Decennial Census: Preliminary Observations on the Results to Date of the
Dress Rehearsal and the Census Bureau's Readiness for 2000 (Testimony,
07/30/98, GAO/T-GGD-98-178).

GAO discussed the Bureau of the Census' dress rehearsal for the 2000
Census and the Bureau's readiness for carrying out the 2000 Decennial
Census, focusing on how key census-taking operations have performed thus
far during the dress rehearsal and the implications that may exist for
2000.

GAO noted that: (1) when it last testified before Congress in March
1998, GAO noted that although the Census Bureau had made progress in
addressing some of the problems that occured during the 1990 Census, key
decennial census activities faced continuing challenges; (2) the census
dress rehearsal, currently underway at three sites, is the last
remaining field test before the decennial census is administered; (3)
within the constraints and limitations imposed by the dress rehearsal
setting, the Bureau to date has shown a general ability to implement the
dress rehearsal at the three locations according to its operational
timetable and plan; (4) certain census activities, such as staffing the
dress rehearsal operations and completing field operations on schedule,
appear to have gone well; (5) however, the dress rehearsal experiences
also have underscored the fact that the Bureau still faces major
obstacles to a cost-effective census; (6) for example, mail response
rates remain problematic, and local partnerships had limited success;
(7) further, the Bureau's general ability to conduct the dress rehearsal
according to its operational plan, while encouraging, is not necessarily
a predictor of success in 2000; (8) because the dress rehearsal was
performed at three sites, the capacity of regional and headquarters
offices, as well as a number of essential census-taking operations,
could not be fully tested under census-like conditions; and (9) the most
important outcome measure--the quality of the census data collected--is
not yet available.

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

 REPORTNUM:  T-GGD-98-178
     TITLE:  Decennial Census: Preliminary Observations on the Results 
             to Date of the Dress Rehearsal and the Census
             Bureau's Readiness for 2000
      DATE:  07/30/98
   SUBJECT:  Census
             Data collection
             Data integrity
             Mailing lists
             Population statistics
             Statistical methods
             Quality assurance
             Human resources utilization
             Interagency relations
             Reengineering (management)
IDENTIFIER:  2000 Decennial Census
             Columbia (SC)
             Menominee County (WI)
             Sacramento (CA)
             Menominee Indian Reservation (WI)
             
******************************************************************
** 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


Before the Committee on Governmental Affairs
U.S.  Senate

For Release on Delivery
Expected at
10:00 a.m.  EDT
Thursday
July 30, 1998

DECENNIAL CENSUS - PRELIMINARY
OBSERVATIONS ON THE RESULTS TO
DATE OF THE DRESS REHEARSAL AND
THE CENSUS BUREAU'S READINESS FOR
2000

Statement of
J.  Christopher Mihm
Associate Director, Federal Management
 and Workforce Issues
General Government Division

GAO/T-GGD-98-178

GAO/GGD-98-178T


(410353)


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


PRELIMINARY OBSERVATIONS ON THE
RESULTS TO DATE OF THE DRESS
REHEARSAL AND THE CENSUS BUREAU'S
READINESS FOR 2000
====================================================== Chapter SUMMARY

When GAO last testified before Congress in March 1998, it noted that,
although the Census Bureau had made progress in addressing some of
the problems that occurred during the 1990 Census, key decennial
census activities faced continuing challenges.  The situation today
is much the same. 

The census dress rehearsal, currently underway at three
sites--Sacramento, CA; 11 counties in the Columbia, SC area; and
Menominee County in Wisconsin, including the Menominee American
Indian Reservation--is the last remaining field test before the
decennial census is administered.  Within the constraints and
limitations imposed by the dress rehearsal setting, the Bureau to
date has shown a general ability to implement the dress rehearsal at
the three locations according to its operational timetable and plan. 
Certain census activities, such as staffing the dress rehearsal
operations and completing field operations on schedule, appear to
have gone well. 

However, the dress rehearsal experiences also have underscored the
fact that the Bureau still faces major obstacles to a cost-effective
census.  For example, mail response rates remain problematic, and
local partnerships had limited success.  Further, the Bureau's
general ability to conduct the dress rehearsal according to its
operational plan, while encouraging, is not necessarily a predictor
of success in 2000.  Because the dress rehearsal was performed at
three sites, the capacity of regional and headquarters offices, as
well as a number of essential census-taking operations, could not be
fully tested under census-like conditions.  Finally, the most
important outcome measure--the quality of the census data
collected--is not yet available. 


