2000 Census: Contingency Planning Needed to Address Risks That Pose a
Threat to a Successful Census (Letter Report, 12/14/1999, GAO/GGD-00-6).

Pursuant to a congressional request, GAO provided information on the
Year 2000 census, focusing on: (1) the need to boost the declining level
of public participation in the census; and (2) the Census Bureau's need
to collect timely and accurate data from nonrespondents.

GAO noted: (1) with less than 4 months remaining until Census Day,
significant operational uncertainties continue to surround the Bureau's
efforts to increase participation in the census and to collect timely
and accurate field data from nonrespondents; (2) key to a successful
census is the level of public participation, as measured by the
questionnaire mail response rate; (3) however, the response rate has
been declining since 1970, in part because of various demographic and
attitudinal factors, such as more complex housing arrangements and
public mistrust of government; (4) based on the 1998 dress rehearsal for
the 2000 Census, the Bureau estimates a 61-percent mail response rate in
2000; (5) however, this goal may be optimistic because: (a) a key
ingredient of the dress rehearsal mail response rate - a second
"replacement" questionnaire - will not be used in 2000 because the
Bureau is concerned that the questionnaire could confuse recipients,
which could lead to duplicate responses, and (b) while the Bureau has
instituted an extensive outreach and promotion effort to help it achieve
its desired response rate, dress rehearsal results suggest the Bureau
still has not resolved the long-standing challenge of motivating public
participation in the census; (6) the Bureau's ability to complete its
field operations on time without compromising data quality is another
significant risk to a successful census; (7) past experience has shown
that following up on nonresponding households is one of the most
error-prone and costly of all census-taking activities, requiring the
Bureau to fill about 860,000 positions and recruit up to 3.5 million
people; (8) even if the Bureau achieves its 61-percent mail response
rate objective, it will have a nonresponse follow-up workload of 46
million housing units; (9) to complete this workload in the 10-week time
frame that the Bureau has allocated, it will need to close an average of
657,000 cases every day; (10) however, a lower-than-expected mail
response rate, difficulties in recruiting a sufficient number of workers
in a tight labor market, and a variety of other factors, could undermine
the Bureau's efforts and result in higher costs and less accurate data;
and (11) while the Bureau has established post-census coverage
improvement procedures to improve the accuracy of the 2000 Census data,
these procedures are similar to 1990 methods that had limited success.

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

 REPORTNUM:  GGD-00-6
     TITLE:  2000 Census: Contingency Planning Needed to Address Risks
	     That Pose a Threat to a Successful Census
      DATE:  12/14/1999
   SUBJECT:  Census
	     Vital records
	     Population statistics
	     Cost analysis
	     Mailing lists
	     Surveys
	     Statistical data
	     Government information dissemination
	     Communication
	     Labor supply
IDENTIFIER:  Census Bureau Complete Count Committee Program
	     2000 Decennial Census

******************************************************************
** 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.                 **
**                                                              **
** 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.                         **
******************************************************************

United States General Accounting Office
GAO

Report to the Subcommittee on the Census,

Committee on Government Reform, House of

Representatives

December 1999

GAO/GGD-00-6

2000 CENSUS
Contingency Planning Needed to Address Risks

That Pose a Threat to a Successful Census

Ordering Information

The first copy of each GAO report and testimony
is free. Additional copies are $2 each. Orders
should be sent to the following address,
accompanied by a check or money order made out
to the Superintendent of Documents, when
necessary. VISA and MasterCard credit cards are
accepted, also. Orders for 100 or more copies to
be mailed to a single address are discounted 25
percent.

Order by mail:

U.S. General Accounting Office
P.O. Box 37050
Washington, DC 20013

or visit:

Room 1100
700 4th St. NW (corner of 4th and G Sts. NW)
U.S. General Accounting Office
Washington, DC

Orders may also be placed by calling (202) 512-
6000 or by using fax number (202) 512-6061, or
TDD (202) 512-2537.

Each day, GAO issues a list of newly available
reports and testimony. To receive facsimile
copies of the daily list or any list from the
past 30 days, please call (202) 512-6000 using a
touch-tone phone. A recorded menu will provide
information on how to obtain these lists.

For information on how to access GAO reports on
the INTERNET, send e-mail message with "info" in
the body to:

[email protected]

or visit GAO's World Wide Web Home Page at:

http://www.gao.gov

United States
General Accounting Office
Washington, D.C. 20548-0001
Official Business
Penalty for Private Use $300
Address Correction Requested

                    Bulk Rate
               Postage & Fees Paid
                       GAO
                 Permit No. G100

(410447)

Contents
Page 341GAO/GGD-00-6 Contingency Planning Needed to Address Risks
Letter                                                                      1
                                                                             
Appendix I                                                                 36
Comments From the
Secretary of Commerce
                                                                             
Appendix II                                                                38
GAO Contacts and Staff
Acknowledgments
                                                                             
Tables                     Table 1:  Dress Rehearsal Mail                   7
                           Response Rates
                           Table 2:  Impact of Second                       8
                           Questionnaire on Dress Rehearsal Mail
                           Response Rates
                           Table 3: Agencies Have Taken Actions            24
                           to Expand the Census Applicant Pool
                                                                             
Figures                    Figure 1: Census Questionnaire Mail              6
                           Response Rates Have Been Steadily
                           Declining
                           Figure 2:  Percentage of Advertising            10
                           Budget Allocated to Targeted Ethnic
                           Groups Through the End of Fiscal Year
                           1999
                           Figure 3:  Paid Advertising Campaign            12
                                                                           13
                           Figure 4:  The Bureau's Nonresponse             20
                           Follow-up Workload Has Increased,
                           While Time Frames Have Been
                           Compressed
                                                                             

Abbreviations

HUD       Department of Housing and Urban
Development
OPM       Office of Personnel Management

B-282565

Page 32GAO/GGD-00-6 Contingency Planning Needed to
Address Risks
     B-282565

     December 14, 1999

The Honorable Dan Miller
Chairman
The Honorable Carolyn B. Maloney
Ranking Minority Member
Subcommittee on the Census
Committee on Government Reform
House of Representatives
 
Taking a complete and accurate census is an
enormous and costly endeavor. For the 2000 Census,
the Bureau of the Census estimates it will need to
properly locate and collect information on more
than 270 million residents living in about 119
million housing units. To do this, the Bureau
expects to open over 520 local census offices;
fill about 1.35 million temporary positions;
receive, at peak, from 40 to 70 million
questionnaires; process about 1.5 billion pages of
data; and use more than 20 million maps for
fieldwork. The Bureau estimates that the cost of
the 2000 Census will be at least $6.8 billion. 1
This represents an increase of 113 percent in real
terms over the $3.2 billion cost of the 1990
Census in 1999 dollars. When measured on a cost-
per-housing unit basis, the 2000 Census will cost
an estimated $57 per housing unit in 2000 compared
to about $31 in 1990 (in 1999 dollars), which is
an increase of 84 percent.

The 1990 Census was the most costly census in U.S.
history and data were less accurate than the 1980
Census, leaving millions of Americans-especially
members of minority groups-uncounted. Throughout
this decade, the Bureau has aggressively planned
and implemented a census in 2000 that seeks to
address the problems with prior censuses.  Despite
the Bureau's efforts, however, our work in recent
years has continued to show that formidable
challenges surround key census-taking operations.
In February 1997, our findings led us to conclude
that there is a high risk that the 2000 Census
will be less accurate and more costly than
previous censuses.2

Two of our long-standing concerns in this regard
have been the need to boost the declining level of
public participation in the census-a significant
problem that is certainly beyond the Bureau's
ability to control-and the Bureau's need to
collect timely and accurate data from
nonrespondents. As agreed with your offices, we
reviewed the Bureau's progress in reducing the
risks that these challenges pose to a successful
census.

