/mail-back
questionnaires as was the 1990 Census. 


      INCREASING NUMBER OF HOUSING
      UNITS
-------------------------------------------------------- Chapter 3:2.1

Rapid population growth has been one of the hallmarks of the American
national experience.  The country's population has grown over 10
percent on average per decade since the 1960 Census, and that fact
has contributed to the ever increasing price tag of the decennial
census.  In recent decades, however, an additional factor has been
important:  the increasing rise in the number of housing units. 

While the number of people in the country has been increasing, there
have been fewer people living in the average household.  The rate of
increase in the number of households therefore has been rising at an
even quicker pace than the population.  In 1960, the Bureau counted
people at nearly 60 million housing units; in 1990, it counted people
at 102.3 million housing units, a 75-percent increase in the number
of units to be either contacted by mail or visited or both.  For
2000, the Bureau estimates there should be 118.6 million housing
units that need to be contacted. 


      TECHNOLOGY
-------------------------------------------------------- Chapter 3:2.2

The Bureau has been a leader in the use of automation technology and
electronic data processing methods for nearly a century.  The Bureau
needed to be inventive because as the population grew and the census
questions and the possible answers to them became more numerous, the
decennial census required increasing numbers of clerks to tally and
cross-tabulate the responses.  To conduct the 1860 Census, the Bureau
had 184 office staff and 4,417 field enumerators and produced 3,189
pages of census reports.  In 1890, the census effort required just
over 3,143 office workers, 46,804 field enumerators, and the pages
published numbered 26,408.  The census was also taking ever longer to
complete as the amount of data collected increased. 

All the tallying of the 1880 Census was still done by hand, and the
Bureau recognized then that the solution to what was becoming a data
crisis was mechanization.  Herman Hollerith, a former Bureau
employee, developed an electrical enumerating machine with its punch
cards for the 1890 Census.\20 With 105 of Hollerith's machines, the
1890 census was completed in 3 years, as opposed to the 7 years it
took to complete the 1880 Census. 

In 1946, the Bureau contracted with a private firm, the
Eckert-Mauchly Computer Corporation, to design a machine for its
statistical purposes that would use electrical impulses rather than
mechanical holes to tabulate census responses.  The machine, known as
UNIVAC, had a processing unit containing 18,000 vacuum tubes and was
delivered in 1951.  Although it was too late for processing much of
the 1950 Census data, it proved the concept and was a precursor to
much greater use of computer processing in subsequent censuses. 
Since 1950, the Census Bureau has taken advantage of improvements and
additional capabilities in electronic data processing developed
during the previous decade. 

While the punch card automation systems of the late nineteenth and
early twentieth centuries saved enormous clerical labor costs, they
were also a practical necessity:  the time needed for completing a
census tabulation was approaching 7 years.  The benefits of
technology in the second half of the twentieth century have produced
some savings in labor costs, but the major benefits have been in
faster census data processing and improved data analysis and
accuracy.  For example, the TIGER maps developed for the 1990 Census
integrated maps, addresses, and other geographical information, thus
solving most problems of inconsistency. 

The 2000 Census will rely on computer technology to a greater extent
than ever before, but most improvements are not primarily aimed at
cost reduction.  For example: 

  -- The census address list and the geographic file will be
     integrated to assist enumerators in finding housing units. 

  -- By providing census data electronically directly on the Internet
     and through libraries, universities, and the Bureau's Data
     Centers, the Bureau intends to make more data available faster
     to the public than ever before. 

  -- The improved data recognition software to be introduced in 2000
     will facilitate the processing of enumeration data. 

  -- The use of laptop computers pre-loaded with census data for the
     enumerators to use during the post-enumeration interviews for
     ICM is planned to speed the process of reconciling differences
     between the census and ICM data. 

While there is a cost to implementing these technologies, the benefit
is wider, faster distribution, and therefore use, of public census
data; greater accuracy and fuller coverage of the census; internal
Bureau efficiencies; and, to a lesser degree, reduction of labor
costs. 


--------------------
\20 Hollerith went on to establish the Hollerith Tabulating Company
to manufacture machines and sell punch cards and later merged his
company with others making time clocks, scales, and grinders.  In
1924, the merged companies were renamed International Business
Machines. 


      STAFFING THE CENSUS
-------------------------------------------------------- Chapter 3:2.3

Taking a decennial census is a very labor intensive and costly
endeavor.  Over the decades, as the population has grown, so have the
different types and numbers of workers needed to complete and report
a census on time.  (App.  II, table II.3 provides information on the
growth of the 21 completed decennial censuses by listing the
population, enumerator staff, office and headquarters staff, and the
actual cost of each census.)