PRELIMINARY OBSERVATIONS ON THE
RESULTS TO DATE OF THE DRESS
REHEARSAL AND THE CENSUS BUREAU'S
READINESS FOR 2000
==================================================== Chapter STATEMENT

Mr.  Chairman and Members of the Committee: 

I am pleased to be here today to provide an update on the Census
Bureau's dress rehearsal for the 2000 Census and the Bureau's
readiness for carrying out the 2000 Decennial Census.  The dress
rehearsal, currently under way at three sites--Sacramento, CA; 11
counties in the Columbia, SC area; and Menominee County in Wisconsin,
including the Menominee American Indian Reservation-- is designed to
demonstrate major operations, procedures, and questions that are
planned for the decennial census. 

At your request, my statement focuses on how key census-taking
operations have performed thus far during the dress rehearsal and the
implications that may exist for 2000.  When we last testified before
Congress in March 1998, we noted that, although the Bureau had made
progress in addressing some of the problems that occurred during the
1990 Census, key activities faced continuing challenges.\1 The
situation today is much the same.  On the one hand, certain census
activities, such as staffing the dress rehearsal operations, appear
to have gone well.  On the other hand, measures of other activities,
such as the mail response rate, suggest that the Bureau still faces
major obstacles to a cost-effective census.  Moreover, while the
dress rehearsal activities done thus far have demonstrated the
Bureau's general ability to execute the dress rehearsal according to
its operational timetable and plan, the important outcome
measure--the quality of the data collected--is not yet available. 
Further, the Bureau's general ability to conduct the dress rehearsal
according to its operational plan, while encouraging, is not
necessarily a predictor of success in 2000.  Because the dress
rehearsal was performed at three sites, the capacity of regional and
headquarters offices, as well as a number of essential census-taking
operations, could not be fully tested under census-like conditions. 

As you know, Mr Chairman, uncertainty continues to surround the final
design of the decennial census.  Congress has not endorsed the
Bureau's planned use of statistical sampling to improve the accuracy
of the population counts because of Congressional concerns over the
validity, legality, and operational feasibility of the Bureau's
statistical sampling and estimation procedures.  The Bureau is now
planning for both a sampling and nonsampling census until a final
decision on the design is made.  Elements of each are being tested
during the dress rehearsal.  The Bureau is using sampling and
statistical estimation methods at the Sacramento site, in accordance
with its plans for a sampling census.  At the South Carolina site,
the Bureau's procedures are to follow up on all nonresponding
households just as it was to do nationwide in the 1990 Census.  At
the Menominee dress rehearsal site, the Bureau is to follow up on all
nonresponding households, but it is also using sampling and
statistical estimation to improve the accuracy of the population
count. 

My comments today are based on our ongoing review of key census-
taking operations that could significantly affect the cost and
accuracy of the 2000 Census.  They include such activities as (1)
creating a complete and accurate address list, (2) obtaining a high
level of public cooperation through an effective census promotion and
outreach effort, (3) staffing census-taking operations with an
adequate workforce, (4) processing census data accurately and using
technology efficiently and effectively, and (5) carrying out field
activities including both nonresponse follow-up and sampling and
statistical estimation procedures. 

To assess these activities, we (1) made several visits to the dress
rehearsal sites and the Bureau's data capture center in
Jeffersonville, IN; (2) observed key census-taking operations; (3)
interviewed Bureau headquarters officials, staff from regional and
local census offices, and individual enumerators and their
supervisors; and (4) reviewed relevant documents and data the Bureau
prepared about these operations.  To obtain a local perspective on
the dress rehearsal, we conducted in-person and telephone interviews
with local officials at the three dress rehearsal sites on their
experiences in reviewing address lists, promoting the census, and
recruiting and hiring census workers. 

Because the dress rehearsal is still under way and more comprehensive
data on the results of the dress rehearsal are not yet available, our
observations today should be considered preliminary and the Bureau's
data are subject to change pending further refinements and analysis. 


--------------------
\1 Decennial Census:  Preparations for Dress Rehearsal Underscore the
Challenges for 2000 (GAO/T-GGD-98-84, Mar.  26, 1998). 