Results in Brief
With less than 4 months remaining until Census Day
(Apr. 1, 2000), significant operational
uncertainties continue to surround the Bureau's
efforts to increase participation in the census
and to collect timely and accurate field data from
nonrespondents. These uncertainties raise concerns
that the 2000 Census may be less accurate than the
1990 Census.

   Operational Uncertainties Surround the 2000
Census

Motivating Public Participation Will Be a
Formidable Task

Achieving the Bureau's mail response rate
objective will be difficult

Outreach and promotion program may have only a
modest impact on the mail response rate

Field Follow-up Efforts Will Be Costly and May
Produce Unreliable Data

Operational challenges could undermine nonresponse
follow-up efforts

The Bureau may be challenged to meet field
staffing goals

Post-census day coverage improvement initiatives
offer little hope of resolving the undercount

Key to a successful census is the level of public
participation, as measured by the questionnaire
mail response rate. A high mail response rate
helps the Bureau obtain more accurate census data
and reduce its costly nonresponse follow-up
workload. However, the response rate has been
declining since 1970, in part because of various
demographic and attitudinal factors, such as more
complex housing arrangements and public mistrust
of government.

The Bureau estimates that, partly on the basis of
its experience in the 1998 dress rehearsal for the
2000 Census, it will achieve a 61-percent mail
response rate in 2000. However, this goal may be
optimistic for at least two reasons. First, a key
ingredient of the dress rehearsal mail response
rate-a second "replacement" questionnaire-will not
be used in 2000 because the Bureau is concerned
that the questionnaire could confuse recipients,
which could lead to duplicate responses. Second,
while the Bureau has instituted an extensive
outreach and promotion effort to help it achieve
its desired response rate, dress rehearsal results
suggest that the Bureau still has not resolved the
long-standing challenge of motivating public
participation in the census.

The Bureau's ability to complete its field
operations on time without compromising data
quality is another significant risk to a
successful census. Past experience has shown that
following up on nonresponding households is one of
the most error-prone and costly of all census-
taking activities. For 2000, the Bureau estimates
that it will need to fill about 860,000 positions
for nonresponse follow-up and other peak field
operations. To fill this many positions, the
Bureau plans to recruit 3.5 million people, which
is a number roughly equivalent to the population
of South Carolina.

Moreover, even if the Bureau achieves its 61-
percent mail response rate objective, it will have
a nonresponse follow-up workload of 46 million
housing units. To complete this workload in the 10-
week time frame that the Bureau has allocated, it
will need to close an average of 657,000 cases
every day. However, a lower-than-expected mail
response rate, difficulties in recruiting a
sufficient number of workers in a tight labor
market-despite the Bureau's recruiting efforts and
offers of competitive, geographically based pay
wages-and a variety of other factors, could
undermine the Bureau's efforts and result in
higher costs and less accurate data. Finally,
while the Bureau has established post-census
coverage improvement procedures to improve the
accuracy of the 2000 Census data, these procedures
are similar to 1990 methods that had limited
success.

     Given the operational uncertainties
surrounding public participation in the census and
the Bureau's field follow-up operations, we
recommend that the Bureau develop contingency
plans to mitigate the impact of a lower-than-
expected response rate. This recommendation is
consistent with suggestions we made in our 1992
summary assessment of the 1990 Census.3 Congress
may also wish to consider opportunities to expand
the potential census applicant pool. The Bureau
has worked with other agencies to identify and
implement several important provisions to expand
the census labor pool.  Additional opportunities
include, for example, legislative actions to
remove financial disincentives that could
discourage a wide range of people from seeking
census employment, including recipients of Social
Security, veterans healthcare, food stamp,
Medicaid, and Temporary Assistance for Needy
Families benefits, federal and military retirees,
and others.

Background
A successful census is critical because, as
required by the Constitution, census data are used
to reapportion seats in the House of
Representatives. In addition, every year, the
government awards around $180 billion in federal
funds to localities on the basis of census
numbers, and states use census data, among other
purposes, to redraw the boundaries of
congressional districts. Businesses and private
citizens also depend on census data for such
purposes as marketing and planning. Census Day is
April 1, 2000, with peak efforts to follow up on
nonresponding households scheduled to run from
April 27 to July 7, 2000. Population counts to be
used to reapportion seats in the House of
Representatives are to be delivered to the
President by January 1, 2001.

In 1998, the Bureau conducted a dress rehearsal
for the 2000 Census during which it tested most of
the procedures and operations planned for the
decennial census under as near census-like
conditions as possible. The dress rehearsal was
the Bureau's last opportunity for an operational
test of its overall design of the 2000 Census and
to demonstrate to Congress and other key
stakeholders the feasibility of its plans. (Dress
Rehearsal Census Day was April 18, 1998). The
dress rehearsal sites included Sacramento, CA; 11
county governments and the City of Columbia, SC;
and Menominee County, WI, including the Menominee
American Indian Reservation.

Scope and Methodology
To review the Bureau's efforts to increase public
participation in the census and to collect timely
and accurate field data from nonrespondents, we
examined documents that described the Bureau's
budget, plans, procedures, progress, and
evaluations relating to these operations. Further,
we examined current laws, regulations, and
legislation pertaining to staffing the Bureau's
field operations. We also interviewed Bureau
officials at headquarters and, where applicable,
regional and local census officials responsible
for planning and implementing the 1998 dress
rehearsal and the 2000 Census.

To obtain a local perspective on the Bureau's
outreach and promotion and field follow-up efforts
during the dress rehearsal, we made several site
visits to the dress rehearsal jurisdictions and
interviewed local officials who were responsible
for organizing and implementing community outreach
and promotion efforts. We also inspected each site
for the scope and prominence of promotional
material and activities, and observed nonresponse
follow-up operations. To determine the dollar
effect of a 1 percentage point decrease in the
mail response rate, we reduced the assumed mail
response rates used in the Bureau's cost model
supporting its fiscal year 2000 amended budget
request.

We did our audit work at the Bureau's Census 2000
dress rehearsal sites; Regional Census Offices in
Charlotte, NC, and Seattle, WA; Bureau
headquarters in Suitland, MD; as well as in
Washington, D.C., between April 1998 and October
1999, in accordance with generally accepted
government auditing standards.

We requested comments on a draft of this report
from the Secretary of Commerce. On December 3,
1999, the Secretary forwarded the Bureau's written
comments on the draft (see app. I), which we
address at the end of this report.

Motivating Public Participation Will Be a
Formidable Task

Achieving the Bureau's Mail Response Rate
Objective Will Be Difficult

Outreach and Promotion Program May Have Only a
Modest Impact on the Mail Response Rate

Public participation is critical to a successful
census because it helps improve the accuracy and
completeness of census information, while reducing
the Bureau's costly and time-consuming nonresponse
follow-up workload. The mail response rate to the
census questionnaire is the most commonly used
indicator of the level of public participation.4
Unfortunately, as shown in figure 1, the mail
response rate has declined with each decennial
census since the Bureau first initiated a national
mailout/mailback approach in 1970.

Figure 1: Census Questionnaire Mail Response Rates
Have Been Steadily Declining

Source: U.S. Census Bureau.

