2000 Census: Analysis of Fiscal Year 2000 Amended Budget Request (Letter
Report, 09/22/1999, GAO/AIMD/GGD-99-291).

Pursuant to a congressional request, GAO provided information on the
Bureau of the Census' fiscal year (FY) 2000 budget, focusing on: (1) an
overall analysis of the key changes in assumptions resulting in the
+$1.7 billion request increase; (2) details on the components of this
increase and which changes, according to the bureau, are related and
which are not related to the inability to use statistical sampling; and
(3) the process the bureau used for developing the increase in the
original FY 2000 budget request and the amended budget request.

GAO noted that: (1) GAO found that the net $1.7 billion requested
increase in the original FY 2000 budget request of $2.8 billion resulted
primarily from changes in assumptions relating to a substantial increase
in workload, reduced employee productivity, and increased advertising;
(2) under the nonsampling design, census costs will increase because the
bureau expects to follow up on more nonresponding households than for a
sampling-based census and plans to use additional programs to improve
coverage because it cannot rely on statistical methods to adjust for
undercounting and other coverage errors; (3) the bureau assumes an
increased workload because the housing units that the bureau expects to
visit have increased from an estimated 30 million to 46 million; (4) the
16 million increase includes visiting 12 million additional
nonresponding housing units and 4 million additional housing units that
the Postal Service says are vacant or nonexistent; (5) this increased
workload, which increased costs for most bureau program activities,
relates primarily to additional salaries, benefits, travel, data
processing, infrastructure, and supplies; (6) another key factor
substantially increasing the FY 2000 budget request is that the bureau
reduced the assumed productivity of its temporary enumerator employees;
(7) because of the assumed increase in workload and reduction in
productivity, the total number of temporary field positions increased
from 780,000 in the original budget to 1,350,000 in the amended budget
request; (8) the bureau also included nearly $72 million of advertising
in the amended budget request to increase public awareness and hopefully
increase response rates for mailed questionnaires; (9) according to the
bureau, about $1.6 billion of this increase is related and $100 million
is not related to the inability to use statistical sampling; (10) the
items unrelated to the sampling issue include costs not included in the
original budget and revisions of prior estimates; (11) the bureau
developed its $4.5 billion amended budget request for FY 2000 using a
cost model consisting of a series of interrelated software spreadsheets;
(12) the original and amended budget requests were developed using this
cost model, with each estimate being developed independently using
different versions of the cost model; and (13) the bureau derived the
$1.7 billion requested increase by calculating the net difference
between the original budget request of $2.8 billion and the amended
budget request of $4.5 billion.

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

 REPORTNUM:  AIMD/GGD-99-291
     TITLE:  2000 Census: Analysis of Fiscal Year 2000 Amended Budget
	     Request
      DATE:  09/22/1999
   SUBJECT:  Cost analysis
	     Statistical methods
	     Human resources utilization
	     Budget amendments
	     Census
	     Presidential budgets
	     Surveys
IDENTIFIER:  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.  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.                         **
******************************************************************

Report to the Chairman, Subcommittee on the Census, Committee on
Government Reform, House of Representatives

September 1999

2000 CENSUS

Analysis of Fiscal Year 2000 Amended Budget
Request
*****************

*****************

GAO/AIMD/GGD-99-291

Letter                                                                     3

Appendixes

Appendix I:Briefing on the Analysis of the Census Bureau's Fiscal Year
2000 Amended Budget Request

                                                                         20

Appendix II:Comments From the Department of Commerce

                                                                         65

Appendix III:GAO Contact and Staff Acknowledgements

                                                                         70

Table 1:  Comparison of the Bureau's Original and Amended
Budget Request                                  15

Figure 1:  Census Full-Cycle Cost per Housing Unit--Fiscal Years
1970 to 2000 (in constant fiscal year 1998 dollars)8

Figure 2:  Increase in Housing Units Visited Between the Original
and Amended Budget Requests                     11

ACE     Accuracy and Coverage Evaluation

FTE     full-time equivalent

ICM     Integrated Coverage Measurement

LCO     Local Census Office

MOU     Memorandum of Understanding

TQA     Telephone Questionnaire Assistance

                                                 Accounting and Information
                                                        Management Division

B-283148

September 22, 1999

The Honorable Dan Miller
Chairman
Subcommittee on the Census
Committee on Government Reform
House of Representatives

Dear Mr. Chairman:

The decennial census is the nation's most comprehensive and expensive data-
gathering program. The Constitution requires a decennial census of the
population in order to apportion seats in the House of Representatives.
Public and private decisionmakers also use census data on population
counts and social and economic characteristics for a variety of purposes.
Since 1970, the Census Bureau has used essentially the same methodology to
count the vast majority of the population during the decennial census. The
bureau develops an address list of the nation's households and mails
census forms (questionnaires) to those households asking the occupants to
mail back the completed forms. The bureau hires temporary census takers,
known as enumerators, by the hundreds of thousands to gather the requested
information for each nonresponding household./Footnote1/