Decennial census staffing can be generally divided into three
different categories--field enumeration, local census and census
field offices, and headquarters staff.  The majority of the field
staff are enumerators and supervisors whose primary job is verifying
addresses prior to a census, doing nonresponse follow-up during the
initial census taking; and doing post-enumeration surveys.  The
majority of enumerators work from 6 to 10 weeks and are paid a few
dollars over minimum wage.  Enumerators do not receive regular
federal employees benefits but have been eligible for unemployment
compensation during past censuses.  Partly because of the high
turnover rate of enumerators during prior censuses, the Bureau plans
to employ part-time workers who may have another job.  The Bureau
estimates it will need over 300,000 field staff working out of 520
local census offices and 402 census field offices for the 2000
Census. 

Local census and census field office staff generally consists of a
variety of occupations, such as clerks, lower to mid-level managers,
data processors, and data scanning operators.  Some of these local
census office staff may be employed for up to 11 months.  Temporary
office staff are generally paid at rates similar to those of
full-time federal employees.  The Bureau has 4 processing offices and
12 regional offices whose primary mission every 10 years becomes
taking the decennial census.  The Bureau estimates it should need
several thousand employees for its local census, census field, and
processing offices in 2000. 

Headquarters staff consists primarily of managers and analytical
staff, such as the Director of the Bureau, high-level managers,
lawyers, computer programmers, statisticians, demographers,
advertising experts, and writers.  The types and amounts of pay
received by headquarters staff is as divergent as the many different
occupations needed to take and report a census.  Several thousand
full-time employees from Bureau headquarters are expected to work on
the 2000 Decennial Census. 

Sufficient staff may allow the Bureau to meet its stated goals of
producing an accurate and timely one-number census in 2000.  But,
doing so also could be costly since it requires the Bureau to
undertake many labor-intensive procedures and special activities to
ensure that all residents of the United States are counted and
included in the 2000 Census. 


MOST FREQUENTLY ASKED QUESTIONS
AND ANSWERS ABOUT THE DECENNIAL
CENSUS
=========================================================== Appendix I

Question                Answer
----------------------  ----------------------------------------------
Why is a census count   For constitutionally mandated reapportionment
taken?                  and other statutory requirements.

What is                 Allocation of the 435 members in the House of
reapportionment?        Representatives among the states according to
                        the population from the decennial census.

What is the latest      248.7 million people per the 1990 Census.
count?

Who is counted?         All persons residing in the United States
                        regardless of their citizenship status.

How long does it take   Nine months with extensive research,
to count and report     pretesting, and planning.
results?

How much does it cost?  $2.6 billion for the 1990 Census; about $4
                        billion is the projected cost for the 2000
                        Census.

How many workers are    Over 500,000 temporary workers and about 6,800
needed?                 permanent employees were used during the 1990
                        Census.

How is the census       Mostly self-enumeration by mail using
taken?                  standardized short and long form
                        questionnaires and door-to-door followup for
                        nonresponding households.

What questions are      Population, economic, demographic, and housing
asked?                  issues.

Are statistical         The long form or sample questionnaire is
procedures used?        filled out by one-sixth of the population and
                        used to project national results. The Census
                        Bureau plans to use statistical sampling and
                        estimation procedures in 2000.

At what levels are the  Computerized data are available by State,
results available for   county, city,
public uses?            municipalities, etc., in various enumerations.

What publications are   Standardized and special publications are
available?              available on paper, computer tapes, CD-ROM,
                        and the Internet.
----------------------------------------------------------------------
Note:  The plans for the 2000 Census are not final.  Depending upon
decisions made by the Bureau and Congress, operational procedures may
change and costs must be adjusted. 



                               Table II.1
                
                    Changes in the Apportionment of
                       Membership of the House of
                  Representatives Between the 1920 and
                 1990 Decennial Censuses, by Region of
                                Country

                                        Decennial census
                          --------------------------------------------
                                                              198  199
Regions/divisions         1920  1930  1940  1950  1960  1970    0    0
------------------------  ----  ----  ----  ----  ----  ----  ---  ---
Northeast                  123   122   120   115   108   104   95   88
New England                 32    29    28    28    25    25   24   23
Middle Atlantic             91    93    92    87    83    79   71   65
Midwest                    143   137   131   129   125   121  113  105
East North Central          86    90    87    87    88    86   80   74
West North Central          57    47    44    42    37    35   33   31
South                      136   133   135   134   133   134  142  149
South Atlantic              56    54    56    60    63    65   69   75
East South Central          39    34    35    32    29    27   28   27
West South Central          41    45    44    42    41    42   45   47
West                        33    43    49    59    69    76   85   93
Mountain                    14    14    16    16    17    19   24   24
Pacific                     19    29    33    43    52    57   61   69
======================================================================
Total for United States    435   435   435   437   435   435  435  435
----------------------------------------------------------------------
Source:  United States Summary, Population and Housing Unit Counts,
Bureau of Census, Department of Commerce, Aug.  27, 1993. 