According to the Bureau, this declining trend is
due, in part, to various demographic, attitudinal,
and other factors, such as concerns over privacy
and mistrust of government. These and related
issues led us to note, in our 1992 summary
assessment of the 1990 Census, the formidable
challenges the Bureau faces in increasing public
cooperation. We identified several opportunities
for improvements in this regard. One of these
opportunities was to simplify the census
questionnaire-which the Bureau has successfully
done. However, even with a simplified
questionnaire and other changes in census design,
we noted the Bureau needed to prepare for a lower
mail response rate in 2000.5

For the 2000 Census, the Bureau is expecting a 61-
percent mail response rate, which is 4 percentage
points lower than what it achieved in 1990.
However, the dress rehearsal results suggest that
even this goal may be optimistic. First, while the
Bureau generally achieved its mail response rate
goals during the dress rehearsal, it did so only
by mailing out a second, "replacement"
questionnaire, which is an approach the Bureau has
since dropped for 2000. Second, the Bureau's
outreach and promotion program does not appear to
have bridged the gap that typically exists between
raising awareness of the census on the one hand
and motivating people to respond on the other. The
significant difficulty in both raising public
awareness and motivating people to mail back their
questionnaires was demonstrated during the 1990
Census, when Bureau research showed that although
93 percent of the public reported being aware of
the census, the mail response rate was just 65
percent.

Achieving the Bureau's Mail Response Rate
Objective Will Be Difficult
The dress rehearsal results raise concerns as to
whether the Bureau can achieve its 61-percent mail
response rate goal in 2000. As shown in table 1,
the Bureau generally met its dress rehearsal mail
response rate objectives, exceeding its goal by
three percentage points in Sacramento and falling
slightly short in South Carolina and Menominee.

Table 1:  Dress Rehearsal Mail Response Rates
Site            Data          Response     Actual
               collection   rate goal   response
               strategy                     rate
South Carolina  Mailout/Mai       55 %         55 %
               lback
                Update/Leav       55          48 
               e
                Overall           55        53.4 
Sacramento, CA  Mailout/Mai       50          53 
               lback
Menominee, WI   Update/Leav       40          39 
               e
Note: In mailout/mailback areas, housing units are
to receive all questionnaires and related material
by mail, and are to return completed
questionnaires by mail. In update/leave areas,
enumerators are to deliver questionnaires to
housing units while updating the Bureau's master
address list. Residents at these housing units are
to then return completed questionnaires by mail.
Source: U.S. Census Bureau.

However, a key ingredient of these response rates
was the Bureau's use of a second, "replacement"
questionnaire that was sent to all housing units
located in mailout/mailback areas in South
Carolina and Sacramento. The Bureau has since
rejected this procedure because it concluded that
the number of duplicate responses suggested that
the second mailing confused the public. Bureau
officials told us that a similar situation in 2000
could cause overwhelming processing problems.

Recognizing the potential impact of not using a
second mailing, the Bureau reduced its initial
goal of a 67-percent mail response rate by 6
percentage points to 61 percent. However, the
results of a subsequent Bureau study suggest that
the second mailing during the dress rehearsal had
an even greater impact on the mail response rate,
and, as a result, the Bureau's current 61-percent
response rate objective could be optimistic.

Although the impact of the second questionnaire is
difficult to measure precisely, as shown in table
2, the Bureau estimates that the second
questionnaire added between 8.2 and 15.8
percentage points to the South Carolina response
rate, and between 7.5 and 14.4 percentage points
to the Sacramento response rate.6 Thus, without
the second mailing, it is likely that the Bureau
would have fallen far short of its response rate
goals-by at least 8.2 percentage points in South
Carolina and 4.5 percentage points in Sacramento.

Table 2:  Impact of Second Questionnaire on Dress
Rehearsal Mail Response Rates
Sitea       Response  Actual Increase   Response
           rate goal             from       rate
                     respons   second    without
                           e  mailing     second
                        rate             mailing
South            55 %    55 %    8.2 - %      46.8 %
Carolina                        15.8
Sacramen         50     53     7.5 -        45.5 
to, CA                          14.4
aThe Menominee, WI, dress rehearsal site is not
shown in this table because it did not test the
mailout/mailback data collection strategy and,
thus, was not used to test a second questionnaire
mailing.
Source: GAO analysis of U.S. Census Bureau data.

Dress rehearsal mail response rates are not
necessarily predictive of decennial response
rates, which are higher because of the greater
public and media attention that the actual census
receives. Nevertheless, the dress rehearsal mail
response rates provide a useful indication of what
might occur during the actual census and, for
2000, raise concerns that the Bureau is at risk of
an even lower response rate than it had estimated.

Outreach and Promotion Program May Have Only a
Modest Impact on the Mail Response Rate
To help combat the downward trend in response
rates, the Bureau has instituted both a national
and locally based outreach and promotion program.
Two key components of the Bureau's outreach and
promotion program include a paid advertising
campaign and partnerships with local governments.

Paid Advertising May Have a Limited Effect on
Census Participation
In October 1997, the Bureau hired a consortium of
private-sector advertising agencies, led by Young
& Rubicam, to develop an extensive paid
advertising program for the 2000 Census. Marketing
the census represents a particular challenge in
that advertisers typically target their best
prospects and specific segments of the population.
In contrast, census advertising is aimed at the
most resistant "customers" and every U.S.
household.

The Bureau estimates it will spend about $167
million on the paid advertising campaign in fiscal
years 1998 through 2000, of which $102.8 million
(62 percent) has been allocated in fiscal years
1999 and 2000 for media (television, radio, print,
and other types of advertising). A substantial
portion of the advertising is to be directed at
minority groups. Through the end of fiscal year
1999, for example, of the $16.4 million allocated
for media purchases, about $7.3 million (45
percent) was to be used to target specific race
and ethnic groups (see fig. 2). The Bureau has not
yet purchased advertising for 2000, although a
similar spending pattern is likely.

Figure 2:  Percentage of Advertising Budget
Allocated to Targeted Ethnic Groups Through the
End of Fiscal Year 1999

Source: U.S. Census Bureau.

According to Bureau officials, the paid
advertising campaign is intended to motivate
people to return their census forms by using a
variety of media to stress the message that
participating in the census benefits one's
community.  We observed this during our visits to
the dress rehearsal sites where we often saw
billboards containing such taglines as "This is
Your Future.  Don't Leave It Blank," "The Future
Takes Just a Few Minutes to Complete," and "Pave a
Road With These Tools" (see fig. 3).  The census
was also publicized through broadcast and print
media and promotional items, such as cups and T-
shirts.

     [Blank page]

Figure 3:  Paid Advertising Campaign

Nevertheless, the effectiveness of the paid
advertising campaign appears to have been limited
during the dress rehearsal. An independent
research firm, which the Bureau hired to evaluate
the effectiveness of the advertising campaign,
reported that the campaign generally had no more
than a "modest" impact on the public's attitudes
and knowledge of the census. Although the Bureau
had expected a 30 percentage point increase in
awareness at the South Carolina and Sacramento
dress rehearsal sites, the evaluation results
indicated that there already was a high level of
census awareness among all demographic groups
before the start of the advertising campaign. In a
telephone survey of residents that was conducted
before the advertising campaign, 86 percent of
those responding in Sacramento and 93 percent of
South Carolina respondents said that they had
heard of the census. (The Bureau believes that
events occurring before the start of the
advertising campaign, such as news coverage about
the census approach, among other factors, may have
contributed to the high level of awareness
observed before the start of the advertising
campaign). Following the advertising campaign,
awareness levels increased by 8 percentage points
in Sacramento and 5 percentage points in South
Carolina (minority groups, the less educated, and
the less affluent experienced a greater increase
in awareness).