   DRESS REHEARSAL EXPERIENCES
   CONFIRM NEED FOR REVISED
   APPROACH TO DEVELOPING ADDRESS
   LIST
-------------------------------------------------- Chapter STATEMENT:1

One of our long-standing concerns has been the Bureau's ability to
build a complete and accurate address list and develop precise maps. 
Accurate addresses are critical for delivering questionnaires,
avoiding unnecessary and expensive follow-up efforts at vacant or
nonexistent residences, and establishing a universe of households for
sampling and statistical estimation.  Precise maps are essential for
counting persons at their proper locations--the cornerstone of
congressional reapportionment and redistricting.  Bureau maps are
also used for certain census- taking operations such as nonresponse
follow-up that entails following up on households that fail to mail
back a census questionnaire. 

To build its address list, which is known as the Master Address File
(MAF), the Bureau initially planned, in part, to (1) use addresses
provided by the Postal Service, (2) merge these addresses with the
address file the Bureau created during the 1990 Census, (3) conduct
limited checks of the accuracy of selected addresses, and (4) send
the addresses to local governments and Indian tribes for verification
as part of a process called Local Update of Census Addresses. 

However, as we reported in March 1998, the Bureau concluded in
September 1997 that its reliance on postal and 1990 Census addresses
to construct its 2000 Census address list would not yield a
sufficiently complete and accurate list.\2 The Bureau therefore
decided that redesigned procedures were needed in order to generate a
MAF for the 2000 Census that, as a whole, was 99 percent complete. 
Under the revised approach, after local address review, the Bureau
plans to verify physically the completeness and accuracy of the
address file for the 2000 Census by canvassing neighborhoods across
the country.  The Bureau expects the new approach will cost an
additional $108.7 million. 

The experiences of the dress rehearsal suggest that the Bureau was
correct in the decision to revise its procedures for building its
address list.  One early indicator of the accuracy of the address
list is the number of census questionnaires returned to the Bureau by
the Postal Service as undeliverable.  Of the approximately 394,000
questionnaires mailed at the Sacramento and South Carolina dress
rehearsal sites, Bureau data show that about 48,000 (12.2 percent)
were returned as undeliverable.\3 According to a Bureau official, the
Bureau had expected to do better than that, based on the 1995 test
census results.  For the 1995 test census, about 7.7 percent of the
census questionnaires were reported to be undeliverable at the
Oakland, California test site and 4.5 percent at the Paterson, New
Jersey test site. 

In addition, the census maps appeared to be of uneven quality and
usefulness at the dress rehearsal locations.  For example, local
census officials in Sacramento and South Carolina said that the
census maps were inaccurate and contained a variety of errors, such
as streets that were incorrectly placed and named.  In both
locations, problems with census maps led some enumerators to use
commercially available maps rather than those supplied by the Bureau. 
In Menominee, because of the rural nature of the site, maps were
particularly important.  Houses generally lacked numbered street
addresses, and, as a result, enumerators had to locate them, in part,
by using maps.  However, Bureau officials told us that while the
quality of the Menominee maps is improving over the course of the
dress rehearsal, the maps still have problems that make it difficult
for enumerators to locate houses. 

As I noted, the Bureau recognized that it needed to revise its
approach to building the census address list and to improve the
quality of its map products.  However, the Bureau's revised approach
to developing its address list is not without risk.  Although
elements of the revised approach have been used and tested in earlier
censuses, the Bureau has not used or tested them together, nor in the
sequence as presently designed for the 2000 Census.  Furthermore,
because the Bureau made the decision to change its address list
development procedures in September 1997--after major dress rehearsal
address list development efforts were already in place--the revised
approach was not used during the dress rehearsal.  As a result, it
will not be known until the 2000 Census whether the Bureau's
redesigned procedures will allow it to meet its goal of a 99 percent
complete address list.  The Bureau is scheduled to begin its 2000
Census field canvassing address list efforts in August.  We will
continue to monitor the Bureau's efforts to build the census address
list. 


--------------------
\2 GAO/T-GGD-98-84, Mar.  26, 1998. 

\3 At the Menominee site, census questionnaires were distributed
using a process called "update/leave" whereby census enumerators
hand-deliver questionnaires to each household, update the address,
and leave the questionnaire for respondents to mail back. 


   OBTAINING PUBLIC COOPERATION
   CONTINUES TO BE A CHALLENGE FOR
   THE CENSUS BUREAU
-------------------------------------------------- Chapter STATEMENT:2

The Bureau plans an extensive outreach and promotion campaign to
boost mail response rates and thus reduce its error prone and costly
nonresponse follow-up workload.  For the 2000 Census, the Bureau
believes that its outreach and promotion program, when combined with
other initiatives such as simplified census questionnaires, should
produce a mail response rate of 66.9 percent, 12 percentage points
higher than the 55 percent response rate that the Bureau expected it
would achieve without these efforts. 