For the 2000 census, the bureau planned to augment the traditional census
methodology with statistical estimation to develop a unified census count.
In November 1997, in the Department of Commerce and Related Agencies
Appropriations Act for 1998, the Congress questioned the constitutionality
of using statistical sampling and directed the bureau to plan to implement
a census in 2000 without using statistical methods. The Census Bureau
reported to the Congress on possible components in a traditional census
plan in April 1998. However, the bureau did not begin detailed budgeting
for a nonsampling-based census until after the Supreme Court ruled that
the Census Act/Footnote2/ prohibited the use of statistical sampling for
purposes of determining the population count used to apportion the House
of Representatives.

As recently as August 1997, the bureau estimated that without sampling,
the cost of the 2000 census would increase by $675 million to $800 million
and would be less accurate than the 1990 census. Following the Supreme
Court's decision, in June 1999, the bureau requested an increase of over
$1.7 billion to its original fiscal year 2000 budget request of nearly
$2.8 billion. Your office advised us that the information provided by the
bureau did not provide a satisfactory explanation of how this amended
budget request was developed and the reasons for the $1.7 billion
requested increase. 

As a result of this greater than expected increase in the bureau's fiscal
year 2000 budget and your need for additional information, you requested
that we (1) provide an overall analysis of the key changes in assumptions
resulting in the $1.7 billion requested increase, (2) provide details on
the components of this increase and which changes, according to the
bureau, are related and which are not related to the inability to use
statistical sampling, and (3) describe the process the bureau used for
developing the increase in the original fiscal year 2000 budget request
and the amended budget request. This report provides a summary of the
information presented at our September 15, 1999, briefing to you and the
House and Senate appropriations subcommittees. The briefing slides in
appendix I have been updated to incorporate additional information you
requested at the briefing as well as information subsequently provided by
the bureau. Also at your request, we will shortly be issuing a status
report on key census-taking operations and preparation for the 2000 census.

Results in Brief 

We found that the net $1.7 billion requested increase in the original
fiscal year 2000 budget request of $2.8 billion resulted primarily from
changes in assumptions relating to a substantial increase in workload,
reduced employee productivity, and increased advertising. Under the
nonsampling design, census costs will increase because the bureau expects
to follow up on more nonresponding households than for a sampling-based
census and plans to use additional programs to improve coverage because it
cannot rely on statistical methods to adjust for undercounting and other
coverage errors. 

The bureau assumes an increased workload because the housing units that
the bureau expects to visit have increased from an estimated 30 million to
46 million. The 16 million increase includes visiting 12 million
additional nonresponding housing units and 4 million additional housing
units that the Postal Services says are vacant or nonexistent. Also
contributing to the workload increase are a number of programs that were
not in the original budget. These programs are primarily aimed at
improving the accuracy of the 2000 census through quality control
operations, such as reinterviewing households that had been previously
visited by an enumerator. However, it is unclear whether these additional
programs will result in a 2000 census that is more accurate than the 1990
census. This increased workload, which increased costs for most bureau
program activities, relates primarily to additional salaries, benefits,
travel, data processing, infrastructure, and supplies.

Another key factor substantially increasing the fiscal year 2000 budget
request is that the bureau reduced the assumed productivity of its
temporary enumerator employees. The bureau reduced the assumed
productivity of its enumerators by 20 percent between the original and
amended budget requests--from 1.28 to 1.03 households per hour. This
reduction relates to all the enumerator employees for nonresponse follow-
up--both those employees to be hired to visit the 30 million nonresponding
housing units in the original budget request as well as those for the 16
million additional housing units. The bureau did not provide any
documented internal or external quantitative analysis or other analysis to
support the initial or revised productivity rates. This reduction was
primarily based on senior management judgments, which the bureau
acknowledges are very conservative, about factors such as the uncertainty
of hiring a sufficient number of quality temporary workers in a tight
labor market. 

Because of the assumed increase in workload and reduction in productivity,
the total number of temporary field positions/Footnote3/ increased from
780,000 in the original budget to 1,350,000 in the amended budget request.
The 570,000 new positions include 200,000 (35 percent) for following up on
nonresponding households, 120,000 (21 percent) for enumerator activities
such as counting people at homeless shelters, and 220,000 (38 percent)
related to additional coverage improvement and quality control programs.
(The remaining 6 percent is for various other positions.) This large
number of positions is due in part to the expected substantial temporary
employee turnover.

The bureau also included nearly $72 million of advertising in the amended
budget request to increase public awareness and hopefully increase
response rates for mailed questionnaires. However, the bureau's amended
budget request did not assume any cost savings from the increased
advertising dollars in the form of increased response rate and,
accordingly, a reduced workload. 