Significantly, much like the 1990 Census, the
public's high level of awareness was not matched
by similarly high mail response rates. As
previously discussed, at the South Carolina and
Sacramento dress rehearsal sites, the mail
response rates were about 55 percent.

Following the dress rehearsal, the Bureau expanded
and enhanced the paid advertising campaign, in
part by adding messages that are to run prior to,
and following Census Day. The campaign, which is
to run from November 1999 through late-May 2000,
is divided into three phases: educational,
motivational, and nonresponse follow-up. The
phases will be similar in that they will all
contain messages about census benefits and
confidentiality. In addition, the motivational
phase, timed to coincide with the census
questionnaire mailings, is to let people know to
expect the census form and to mail it back. The
nonresponse follow-up phase, which is to occur
when the Bureau is going door-to-door collecting
data from nonrespondents, is to encourage people
to cooperate with census enumerators.

Still, the impact that this additional advertising
might have on people's willingness to respond to
the census is difficult to gauge. According to the
Bureau, there did not appear to be a direct
relationship between advertising exposure during
the dress rehearsal and the likelihood of
returning a census form.7 However, the Bureau
suspects that the campaign had an "indirect
effect" on public response to the census in that
the campaign may have made people expect the
census form in the mail, which, in turn, increased
the likelihood that they would return it.
Moreover, as noted earlier, even though the
advertising campaign for 2000 has been greatly
enhanced since the dress rehearsal, high levels of
awareness do not guarantee high mail response
rates.

Mismatch Between Bureau Expectations and Local
Governments' Capabilities Could Undermine
Partnership Efforts
In addition to the paid advertising campaign, the
Bureau is seeking to form partnerships with local
governments, community groups, businesses and
nongovernmental organizations to promote the
census on a grassroots basis. The Bureau has
allocated $108 million for its partnership
initiatives in fiscal years 1999 and 2000.

A key element of the Bureau's local partnership
effort will be Complete Count Committees, which
are to consist of local government, religious,
media, education, and other community leaders. The
committees are to promote the census by sponsoring
promotional events, placing articles in local
newspapers, and holding press conferences that
convey the importance of the census, among other
activities. For 2000, as a matter of long-standing
policy, the Bureau is not directly funding local
outreach and promotion activities. Instead, for
fiscal years 1999 and 2000, the Bureau is to
distribute about $1.2 million to each of the
Bureau's 12 Regional Census Centers for in-kind
services, such as printing handouts. The Bureau
also plans on assigning employees, known as
partnership specialists, to work with local groups
to help them initiate and sustain grassroots
marketing activities, such as the Complete Count
Committees.

The Complete Count Committee program stems from
the Bureau's recognition that the paid advertising
campaign alone will not get the message across to
everybody-particularly the hard-to-count-that
participating in the census is important. The
Bureau hopes that local people who are trusted by
members of the community can more effectively
market the census to those who are difficult to
convince through traditional advertising media.
Thus, while the Bureau plans on partnering with a
number of religious, service, community, and other
organizations-often to increase census
participation among certain groups or areas-the
Bureau believes that Complete Count Committees are
the key to making each and every community aware
of the census and persuading everyone to respond.

However, during the dress rehearsal, we found that
the effectiveness of the Complete Count Committee
program was undermined by an apparent mismatch
between the Bureau's expectations of the
committees and what the committees could
realistically accomplish with their limited
resources. While the Bureau expected local
governments to plan and execute an outreach and
promotion program largely on their own with
minimal direct support from the Bureau, we found
that many local governments lacked the money,
people, and/or expertise to launch an adequate
marketing effort during the dress rehearsal. If
such expectations remain misaligned for 2000,
these disappointing results could continue.

Regarding money, officials representing 9 of the
14 local governments participating in the dress
rehearsal told us that they were unable or
unwilling to fund promotional activities. For
example, while the Sacramento committee initially
developed a list of several dozen promotional
activities involving local media and other
organizations, a committee representative told us
that many activities were dropped because of a
lack of money. Although the Bureau encouraged
committees to turn to local businesses for
support, the committees (1) were generally too
small to organize an effective outreach effort or
(2) viewed such an effort as a federal function.
As one South Carolina committee representative
said, "Fundraising for the federal government
doesn't go over well.. That's what taxes are for."

The Bureau may also have overly optimistic
expectations of the level of staff and expertise
available at the local level to plan and implement
outreach and promotion activities. This was
evident during the dress rehearsal where some
local governments had difficulty getting staff to
volunteer to help plan and organize promotion
activities. At the South Carolina and Menominee
sites, for example, some local officials expressed
frustration and others resentment for what they
perceived as the burden of promoting the census
and the time it was taking from their other
responsibilities. In addition, local governments
may lack the know-how to launch an effective
marketing effort. During the dress rehearsal, for
example, the Bureau's South Carolina partnership
specialist said that the Bureau assumed that the
South Carolina counties had the experience and
knowledge to market the census. However, she noted
that, in hindsight, the opposite was often the
case in those counties.

In addition, while the Bureau's partnership
specialists are to provide needed expertise and
assistance to local governments and other groups,
the dress rehearsal suggested that these
specialists may be spread too thin to offer
meaningful support in 2000. In our past work,8 we
reported that some South Carolina committees never
formed, while others became inactive, partly
because the Bureau's two partnership specialists
were responsible for assisting 11 county
governments and the City of Columbia-a geographic
area covering more than 6,700 square miles.

In 2000, the partnership specialists will likely
have a far greater workload. The Bureau plans to
fill 542 partnership specialist positions to
assist local governments. According to the Bureau,
as of the end of July 1999, about 6,800 local
governments had formed Complete Count Committees,
(including 50 of the 51 largest cities). The
Bureau expects that as many as 8,000 committees
will ultimately be formed. Thus, on average, each
partnership specialist could be responsible for
assisting between 13 and 15 local governments. By
comparison, the problems we observed at the South
Carolina dress rehearsal site occurred when each
partnership specialist was responsible for
assisting an average of six local entities.

Further, as local governments have been forming
Complete Count Committees for the 2000 Census,
early indicators suggest that the potential impact
of this program may not be fully realized. For
example, the effectiveness of the Complete Count
Committee program will be partly determined by the
number of governments that decide to participate.
In the spring of 1998, the Bureau formally invited
all 39,000 local and tribal governments in the
United States to establish such committees.
Although the Bureau did not expect that all 39,000
local governments would do so, the 6,800
committees formed so far represent about 40.4
million people-or only 16 percent of the U.S.
population. Moreover, those local governments that
do not form Complete Count Committees could add to
the partnership specialists' workload, because the
partnership specialists will need to develop some
other method of publicizing the census in those
locations.

Field Follow-up Efforts Will be Costly and May
Produce Unreliable Data

Operational Challenges Could Undermine Nonresponse
Follow-up Efforts

The Bureau May Be Challenged to Meet Field
Staffing Goals

Post-Census Day Coverage Improvement Initiatives
Offer Little Hope of Resolving the Undercount

The Bureau implements a nationwide field follow-up
operation in an attempt to count those individuals
who did not mail back their census questionnaires.
Specific activities include (1) nonresponse follow-
up, during which temporary Bureau employees, known
as enumerators, visit and collect census
information from each nonresponding housing unit
and (2) additional coverage improvement
initiatives, which are aimed at collecting data
from people missed during the initial enumeration
and nonresponse follow-up.