The Bureau always finds that mail response rates during census tests,
including the dress rehearsal, are lower than those obtained during
an actual decennial census, when public awareness of the census is
generally much greater.  Table 1 shows the anticipated dress
rehearsal mail response rates for the three sites and the rates the
Bureau actually achieved. 



                                Table 1
                
                 Anticipated and Actual Dress Rehearsal
                 Mail Response Rates as of May 7, 1998

                                  Anticipated mail         Actual mail
Site                                 response rate       response rate
------------------------------  ------------------  ------------------
South Carolina                                 55%               54.1%
Sacramento                                      50                53.7
Menominee                                       40                40.6
----------------------------------------------------------------------
Note:  The Bureau calculates mail response rates by dividing the
total number of questionnaires returned (either by mail or by
enumerators) by the total number of questionnaires delivered. 

Source:  U.S.  Census Bureau. 

Despite the fact that the Bureau generally met its response rate
goals for the dress rehearsal, significant concerns remain about the
degree to which the Bureau will be able to meets its mail response
goal for 2000.  By way of comparison, the 1988 dress rehearsal for
the 1990 Census generated mail response rates that ranged from 49
percent to 56 percent for mailout/mailback operations, and 58 percent
for update/leave operations.\4 The mail response rate to the 1990
Census was 65 percent--slightly less than the 67 percent response
rate that the Bureau hopes for in 2000. 

More importantly, the Bureau does not currently plan to use in 2000 a
key ingredient of the response rate achieved during the dress
rehearsal--a second mailing.  According to a Bureau official,
concerns about public confusion have contributed to the Bureau's
decision not to use a second questionnaire mailing in 2000.  The
preliminary results of the dress rehearsal suggest that the Bureau
may need to reconsider its decision.  At both the South Carolina and
Sacramento sites, the Bureau obtained approximately a 7-percentage
point "bump" in response rates by sending a second questionnaire to
all households located in mailout/mailback areas.  According to a
senior Bureau offical, this 7 percentage point increase represents
real additions to the count and does not include duplicate
submissions from households that already had responded. 

The Bureau traditionally has found that simply raising awareness of
the census is insufficient; through its various outreach and
promotion programs, the Bureau must also motivate people to return
their questionnaires.  The difficulty in doing this was demonstrated
during the 1990 Census when the Bureau found that, although about 93
percent of the public was aware of the census, the mail response rate
was only 65 percent--10 percentage points lower than the mail
response rate to the 1980 Census. 

Today, I will highlight two of the more important components of the
Bureau's efforts to build public awareness and cooperation through
its outreach and promotion campaign:  paid advertising and
partnerships and community outreach. 


--------------------
\4 The 1988 dress rehearsal was carried out in St.  Louis, east-
central Missouri, and eastern Washington. 


      PAID ADVERTISING
------------------------------------------------ Chapter STATEMENT:2.1

With regard to the Bureau's paid advertising campaign, in October
1997, the Bureau announced it had hired Young & Rubicam, a private
advertising agency, to market the census.  The advertising campaign
is based on the theme "This is your future-- don't leave it [blank]"
and stresses how responding to the census questionnaire benefits
one's community.  This advertising effort was evident during our
visits to the dress rehearsal sites, where we often observed
billboards bearing Census 2000 advertising messages, such as "How
America Knows What America Needs," "The Future Takes Just a Few
Minutes to Complete," and "Pave a Road With These Tools." In
convenience stores, we observed signs that told passers-by that "[The
Census] Gives Life to New Healthcare Centers." In Sacramento, we
observed outdoor advertising in languages appropriate for the
neighborhood.  The census was also promoted through broadcast and
print media, as well as through less traditional methods such as
advertisements on shopping bags at a chain of discount stores. 

About $100 million has been budgeted for the paid advertising
campaign for the 2000 Census, of which about 70 percent is earmarked
for buying advertising in print and broadcast media.  According to
the Bureau, for fiscal year 1998, dress rehearsal paid advertising
expenses included such activities as

  -- $0.31 million for research, including pre- and post-campaign
     focus groups, and telephone interviews at the dress rehearsal
     sites;

  -- $2.0 million for development, including the production of radio
     and television advertisements;

  -- $0.82 million for South Carolina media costs;

  -- $0.35 million for production and media costs for nontraditional
     advertising in South Carolina;

  -- $0.23 million for Menominee media costs; and

  -- $1.12 million for Sacramento media costs. 