                               Table II.2
                
                Changes in the United States Population
                and its Undercount by Race and Ethnicity
                  Between the 1950 and 1990 Decennial
                                Censuses

                                          1950  1960  1970  1980  1990
----------------------------------------  ----  ----  ----  ----  ----
Population distribution
 (percentage)
White (non-Black)                         89.3  88.6  87.6  79.9  75.7
Black                                      9.9  10.5  11.1  11.5  11.8
American Indian, Alaska Native              NA    NA    NA   0.6  00.7
Asian Pacific Islander                      NA    NA    NA   1.6  02.8
Hispanic                                    NA    NA    NA   6.4  09.0
Other \                                    0.7   0.9  01.3    --    --
Net undercount estimates (percentage)
White (non-Black)                          3.8   2.7   2.2   0.8   1.3
Black                                      7.5   6.6   6.5   4.5   5.7
American Indian, Alaska Native              NA    NA    NA    NA   4.5
Asian Pacific Islander                      NA    NA    NA    NA   2.3
Hispanic                                    NA    NA    NA    NA   5.0
----------------------------------------------------------------------
Source:  Statistical Abstract of the United States--The National Data
Book, Bureau of Census, Department of Commerce, October 1996.

Economic and Statistics Administration, Bureau of Census, Department
of Commerce, November 1996. 



                               Table II.3
                
                Growth of the Decennial Census from 1790
                                to 1990

                                                 Number of  Total cost
                                                headquarte   of census
                        Total U.S.   Number of   rs and/or  (thousands
                        population  enumerator      office          of
Census year             (millions)       staff       staff    dollars)
----------------------  ----------  ----------  ----------  ----------
1790                           3.9         650          \a         $44
1800                           5.3         900          \a          66
1810                           7.2       1,100          \a         178
1820                           9.6       1,188          \a         208
1830                          12.9       1,519          43         378

1840                          17.1       2,167          28         833
1850                          23.2       3,231         160       1,423
1860                          31.4       4,417         184       1,969
1870                          38.6       6,530         438       3,421
1880                          50.2      31,382       1,495       5,790

1890                          63.0      46,804       3,143      11,547
1900                          76.2      52,871       3,447      11,854
1910                          92.2      70,286       3,738      15,968
1920                         106.0      87,234       6,301      25,117
1930                         123.2      87,756       6,825      40,156

1940                         132.2     123,069       9,987      67,527
1950                         151.3     142,962       9,233      91,462
1960                         179.3     159,321       2,960     127,934
1970                         203.2     166,406       4,571     247,653
1980                         226.5     458,523       4,081   1,136,000

1990                         248.7     510,200       6,763   2,600,000
----------------------------------------------------------------------
\a There was no official headquarters staff for the first four
censuses.  In addition, the records for the 1790, 1800, and 1810
Censuses were accidentally destroyed; the numbers shown are
estimates. 

Source:  The Story of the Census--1790-1915, Bureau of Census,
Department of Commerce, 1915.

The Bureau of Census, adapted by A.  Ross Eckler, 1972.

American Census:  A Social History, Anderson, Margo J., 1988.

Modernizing the U.S.  Census, National Research Council, 1995.

Statistical Abstract of the United States--The National Data Book,
Bureau of Census, Department of Commerce, October 1996. 


MAJOR CONTRIBUTORS TO THIS REPORT
========================================================= Appendix III

GENERAL GOVERNMENT DIVISION

J.  Christopher Mihm, Associate Director, Federal Management and
 Workforce Issues
James H.  Burow, Assistant Director
Jacqueline E.  Matthews, Evaluator-in-Charge
Timothy A.  Bober, Senior Evaluator
Ruth G.  Kassinger, Writer Consultant
Victoria E.  Miller, Senior Evaluator


BIBLIOGRAPHY OF NON-GAO
PUBLICATIONS AND SOURCES
=========================================================== Appendix 1

Anderson, Margo J.  American Census:  A Social History.  New Haven,
CT:  1988. 

Bohme, Frederick G.  "200 Years of Census Factfinding", Historian, U. 
S.  Department of Commerce, Bureau of Census, 1989. 

Bohme, Frederick G.  and David M.  Pemberton.  "Privacy and
Confidentiality in the U.S.  Censuses--A History", U.  S.  Department
of Commerce, Bureau of Census, 1991. 