According to the Bureau, even if it achieves its
anticipated 61-percent mail response rate,
enumerators will need to follow up with 46 million
nonresponding housing units. However, completing
this workload in the 10-week time period the
Bureau has allotted for nonresponse follow-up,
without compromising data quality, could prove
extremely difficult.

During the 1990 Census, for example, field follow-
up operations proved to be error-prone and costly,
in part because a higher than expected nonresponse
follow-up workload required the Bureau to hire
more enumerators than originally anticipated.
However, some local census offices could not meet
the demand for additional enumerators, which
delayed the completion of nonresponse follow-up.
As the time spent on data collection dragged on,
the rate of errors appeared to increase because
people moved or could not recall who had been
residing at their home on Census Day. Furthermore,
to complete nonresponse follow-up, enumerators
collected data from secondhand sources, such as
neighbors and mail carriers-referred to as "proxy"
data. However, the Bureau-on the basis of its work
evaluating past census operations-has found that
proxy data are not as reliable as data obtained
directly from household residents.

In addition, field follow-up operations are
expensive. The Bureau estimates that, in 2000, the
cost to enumerate a household that mails back the
census questionnaire will be about $3. For those
households that do not return a
questionnaire-requiring enumerators to obtain the
information-costs could be as high as $35 per
questionnaire.

Operational Challenges Could Undermine Nonresponse
Follow-up Efforts
The combined challenges that affected the success
of the Bureau's 1990 nonresponse follow-up
operations-completing nonresponse follow-up on
time, maintaining data quality, and recruiting a
sufficient number of enumerators-may pose similar,
if not greater, challenges for the Bureau in 2000.

Completing Nonresponse Follow-up on Schedule
Without Compromising Data Quality May Be Difficult
For the 2000 Census, the Bureau has based its $1.5
billion nonresponse follow-up budget on the
assumption that it will achieve a 61-percent mail
response rate, which corresponds to a follow-up
workload of about 46 million of the 119 million
housing units estimated to comprise the nation.
However, given the Bureau's experiences during
past censuses, it will be challenged to complete
the nonresponse follow-up workload on time and
minimize the collection of proxy data.

For example, during the 1990 Census, because of
unanticipated workload, staffing, and scheduling
problems, it took the Bureau 14 weeks to complete
nonresponse follow-up on 34 million housing
units-8 more weeks than the 6-week period that the
Bureau initially estimated for that operation. For
2000, the Bureau has scheduled 10 weeks to follow
up on an expected 46 million housing units. Under
this timetable, the Bureau has 4 weeks less time
to follow up on 12 million more households, when
compared to 1990 (see fig. 4).

Figure 4:  The Bureau's Nonresponse Follow-up
Workload Has Increased, While Time Frames Have
Been Compressed

Source: GAO analysis of Census Bureau data.

Thus, to follow up on 46 million households within
the 10-week time frame, the Bureau will need to
complete over 657,000 cases each day for the
entire 10-week period. In addition, the Bureau's
quality assurance procedures, which call for
enumerators to revisit certain households to
identify and correct enumeration errors, will add
more than 17,000 households to the Bureau's
average daily workload. Maintaining this pace
could prove difficult for a variety of factors
that range from the availability of a productive,
temporary workforce, to local weather conditions.

According to senior Bureau officials, a mail
response rate as little as 2 or 3 percentage
points less than the Bureau's 61-percent goal
could cause serious problems. For example,
according to Bureau officials, the Bureau has a
limited number of needed materials for nonresponse
follow-up. Furthermore, while the amount added to
total field data collection costs as a result of
any increased workload will ultimately depend on
where this workload is located and how the Bureau
manages its resources in completing this workload,
additional costs could, nonetheless, be
substantial.

Each percentage point drop in the mail response
rate would increase the nonresponse follow-up
workload by about 1.2 million households. In 1995,
the Bureau estimated that a 1 percentage point
increase in workload could add approximately $25
million to the cost of the census. However, on the
basis our analysis of fiscal year 2000 Bureau
budget estimates, we project that a 1 percentage
point increase in workload could add at least $34
million in direct salary, benefits, and travel
costs to the $1.5 billion budgeted for nonresponse
follow-up. This $34 million in direct costs
exclude, for example, indirect costs for
headquarters and field support personnel, quality
control operations, rent, and data processing,
which may or may not be incurred. The Bureau's
ability to absorb these additional costs in its
fiscal year 2000 budget will be a function of the
actual outcome of other assumptions, such as
enumerator productivity, and the Bureau's ability
to manage other uncertainties. Of course, a higher
than expected mail response rate is possible and
could result in significant savings, which the
Bureau said it would use to augment its coverage
improvement programs for hard-to-count
populations.

Completing the nonresponse follow-up workload in a
timely manner will be critical to the Bureau's
collection of quality field data in 2000.
According to Bureau officials, the Bureau does not
plan to extend the nonresponse follow-up schedule
as it did during the 1990 Census. They noted that
the Bureau must meet the 10-week nonresponse
follow-up schedule to have time to complete other
census operations, including the coverage
evaluations that will be used to estimate census
under- and over-counts, and the processing and
preparation of census data for publication.
However, during the 1990 Census, enumerators
collected proxy data before they had collected
data from 95 percent of the cases within each
predefined housing district, as required by Bureau
procedures. Indeed, about 36 percent of the
Bureau's district offices used proxy data when the
caseload completion level was 90 percent or less.
Just 16 percent of the district offices began
collecting proxy data when the caseload was 95
percent or more complete.

During the dress rehearsal, while nonresponse
follow-up operations were completed on schedule in
both Menominee and Sacramento, and 6 days ahead of
schedule in South Carolina, the Bureau found that
obtaining interviews with household members proved
to be more difficult than it had anticipated. As a
result, the Bureau relied more heavily on proxy
data than it had planned. Although the Bureau
hoped to limit the portion of the nonresponse
follow-up universe that was proxy data to less
than 6 percent, the Bureau did not achieve this
objective at any of the three dress rehearsal
sites. In Sacramento, 20.1 percent of the occupied
nonresponse follow-up universe was proxy data; in
South Carolina, the proportion was 16.4 percent,
and in Menominee, it was 11.5 percent.

Because of the comparatively high use of proxy
data, a Bureau evaluation of the dress rehearsal
nonresponse follow-up operation noted that data
quality, especially for the long-form
questionnaire-which was somewhat more likely than
the short form to be enumerated via proxy-was a
concern.9 According to the evaluation, the data
obtained from the long-form questionnaire are
"especially suspect when obtained from a non-
household member."

The Bureau May Be Challenged to Meet Field
Staffing Goals
A number of questions also surround the Bureau's
ability to staff its nonresponse follow-up
operations, which also has implications for timely
and accurate field data collection. For example,
while the Bureau was generally successful in
staffing its dress rehearsal and initial census
operations and kept turnover rates to manageable
levels, the larger number of people the Bureau
needs to hire in 2000, combined with a tight labor
market and other factors, could pose problems.