      COMMUNITY OUTREACH
------------------------------------------------ Chapter STATEMENT:2.2

The Bureau's use of partnership and community outreach activities
and, in particular, its use of Complete Count Committees to help
promote the census are other key components of the Bureau's outreach
and promotion campaign.  According to the Bureau, Complete Count
Committees are intended to help the Bureau take the census by, among
other activities, planning and implementing a locally-based promotion
effort to publicize the importance of the 2000 Census.  The
committees are to consist of local leaders, such as representatives
of government, education, media, community, religious, and businesses
organizations.  For the dress rehearsal, the Bureau attempted to form
committees in Sacramento and Menominee, as well as in the City of
Columbia and the 11 surrounding counties participating in the dress
rehearsal.  The Bureau recommended that the committees could, among
other initiatives,

  -- form subcommittees to reach specific segments of the population
     such as senior citizens;

  -- sponsor promotional events;

  -- obtain commitments from businesses to promote and support the
     census;

  -- provide the Bureau with testing and training space to assist in
     the employment of enumerators; and

  -- work with local media to cover and publicize census activities. 

This past spring, the Bureau sent a Complete Count Committee
handbook, in which the Bureau described its plan for implementing the
Complete Count Committee program for the 2000 Census, to the highest
elected officials in about 39,000 local and tribal governments.  The
handbook suggested a structure for organizing a grassroots outreach
campaign and provided an outline and schedule of nearly five-dozen
activities that governments could undertake not only to promote the
census, but also assist the Bureau in its data collection and
enumerator recruiting responsibilities as well. 

The success of the Bureau's efforts to form partnerships with local
governments through Complete Count Committees depends, in large part,
on the Bureau having realistic and clearly communicated expectations
for what the Bureau can anticipate from the committees and, just as
important, what the committees can expect from the Bureau.  The
Bureau expects that the committees will secure their own funding and
will rely on the Bureau for only a very limited amount of direct
assistance.  For example, at the dress rehearsal site in South
Carolina, the Bureau hired two partnership specialists to help
mobilize local groups.  These specialists had to distribute their
time and energy among the City of Columbia and the 11 surrounding
counties included in the dress rehearsal--a workload that is
consistent with what will be expected in 2000 when the Bureau plans
to have 320 partnership specialists in place across the nation. 

Our work at the dress rehearsal sites suggests that the effectiveness
of the partnership effort was undermined by an apparent mismatch
between the Bureau's expectations of the committees and what the
committees could realistically accomplish.  In both South Carolina
and Menominee, a message we consistently heard from local officials
associated with the committees was that they lacked the human and
financial resources to promote the census, communication and guidance
from the Bureau were insufficient, and Bureau assistance was limited. 
As a result, Complete Count Committees in some South Carolina
counties were never formed, while others became inactive and some
local officials expressed confusion and frustration over what was
expected. 

Local outreach and promotion appeared to go more smoothly in
Sacramento.  This was likely due in part to the fact that there was
only one Bureau partnership specialist in Sacramento assisting only
one Complete Count Committee for Sacramento.  However, as I have
noted, for the 2000 Census, workloads for the Bureau partnership
specialists closer to those I have described for South Carolina are
more likely be the norm. 

Overall, therefore, the dress rehearsal experience suggests that the
Bureau needs to ensure that it has realistic expectations about the
contributions that Complete Count Committees will be able to make in
promoting the census, building the response rate, and assisting the
Bureau. 


   THE DRESS REHEARSAL HAS NOT
   ENCOUNTERED SIGNIFICANT
   STAFFING PROBLEMS TO DATE
-------------------------------------------------- Chapter STATEMENT:3

Recruiting, hiring, training, and retaining a workforce to carry out
the 2000 Census is clearly one of the government's great human
resource challenges.  The Bureau estimates that, under its current
design, it will need to fill about 295,000 office and field positions
to carry out various census-taking activities during the 2000 Census. 
To fill this many positions, the Bureau estimates, based on past
experience, that it will need to recruit as many as 2.6 million
applicants, because for a variety of reasons, most applicants never
make it through the employment process. 