Bryant, Barbara Everitt (Bureau Director,1989-1993) and William Dunn. 
Moving Power and Money, The Politics of Census Taking.  Ithaca, NY: 
1995. 

Cassedy, James H.  Demography in Early America:  Beginnings of the
Statistical Mind, 1600-1800.  Cambridge, MA:  1969. 

Demos, John.  A Little Commonwealth:  Family Life in Plymouth Colony. 
New York, NY:  1970. 

Durbin, Thomas M.  and L.  Paige Whitaker.  "Congressional and State
Reapportionment and Redistricting:  A Legal Analysis", Congressional
Research Service, The Library of Congress, 1996. 

Eckler, A.  Ross (Bureau Director, 1965-1969).  The Bureau of the
Census.  New York, NY:  1972. 

Huckabee, David C.  "Apportioning Seats in the House of
Representatives:  The Method of Equal Proportions", Congressional
Research Service, Library of Congress, 1988 _____"House Apportionment
Following the 2000 Census:  Preliminary Projections", Congressional
Research Service, The Library of Congress, 1997. 

Melnick, Daniel.  "The 1980 Census:  Recalculating the Federal
Equation", Congressional Research Service, Library of Congress, 1981. 

National Research Council, National Academy Press.  Modernizing the
U.S.  Census, Panel on Census Requirements in the Year 2000 and
Beyond.  Washington, D.C.:  1995. 

U.  S.  Department of Commerce, Bureau of the Census.  The Story of
the Census--1790-1915, 1915 _____ 200 Years of U.S.  Census Taking: 
Population and Housing Questions, 1790 - 1990, 1989.  _____ 1990
Census of Population and Housing, Guide, Parts A, B, C, and D, 1992,
1993, 1994, 1995, and 1996.  (History Series:  CPH-R-1A, CPH-R-2A,
2B, 2C, and 2D)

Wright, Carroll D.  (Commissioner of Labor).  The History and Growth
of the United States Census.  Washington, D.C.:  1900. 

SELECTED GAO PRODUCTS RELATED TO THE DECENNIAL CENSUS

2000 Census:  Preparations for Dress Rehearsal Leave Many Unanswered
Questions (GAO/GGD-98-74, Mar.  26, 1998). 

2000 Census:  Progress Made on Design, but Risks Remain
(GAO/GGD-97-142, July 14, 1997). 

Addressing the Deficit:  Budgetary Implications of Selected GAO Work
for Fiscal Year 1998 (GAO/OCG-97-2, Mar.  14, 1997). 

High Risk Series (GAO/HR-97-1 and 97-2, Feb.  1997). 

Decennial Census:  Fundamental Design Decisions Merit Congressional
Attention (GAO/T-GGD-96-37, Oct.  25, 1995). 

Decennial Census:  1995 Test Census Presents Opportunities to
Evaluate New Census-taking Methods (GAO/T-GGD-94-136, Sept.  27,
1994). 

Decennial Census:  Promising Proposals, Some Progress, But Challenges
Remain (GAO/T-GGD-94-80, Jan.  26, 1994). 

Decennial Census:  Test Design Proposals Are Promising, But
Fundamental Reform Is Still at Risk (GAO/T-GGD-94-12, Oct.  7, 1993). 

Decennial Census:  Focused Action Needed Soon to Achieve Fundamental
Breakthroughs (GAO/T-GGD-93-32, May 27, 1993). 

Census Reform:  Early Outreach and Decisions Needed on Race and
Ethnic Questions (GAO/GGD-93-36, Jan.  28, 1993). 

Census Reform:  Questionnaire Test Shows Simplification Holds Promise
(GAO/T-GGD-92-59, July 1, 1992). 

Decennial Census:  1990 Results Show Need for Fundamental Reform
(GAO/GGD-92-94, June 9, 1992). 

1990 Census:  Limitations in Methods and Procedures to Include the
Homeless (GAO/GGD-92-1, Dec.  30, 1991). 

1990 Census:  Reported Net Undercount Obscured Magnitude of Error
(GAO/GGD-91-113, Aug.  22, 1991). 

1990 Census Adjustment:  Estimating Census Accuracy - A Complex Task
(GAO/GGD-91-42, Mar.  11, 1991). 

Components of the 1990 Census Count (GAO/T-GGD-91-8, Feb.  21, 1991). 

Decennial Census:  Preliminary 1990 Lessons Learned Indicate Need to
Rethink Census Approach GAO/T-GGD-90-18, Aug.  8, 1990). 

1990 Census:  Overview of Key Issues (GAO/GGD-89-77BR, July 3, 1989). 


*** End of document. ***