The Bureau plans to fill about 860,000 positions
for peak field operations, including 539,000 for
nonresponse follow-up. To do this-on the basis of
the anticipated workload and the fact that the
vast majority of people offered a position may not
accept a census job or may resign before work
assignments are completed-the Bureau estimates it
will need to recruit nearly 3.5 million applicants
(a number roughly equivalent to the population of
South Carolina). However, achieving this staffing
goal will not be easy because the labor market has
become increasingly tight. According to the
Bureau, it took this factor into consideration in
setting an assumed enumerator productivity rate of
1.03 households per hour, which is based on
conservative senior management judgments. This
assumed productivity rate represents a 20-percent
reduction from an original assumption of 1.28
households per hour.  Higher than expected
productivity rates could reduce the Bureau's
staffing needs.  Nevertheless, as we have reported
in the past, staffing the census could still be
difficult because census jobs tend to be temporary
and do not offer benefits, such as health or life
insurance, sick or annual leave, retirement plans,
and childcare, and thus, may not be as attractive
to applicants as other employment opportunities.10

While the Bureau's recruiting initiatives appeared
to be effective during the dress rehearsal and
early operations for the 2000 Census, the Bureau
still encountered pockets of problems in specific
geographic locations. For example, when the Bureau
was conducting field operations to build the
address list for the 2000 Census in the resort
areas of Michigan's Upper Peninsula and Vail, CO,
the Bureau was competing for workers during the
seasonal vacation period. To help attract workers,
the Bureau quickly responded by raising hourly
wages. According to a Bureau official, the Bureau
anticipates similar pockets of recruiting problems
to occur during nonresponse follow-up operations
in 2000. Thus, it will be important for the Bureau
to monitor the progress of nonresponse follow-up
and respond quickly so that it can attract needed
staff.

To expand the applicant pool, the Bureau plans,
among other things, to (1) focus its recruiting
efforts on employed individuals seeking additional
jobs, retirees, and homemakers, among others; (2)
develop partnerships with state, local, and tribal
governments, community groups, and other
organizations to assist in recruiting efforts; (3)
expand its employment advertising; and (4) use a
geographic pay scale to set wages at 65 to 75
percent of local prevailing wages (from about
$8.25 to $21.50 per hour) to help make census jobs
more competitive.

The Bureau has also worked with other federal
agencies to waive regulations and policies that
restricted or financially discouraged certain
groups of people from seeking census employment.
As shown in table 3, the Department of Commerce
has authorized the Bureau to bypass Commerce's
policy preference against hiring noncitizens.11
Also, the Department of Housing and Urban
Development (HUD) waived regulations that could
have reduced the benefits for heads of households
receiving housing assistance because the
regulations would have required census income to
be included in the calculations used to determine
program eligibility. To make census employment
more attractive to former and current federal
employees, the Office of Personnel Management
(OPM) used its authority to allow federal and
military retirees, as well as current federal
workers, to work on the census without reducing
their benefits or income. (The requirement that
military retirees are to receive reduced annuities
upon federal reemployment was repealed by P.L. 106-
65, effective Oct. 1, 1999.) Regarding current
federal employees, over 80 federal agencies
employing over 2.4 million workers have authorized
that their employees may hold second appointments
with the Bureau. GAO, seeking to do its part to
help ensure a successful census, is participating
in this initiative.

Although the precise impact these actions might
have on census employment cannot be determined,
the agencies' actions could expand the potential
census applicant pool by millions of people.

Table 3: Agencies Have Taken Actions to Expand the
Census Applicant Pool
                                           Potential             Approximate
  Agency action                          applicanta                number of
                                                            people affected
                                                                 (millions)
In July 1999, Commerce authorized the    Noncitizens who are              6
Bureau to hire                           legal permanent
noncitizens, subject to annual           residents
appropriation act restrictionsb
and immigration law. Previously, under
Commerce policy, the
Bureau could hire noncitizens only if
there was a shortage of
qualified U.S. citizens in a particular
area.
In February 1998, OPM revised personnel  Civilian federal                 2
regulations to allow                     employees
federal employees working on the census
to retain the retirement
and insurance benefits to which they are
entitled under their
primary federal job, while earning
additional wages in their
second, census job.c
In May 1996, HUD agreed to exclude       Heads of households              4
census income from                       receiving housing
calculations used to determine           assistance
eligibility for its public and
assisted housing programs, including
Indian housing assistance.
In April 1996, OPM delegated to the      Federal civilian                 2
Bureau the authority to waive            annuitants
the requirement that federal civilian
retirees reemployed by the
federal government are to receive
reduced salaries, while
reemployed military personnel are to
receive reduced annuities.d Retirees can
only obtain this exemption subject to
the approval
of key Bureau headquarters officials or
Bureau regional directors.
aPotential applicant categories are not
necessarily mutually exclusive.
bAccording to OPM, since 1939, Congress has
prohibited most federal agencies from using their
appropriated funds to employ noncitizens within
the United States. However, certain groups are
currently excluded from this ban. They include (1)
persons from countries allied with the United
States in a defense effort; (2) citizens of
Ireland, Israel, and the Republic of the
Philippines; and (3) persons who owe permanent
allegiance to the United States, such as natives
of American Samoa.
cBefore this action, the Bureau would have had to
work closely with each employee's agency to
determine the amount of additional retirement
deductions, as well as life and health insurance
premiums, to withhold from the employee's
paycheck, a potentially error-prone and
administratively burdensome process, according to
OPM.
dSee 5 U.S.C. 5532, 8344, 8468. The requirement
that military retirees are to receive reduced
annuities upon federal reemployment was repealed
by Public Law 106-65, effective October 1, 1999.

During the 1990 Census, exemptions were made for
recipients of public housing assistance and
federal civilian and military retirees, and both
actions helped expand the census applicant pool.
According to the Bureau, about 20,000 federal and
military retirees worked on the 1990 Census, which
was 3.6 percent of the more than 550,000 people
hired overall. Similarly, when the public housing
assistance exemption was used in 1990, the Bureau
found that it helped generate applicants in
difficult-to-recruit areas, such as high crime,
inner city areas, and Indian reservations.

Pending Legislation Could Expand the Census
Applicant Pool
     Congress is considering three pieces of
legislation designed to improve the recruitment of
temporary census workers. H.R. 683, S. 752, and S.
1588 are similar in that each would (1) exempt
census income from calculations used to determine
eligibility for, or the amounts payable under, any
federal, state, or local program financed in whole
or in part with federal funds and (2) provide a
blanket exemption from income/annuity offset
provisions for federal civilian annuitants and
military retirees. As previously noted, this
annuity offset requirement was repealed for
military retirees by Public Law 106-65.

     The legislation, if enacted, would remove
financial disincentives that could discourage a
wide range of people from seeking census
employment. They include recipients of Social
Security, veterans healthcare, food stamp,
Medicaid, and Temporary Assistance for Needy
Families benefits, as well as federal and military
retirees.

     The broadest of the measures-S. 1588-also
includes financial incentives for volunteers who
help with the census, namely: reimbursements for
expenses, such as gasoline and food, and a program
of undergraduate or graduate debt relief.

Opportunities Exist for Expanding the Census
Applicant Pool Still Further
     If enacted, the legislation could make census
employment more attractive to millions of people.
However, the bills contain restrictions that limit
their applicability. In addition, other statutory
provisions exist that prohibit or create financial
disincentives for certain groups of people who
might be interested in census employment.

     Regarding the restrictions contained in the
measures currently before Congress, the exemption
in all three bills would not apply to federally
funded program beneficiaries who were appointed to
temporary census positions before January 1, 2000.
This could discourage some people who had worked
on early census-taking operations, such as address
list development activities, from seeking further
census employment.

     At the same time, the census applicant pool
is not as large as it could be because provisions
contained in current laws continue to prohibit or
potentially discourage large groups of people from
considering census jobs. Although our review was
not exhaustive, and we did not comprehensively
weigh the pros and cons of each option, we
identified three large sources of potential
applicants who might be interested in census
employment were it not for these provisions.