Despite the uncertainties surrounding the Bureau's ability to staff
the 2000 Census, staffing the dress rehearsal appears to have gone
better than expected thus far.  As shown in table 2, one measure of
the success of the Bureau's staffing efforts, applicants' acceptance
of job offers for nonresponse follow-up (where the demand for
employees is greatest), far exceeded the Bureau's expectations. 



                                Table 2
                
                     Job Offer Acceptance Rate for
                    Nonresponse Follow-up Operations

                                       Anticipated   Actual acceptance
Site                               acceptance rate                rate
------------------------------  ------------------  ------------------
South Carolina                                 50%                 90%
Sacramento                                      50                  78
Menominee                                       50                  71
----------------------------------------------------------------------
Source:  U.S.  Census Bureau. 

Moreover, managers of the local Census Bureau offices at the dress
rehearsal sites we spoke to said that the quality of the newly hired
employees' work was typically good.  According to Bureau data, at all
three dress rehearsal sites, enumerator productivity came very close
to the Bureau's goal of 1.5 nonresponse follow-up cases completed per
hour and enumerator turnover appears to have been lower than
expected. 

The Bureau attributes its apparently successful dress rehearsal
staffing efforts to several factors, including a competitive pay plan
and aggressive recruitment.  Key features of the Bureau's pay plan
include locality-based wages and bonuses for exceeding production
targets.  In addition, when the Bureau recognized that it was having
difficulty recruiting a large enough pool of qualified applicants to
fill its needs for nonresponse follow-up and later census operations
in South Carolina, the Bureau raised enumerator pay rates from $9.50
per hour to $10.50 per hour effective April 3, 1998.  Enumerator pay
was $12.50 per hour in Sacramento and $11.25 per hour in Menominee. 

The Bureau used a wide variety of methods to recruit applicants. 
Among these were the mass mailing of recruiting literature to
residents, posters, flyers, and newspaper advertisements and
articles.  During our visits to the dress rehearsal sites, we
frequently observed recruiting notices in such public locations as
post offices, local government office buildings, and public
libraries.  In fact, recruiting literature appeared to be more
prevalent than materials that promoted the census itself. 


   QUESTIONS REMAIN ABOUT THE
   QUALITY OF DATA PROCESSING AND
   THE USE OF TECHNOLOGY
-------------------------------------------------- Chapter STATEMENT:4

Translating data from completed census forms into a useable format
represents another challenge for the Bureau.  The Bureau plans to
have data capture centers process a total of about 1 billion pages of
census questionnaires in 99 work days beginning in March 2000.  The
Bureau plans to take advantage of commercial off-the-shelf hardware
and software through its contractor Lockheed Martin, rather than rely
on in-house products.  During the dress rehearsal, the Bureau is
testing the accuracy of the data input by the new scanning equipment
and software designed to perform this operation.  Bureau officials
reported that this operation met all high-priority processing
deadlines, despite experiencing system bugs that will need to be
addressed before 2000. 

The purpose of the dress rehearsal was to test and debug the system
in an operational environment in advance of Census 2000.  However,
additional load testing is still necessary because the system could
not be run during the rehearsal at performance levels that will be
needed in 2000.  During the dress rehearsal, the scanning equipment
used to electronically record responses off census forms experienced
system crashes due to flaws in the software.  To deal with this
problem, the Bureau was forced to cut back the number of scanners in
operation at any one time.  According to Bureau officials, the
software subcontractor, is resolving this and other problems through
intensive testing, and will have a new version of its software
available for further testing in late August. 

According to Bureau officials, another problem related to scanning is
the frequency at which the scanners needed to be cleaned of
accumulated dust.  Initially, the Bureau had planned to clean the
machines every 2 hours.  However, dust accumulated faster than
expected, which necessitated a 5-minute cleaning after each 15
minutes of use.  Bureau officials said that poor paper quality
appears to be one factor that led to the accumulation of dust.  The
Bureau and the Government Printing Office are studying the problem. 

To date, test results of the accuracy of scanned data during the
dress rehearsal are not available.  The Bureau plans to continue to
test its data capture system using census-like workloads in late
August and complete its final evaluation later this year.  The Bureau
plans to run actual dress rehearsal forms through the system--as well
as a database of computer-generated images--to test the performance
of its scanning equipment.  Bureau officials believe that sufficient
time remains to complete more testing, incorporate lessons learned
from the dress rehearsal, and make technology enhancements before
Census 2000. 