According to the bureau, about $1.6 billion of this increase is related
and $100 million is not related to the inability to use statistical
sampling. As discussed above, the additional costs primarily involve
increased workload, reduced productivity, and increased advertising. The
items unrelated to the sampling issue include costs not included in the
original budget and revisions of prior estimates. For example, the bureau
did not include $52 million for rent and long distance telephone service
in the original budget request.

The bureau developed its $4.5 billion amended budget request for fiscal
year 2000 using a cost model consisting of a series of interrelated
software spreadsheets. The original and amended budget requests were
developed using this cost model, with each estimate being developed
independently using different versions of the cost model. The bureau
derived the
$1.7 billion requested increase by calculating the net difference between
the original budget request of $2.8 billion and the amended budget request
of $4.5 billion.

Background 

The President's fiscal year 2000 budget, which was submitted to the
Congress on February 1, 1999, included nearly $2.8 billion for the Census
Bureau to perform the 2000 decennial census. This original budget request
reflected the bureau's plan to gather information based in part on using
statistical estimation for nonresponding households and to adjust for
undercounting and other coverage errors. However, on January 25, 1999, the
Supreme Court ruled that the Census Act prohibits the bureau from using
sampling for purposes of congressional apportionment. According to the
bureau, it must therefore enumerate an estimated 12 million additional
nonresponding households and expand programs designed to address
undercounting and limit other coverage errors. The bureau also plans to
visit 4 million additional addresses for which the Postal Service has
returned the questionnaire because it believes that the housing unit is
vacant or nonexistent. 

In light of the need for the additional enumeration, the bureau requested
a substantial increase to its original budget request for fiscal year 2000-
-approximately $1.7 billion--for a total budget of approximately $4.5
billion. Thus, for purposes of apportionment, the 2000 census will be done
similarly to those for the last several decades in which questionnaires
are mailed to the majority of the nation's households asking the occupants
to mail back the completed questionnaires./Footnote4/ In addition, for the
2000 census, the bureau plans to incorporate statistical estimation for
other purposes, such as providing states with information for
redistricting. As part of the 2000 census, the bureau will allow the
public to respond in a variety of new ways, such as through a toll-free
telephone number, the Internet, and unaddressed questionnaires made
available at public locations. 

When compared to the 1970 census, the projected full-cycle cost per
housing unit of the 2000 census will quadruple (in constant fiscal year
1998 dollars). As shown in figure 1, the cost will nearly double from the
previous census (1990). 

Figure****Helvetica:x11****1:    Census Full-Cycle Cost per Housing Unit--
                                 Fiscal Years 1970 to 2000 (in constant
                                 fiscal year 1998 dollars) 

*****************

*****************

Note: Cost per housing unit is based on the full-cycle cost of performing
a decennial census. The 2000 decennial census cycle began the year after
the 1990 decennial census and is estimated to end in 2003. Therefore, the
cost per housing unit for the 2000 census is based on reported costs
incurred for fiscal years 1991 through 1998, the fiscal year 1999 total
enacted appropriations, the fiscal year 2000 amended budget request, and
the bureau's estimate of costs for fiscal years beyond 2000. The 2000
census estimate assumes 119 million total households. The Gross Domestic
Product (GDP) price index was used to adjust for inflation.

Source: GAO analysis of unaudited bureau data.

The accuracy of the 1990 census decreased compared to the 1980 census.
Specifically, the net undercount for the 1990 census was estimated at 1.6
percent of the population (about 4 million persons) compared to 1.2
percent for the 1980 census. The bureau estimated that about 4.4 million
persons were counted twice, and other improper inclusions, in 1990, while
8.4 million were missed. Moreover, the sum of these numbers--
12.8 million--represents a minimum tally of gross errors since it does not
include other errors, such as persons assigned to the wrong locations.

Scope and Methodology

We interviewed bureau officials who provided us with an overview of the
budget formulation process for developing the $4.5 billion amended budget
request. Bureau officials provided us with copies of the cost models used
to develop the original and amended budget requests. The $1.7 billion
requested increase represents the difference between the original and
amended budget requests. Therefore, to ensure that the difference between
the original and amended budget requests was supported by the two models,
we reviewed the summarized output and structure of each model and verified
the mathematical accuracy of the difference. We reconciled the output from
these models to the budget request amounts to determine whether they
mathematically agree with the original and amended budget requests. 

We then isolated the differences in the models by the bureau's eight
program activities or frameworks./Footnote5/ After isolating the
differences, we selected for further analysis cost increases/decreases and
related changes in assumptions that contributed most to each framework's
total net increase or otherwise warranted explanation. We focused our
analysis on the "field data collection and support systems" program
activity because nearly $1.5 billion of the $1.7 billion requested
increase was related to this framework.