ï¿½    Active duty military personnel. The Census
Act allows uniformed personnel to take census jobs
to enumerate members of the uniformed services.12
However, active duty military personnel are
generally not permitted to accept outside federal
employment in the absence of specific statutory
authority to do so.13 Thus, additional statutory
authority would be needed to authorize military
personnel to work on the census. Doing so could
increase the potential census applicant pool by
over 1 million individuals.

ï¿½    Recipients of federal government voluntary
separation incentive payments. Since the early
1990s, as part of an effort to restructure the
federal government, the Department of Defense and,
later, civilian federal agencies, have had the
authority to offer voluntary separation incentive
payments (also known as buyouts) of as much as
$25,000 to eligible employees who left federal
service. For nondefense agencies, Congress has
authorized both governmentwide buyouts and over 15
agency-specific buyout programs. According to
information provided by OPM, most of these buyout
programs contain provisions that generally require
buyout recipients to repay their buyout if they
accept a federal job within 5 years of their
separation date. According to Bureau officials,
some buyout recipients have decided against census
jobs because of the repayment requirement.
Approximately 59,000 buyout recipients could
potentially still be covered by these provisions
during peak census field operations.
ï¿½    Noncitizens from certain countries. Most
federal agencies have historically been prohibited
by statute from using their appropriated funds to
employ noncitizens of the United States, with
certain exceptions. A statutory exemption from
this appropriations restriction (currently
contained in section 605 of Public Law 106-58)
exists that allows agencies to use appropriated
funds to employ noncitizens in limited
circumstances, such as for the temporary
employment of translators or in the field service
as a result of emergencies.  According to Bureau
officials, the Bureau has used this exemption to
hire temporary workers in the past and is
exploring its further use for the 2000 Census.
Nevertheless, a broad statutory exemption from
this appropriations restriction would make it
easier for the Bureau to hire noncitizens from
currently nonexempt countries, such as India,
Pakistan, and Brazil. Many of these individuals
could better enumerate members of their own
community because some hard-to-enumerate foreign-
born residents may feel more comfortable providing
information to persons with whom they share a
common cultural heritage.
Given the Bureau's past history of staffing
problems, the magnitude of the Bureau's staffing
challenge for 2000, and the importance of an
adequate workforce to the collection of timely and
accurate census data, it will be important for the
Bureau to have as large an applicant pool as
possible from which to hire census workers.

Post-Census Day Coverage Improvement Procedures
Offer Little Hope of Resolving the Undercount
The Bureau's post-census coverage improvement
procedures planned for 2000, while designed to
improve the census count, are similar to 1990
methods that had limited success. Bureau officials
believe that these procedures represent the best
the Bureau can reasonably do to enhance the
accuracy of the census. However, these officials
also said that they doubt that either the overall
accuracy levels or differential undercount rates
will show much improvement over 1990 levels
because societal factors that led to a high
undercount in 1990 are even more prevalent today.

Congress directed the Bureau to begin preparing
for a traditional census in November 1997.
However, the Bureau, awaiting the Supreme Court's
decision on the legality of sampling, did not
complete plans for a traditional census until
January 1999, when the Supreme Court ruled that
the Census Act prohibited the use of statistical
sampling for purposes of determining the
population count used to apportion the House of
Representatives.14 As a result, according to Bureau
officials, there was no time to conduct additional
research to estimate the effectiveness of new
coverage improvement procedures and, therefore,
too much risk to justify implementing them.  Thus,
for 2000, the Bureau will primarily use post-
census coverage improvement procedures used in
1990, which added just 3.68 million persons (1.5
percent) to the population count.

Nevertheless, coverage improvement programs in the
1990 Census were costly and yielded data of uneven
quality. For example, the Bureau's 1990 Recanvass
program-where enumerators did a second, post-
Census Day canvass of addresses in selected
neighborhoods to look for missed housing
units-added 139,000 housing units to the census,
which was 0.1 percent of the total, according to
Bureau data. However, while the $14.7 million
program added 178,000 people to the decennial
count, the Bureau later estimated that nearly 22
percent were added in error.

Overall, Bureau officials have acknowledged that
post-census coverage improvement programs are
expensive and do not always produce expected or
hoped for results. Thus, it is unlikely that the
Bureau's post-census coverage improvement
programs-programs that had limited success in
1990-will address the overall and differential
undercount in 2000.

Conclusions
With less than 4 months until Census Day, the
Bureau faces some significant risks that, taken
together, continue to jeopardize the success of
the 2000 Census. Securing an adequate level of
public participation is a great challenge with
implications for the size of the nonresponse
follow-up workload. Having to complete an even
greater nonresponse follow-up workload than
anticipated, or difficulty in filling the number
of enumerator positions that the Bureau estimates
it will need for this operation, would have
implications for scheduling as well as data
quality. Because of these combined risks, the 2000
Census may be less accurate than 1990.

Given the operational uncertainties surrounding
public participation in the census and the
Bureau's field follow-up operations, it will be
important for the Bureau to have contingency plans
in place to mitigate the impact of a lower-than-
expected response rate. Because of the little time
remaining and the need for senior Bureau managers
to devote the bulk of their attention to effective
execution of the census plans already in place,
such contingency plans will be most useful if they
focus on the critical challenges and trade-offs
that the Bureau will face-such as the need to
balance schedule pressures with the need to
protect data quality-if its response rate goals
are not met. In addition, even though the Bureau
has already taken steps to expand the census
applicant pool, additional statutory measures
could be needed, given the Bureau's past history
of staffing problems and the magnitude of the
Bureau's staffing challenge for 2000.

Matters for Congressional Consideration
To help expand the census applicant pool, Congress
may wish to consider legislative actions to modify
legal provisions that potentially discourage or
prohibit specific groups of people from seeking
census employment. Options could include:

ï¿½    expediting its consideration of H.R. 683, S.
752, and S. 1588, which among other things, would
remove financial disincentives that could
discourage recipients of Social Security, veterans
healthcare, food stamp, Medicaid, and Temporary
Assistance for Needy Families benefits, as well as
federal and military retirees from seeking census
employment;
ï¿½    allowing active duty military personnel to
hold temporary census employment;
ï¿½    exempting former federal employees who
received voluntary separation incentives (buyouts)
from requirements to repay their buyout amount if
they work on the census; and
ï¿½    providing a statutory exemption from the
appropriations restriction currently contained in
section 605 of Public Law 106-58, for purposes of
temporary census employment.

     Although we recognize that each of these
options entails policy, budgetary, and
implementation considerations that would need to
be addressed by Congress, they represent an
initial list of options that Congress could
consider to help reduce the Bureau's staffing
burden.

Recommendations
To help ensure an accurate and cost-effective
census, we recommend that the Director, Bureau of
the Census, develop a contingency plan of actions
the Bureau can take to address the operational
challenges that would result from a questionnaire
mail response rate that is lower-than-anticipated.
At a minimum, the Bureau's plan should address the
budgetary, scheduling, staffing, and other
logistical implications of collecting data from a
larger number of nonresponding households. The
contingency plan should also include options and
procedures to balance the pressure to meet census
schedules with the need to limit the use of proxy
data. The Bureau should share its plan with
Congress and others to demonstrate its
preparedness for collecting accurate census data
in the event of lower-than-expected levels of
public cooperation with the census.

Agency Comments and Our Evaluation
The Secretary of Commerce forwarded written
comments from the Bureau of the Census on a draft
of this report.  Overall, the Bureau commented
that the draft report conflicted with the intent
of our earlier reports, which, according to the
Bureau, concluded that there is little time to
make final census design changes and to implement
them, as Census Day approaches.