   CRITICAL OPERATIONS REMAIN TO
   BE COMPLETED BUT FIELD
   OPERATIONS TO DATE FINISHED ON
   SCHEDULE
-------------------------------------------------- Chapter STATEMENT:5

Of the Bureau's numerous field operations, two of the largest and
most logistically challenging under the Bureau's current design are
nonresponse follow-up and a procedure called Integrated Coverage
Measurement (ICM), a survey in which residents in a sample of blocks
are interviewed.  ICM and enumeration data are used in dual system
estimation to adjust for coverage errors in the enumeration. 

As currently planned, the Bureau is to reduce its nonresponse
follow-up workload for the 2000 Census by sampling nonresponding
households.  By using a sample-based nonresponse follow-up, the
Bureau would reduce the time necessary to complete this activity. 
This in turn would expedite the beginning of ICM data collection,
improving the Bureau's ability to meet the target date for delivery
of census data at the end of December.  In addition, compressing the
nonresponse follow-up data collection period could shorten the
average time between census day and visits to households, thereby
reducing the likelihood of enumeration errors caused by households
that move between census day and nonresponse follow-up. 

The Bureau plans to conduct a nationwide ICM.  However, as noted
earlier, for the dress rehearsal, the Bureau only sampled for
nonresponse and is conducting the ICM in Sacramento.  In South
Carolina, the Bureau procedures are to follow up on all nonresponding
households and do a coverage evaluation operation, just as it did
nationally in the 1990 Census.  At the Menominee site, the Bureau is
to follow up on all nonresponding households and additionally, is
using the ICM. 

To preserve the integrity of the adjustment process, enumeration and
ICM operations must be kept independent of one another.  If ICM
respondents became aware of their selection for the ICM, this could
alter their responses to, or willingness to participate in, the
enumeration.  To the extent that this occurred, it could bias the
data and result in a less accurate measure of any coverage errors in
the enumeration.  In an attempt to maintain their necessary
independence, nonresponse follow-up and ICM activities are to be
completed according to a tight schedule, with ICM activities
commencing only after nonresponse follow-up has been completed. 
During the dress rehearsal, nonresponse follow-up was completed on
time at the Sacramento and Menominee sites, and about a week ahead of
schedule in South Carolina.  We observed that ICM operations began as
scheduled in Sacramento.  Major ICM field operations are scheduled to
last until late August 1998. 

The Bureau's procedures called for it to take additional steps to
prevent contamination of the ICM data.  According to Bureau
officials, these included efforts to separate the management and
implementation of the ICM operation from the nonresponse follow-up
operation.  For example, the ICM operation was administered entirely
by the Bureau's Seattle Regional Office rather than the local census
office in Sacramento.  Additionally, nonresponse follow-up
enumerators were not told which blocks were included in the ICM, and
ICM enumerators were told that they could not tell anyone which
blocks had been assigned to them. 

The quality of the dress rehearsal data, as measured by the extent to
which it is complete and accurate, is still to be determined.  With
the ICM still in progress, the full results of the ICM will not be
known for several months.  Moreover, a key question for which
information is not yet available is the degree to which the Bureau
had to rely on proxy responses from neighbors, letter carriers, and
others to complete its nonresponse workload in a timely manner.  As
part of our ongoing work, we will review the quality of the data
collected during the ICM and nonresponse follow-up operations, the
Bureau's procedures for maintaining the independence of enumeration
and ICM data, and, more generally, the extent to which the Bureau was
able to implement its field operations as planned. 

In summary, Mr.  Chairman, within the constraints and limitations
imposed by the dress rehearsal setting, the Bureau to date has shown
a general ability to implement the dress rehearsal at the three sites
according to its operational timetable and plan.  The Bureau has also
shown an ability to adapt to changing requirements as demonstrated by
such actions as redesigning its address list development procedures
to produce a more accurate and complete list and by increasing wage
rates in South Carolina to improve recruiting. 

However, the dress rehearsal also emphasized the formidable
challenges to implementing a successful census that confront the
Bureau.  For example, mail response rates remain problematic and
local partnerships have had limited success.  Moreover, key dress
rehearsal operations remain to be completed and, as noted earlier,
the fundamental indicators of a successful census--the quality of the
data, including the accuracy of the population count and the extent
to which proxy data are used--are not yet available. 

Mr.  Chairman, this concludes my prepared statement.  I would be
pleased to answer any questions you or Members of the Committee may
have. 

*** End of document. ***