The Bureau noted that early in the decade, it
recognized the challenges to conducting a complete
and thorough nonresponse follow-up operation, and
it planned to address these challenges using
statistical sampling to adjust census population
counts. However, as we noted in our draft, in
January 1999, the Supreme Court ruled that the
Census Act prohibited the use of statistical
sampling for purposes of determining the
population count used to apportion the House of
Representatives. Our draft also noted that
according to the Bureau, there was insufficient
time to develop, test, and implement new coverage
improvement programs.

The Bureau also said that our contention that it
has not sufficiently planned for potential
shortcomings in nonresponse follow-up or outreach
and promotion operations appears to contradict our
September 1999 report on the Bureau's fiscal year
2000 amended budget request.15 According to the
Bureau, in that report, we characterized the
Bureau's expectations of enumerator productivity
during nonresponse follow-up, as well as the
effectiveness of the advertising campaign, as
being "generally conservative."

We revised the draft to include the language from
our September report, which noted the Bureau's
assumed enumerator productivity rate is 1.03
households per hour.  This new assumed
productivity rate represents a 20-percent
reduction from an original assumption of 1.28
households per hour and was primarily based on
senior management judgments-which the Bureau
acknowledged are very conservative-about factors
such as the uncertainty of hiring a sufficient
number of quality temporary workers in a tight
labor market.  We also have added language to this
report noting that higher-than-expected
productivity rates could reduce staffing needs.

Our September report did not state that the
Bureau's expectations of the impact of its
outreach and promotion campaign are conservative.
Rather, the September report states that the
Bureau had no data available to support how much,
if any, the Bureau's plans to increase the amount
of census advertising would increase the response
rate. Overall, the Bureau acknowledges that
completing nonresponse follow-up on time, hiring
and training needed staff, and implementing a
successful outreach and promotion campaign will be
a challenge-overriding themes of our report.

The Bureau further commented that there is
inconsistent analysis supporting the conclusions
in our draft. According to the Bureau, in some
instances, the dress rehearsal is used to support
certain conclusions (e.g., the mail response rate
in 2000 will be difficult to achieve). In other
cases, the Bureau notes that conclusions are drawn
that directly contradict the dress rehearsal
findings. We disagree with the Bureau's reading of
the draft report.  We were very careful in drawing
the lessons from the dress rehearsal and applying
them to the 2000 Census.  For example, in
discussing the challenges to motivating the public
to respond to the census, we also noted that an
augmented advertising campaign is planned for the
2000 Census and that mail response rates for the
actual census tend to be higher than response
rates obtained for a dress rehearsal.  Regarding
staffing, we noted that overall, the Bureau met
its dress rehearsal staffing goals, but that the
Bureau encountered pockets of problems in areas
with especially difficult labor markets.  The
draft noted that due to the number of staff the
Bureau will need to hire in 2000 and the
historically tight labor market, the Bureau faces
a substantial challenge-a view consistent with the
Bureau's.

The Bureau's final comment concerned our
recommendation calling on the Bureau to develop a
contingency plan of actions it can take if it
receives a lower-than-expected mail response rate
to the census questionnaire. The Bureau noted that
its paramount objective in the months remaining
before Census Day is to implement the procedures
and operations that have already been planned. The
Bureau commented that the only serious contingency
would be to request a supplemental appropriation.

We agree that the Bureau needs to concentrate on
successfully implementing the procedures and
operations already planned for the 2000 Census
and, as our draft noted, a lower-than-expected
mail response has major cost implications.
Nevertheless, prudent management and past history
suggest that developing a reasonable contingency
plan is an appropriate course of action.
Additional funding will not by itself make up for
a lower-than-expected mail response rate. It is
questionable whether the additional enumerators
that the Bureau will need to complete the
resulting increase in the nonresponse follow-up
workload will be available given the staffing
challenges described in our draft.  By considering
scheduling, data quality, staffing, and other
logistical implications of a lower-than-expected
mail response now-while time is still
available-the Bureau could be better prepared to
maintain the accuracy of census data.

We are sending copies of this report to the
Honorable William M. Daley, Secretary of Commerce,
and the Honorable Kenneth Prewitt, Director of the
Bureau of the Census. Copies will be made
available to others on request.

This report was prepared under the direction of J.
Christopher Mihm, Associate Director, Federal
Management and Workforce Issues. Please contact
Mr. Mihm on (202) 512-8676 if you have any
questions. Key contributors to this report are
included in appendix II.

Nancy Kingsbury
Acting Assistant Comptroller General
_______________________________
1Of the $6.8 billion, $4.5 billion is budgeted for
fiscal year 2000. For more information on the
Bureau's fiscal year 2000 budget request, see 2000
Census: Analysis of Fiscal Year 2000 Amended
Budget Request (GAO/AIMD/GGD-99-291, Sept. 22,
1999).
2See GAO High-Risk Series: Quick Reference Guide
(GAO/HR-97-2, Feb. 1997) and GAO Performance and
Accountability Series: Major Management Challenges
and Program Risks - Department of Commerce
(GAO/OCG-99-3, Jan. 1999).
3Decennial Census: 1990 Results Show Need for
Fundamental Reform (GAO/GGD-92-94, June 9, 1992).
4The mail response rate is calculated by dividing
the total number of questionnaires returned by the
total number of questionnaires delivered (either
by mail or by enumerators).
5GAO/GGD-92-94.
6According to a Bureau evaluation, the lower bound
represents the percentage of housing units that
mailed back only the replacement questionnaire.
The upper bound represents the percentage of
housing units that mailed back (1) the initial
questionnaire after the second was sent, (2) the
replacement questionnaire, or (3) both.
7See Census 2000 Dress Rehearsal Evaluation
Results Memorandum E1b, Effectiveness of the Paid
Advertising Campaign: Reported Exposure to
Advertising and Likelihood of Returning a Census
Form, U.S. Census Bureau, April 1999.
82000 Census: Preparations for Dress Rehearsal
Leave Many Unanswered Questions (GAO/GGD-98-74,
Mar. 26, 1998).
9The short-form questionnaire includes basic
population and housing questions and is to be
delivered to about 83 percent of all housing units
in 2000. The long-form questionnaire includes the
short-form questions, as well as additional
questions on the social and economic
characteristics of the population, and the
physical and financial characteristics of the
housing unit. The long form is to be delivered to
a sample of housing units that make up the
remaining 17 percent of the population.
10GAO/GGD-98-74.
11Immigration law and appropriation act
restrictions must still be considered in
determining noncitizen eligibility for temporary
census employment.
12See 13 U.S.C. 24 (c).
13See 64 Comp. Gen. 395 (1985).
14Department of Commerce v. U.S. House of
Representatives, 525 U.S. 316 (1999).
15GAO/AIMD/GGD-99-291.

Appendix I
Comments From the Secretary of Commerce
Page 37GAO/GGD-00-6 Contingency Planning Needed to
Address Risks

Appendix II
GAO Contacts and Staff Acknowledgments
Page 39GAO/GGD-00-6 Contingency Planning Needed to
Address Risks
GAO Contacts
J. Christopher Mihm, (202) 512-8676
Robert N. Goldenkoff, (202) 512-2757

Acknowledgments
     In addition to those named above, Victoria
Miller O'Dea, Victoria E. Miller, Lynn M.
Wasielewski, Anne K. Rhodes-Kline, James M. Rebbe,
Scott McNulty and Cindy S. Brown Barnes made key
contributions to this report.

*** End of Document ***