We obtained from the bureau documentation supporting or explaining the
basis for the change in assumptions and how the increase/decrease related
to the inability to use statistical sampling. The bureau categorized the
components of the $1.7 billion requested increase into those it considered
related to the statistical sampling issue and those it considered to be
unrelated. We did not conclude whether the bureau properly categorized the
components of the $1.7 billion requested increase as related and
unrelated. We did not assess the efficiency or effectiveness of the
bureau's plans to conduct the 2000 census or validate the bureau's data
and assumptions used to develop the budget requests. 

We performed our work in Washington, D.C., and at the bureau's
headquarters in Suitland, Maryland, from June through September 1999 in
accordance with generally accepted government auditing standards.

Overall Observations on the Fiscal Year 2000 $1.7 Billion Requested Increase

The bureau's request for an additional $1.7 billion for fiscal year 2000
primarily involves changes in assumptions related to increased workload,
reduced employee productivity, and increased advertising. The following
sections discuss the changes in these key assumptions between the original
and amended budget requests.

Key Assumption--Increased Workload 
-----------------------------------

An assumed increase in workload substantially increased costs in the
fiscal year 2000 amended budget request for enumerator and support
salaries, benefits, travel, data capture requirements, infrastructure, and
supplies. The primary reason for the increase in workload is the bureau's
plan to visit an additional 16 million households using a traditional
census approach instead of sampling to estimate these housing units. Also
adding to the workload are additional programs intended to improve the
coverage and quality of the census.

The bureau's amended budget request assumes that it will enumerate the
12 million nonresponding housing units that under the original budget
request would have been statistically estimated. The bureau also estimated
that the Postal Service would determine that of the 119 million total
estimated housing units, about 6 million would be vacant or nonexistent.
The purpose of visiting these 6 million housing units is to confirm that
the Postal Service was correct in concluding that a housing unit was
vacant or nonexistent. In 1990, the bureau found that about 20 to 30
percent of the housing units identified as vacant or nonexistent were
actually occupied. As shown in figure 2, under the original budget
request, the bureau assumed that it would visit only 2 million of the 6
million housing units and that the remaining 4 million housing units would
be accounted for through statistical estimation. Under the amended budget
request, the bureau assumes that it will visit all 6 million housing units.

Figure****Helvetica:x11****2:    Increase in Housing Units Visited Between
                                 the Original and Amended Budget Requests

*****************

*****************

Source: GAO analysis of unaudited bureau data.

Also contributing to the workload increase are a number of programs
intended to improve the coverage and quality of the census that were not
in the original budget request. According to the bureau, these programs
were added to improve the accuracy of the 2000 census since the bureau
would be using a traditional approach and attempting to obtain information
directly from all 119 million estimated households. In the bureau's
original budget request, statistical estimation was intended to improve
the accuracy of the 2000 census by adjusting census counts for
undercounting and other coverage errors. 

An example of an additional program includes reinterviewing a sample of
housing units for which enumerators had previously completed census
questionnaires. This program, which is budgeted to cost $22.7 million,
involves sampling housing units with questionnaires completed by
enumerators, reinterviewing to confirm responses, and adjusting the
responses to the census questionnaires to correct any errors found.
Another program, budgeted at $25.2 million, involves attempting to
redeliver certain census questionnaires that are returned by the Postal
Service. This program is targeted at housing units that have a change in
address or zip code between the fall of 1999, when the bureau delivers its
address files to its printing vendors, and March 2000, when questionnaires
are to be delivered. According to the bureau, the improved coverage and
accuracy resulting from these programs would have been accounted for in
the original plan through statistical estimation techniques. Although the
programs the bureau added in the amended budget request are intended to
improve the accuracy of the 2000 census, it is unclear whether these
additional programs will result in a 2000 census that is more accurate
than the 1990 census.

It is important to note that another key assumption--the average mail
response rate for all questionnaire types--remained constant between the
original and amended budget requests at 61 percent. Given an estimated 119
million households in the census population, a 1-percent change in the
response rate increases or decreases the number of households that must be
enumerated by about 1.2 million. The mail response rate has decreased
substantially in recent years, dropping from 78 percent in 1970 to 65
percent in 1990. If the response rate is substantially different from the
61 percent that the bureau projects, assuming that all other assumptions
prove to be accurate, the bureau's fiscal year 2000 amended budget request
could differ substantially from the bureau's needs. 

Key Assumption--Reduced Productivity
------------------------------------

A significant factor increasing the bureau's budget request is a 20-
percent reduction in the assumed productivity rate for temporary employee
enumerators following up on nonresponding housing units. In the original
budget request, the bureau used an average productivity rate of about 1.28
households per hour, which was reduced in the amended budget request to
about 1.03. This revised productivity rate was applied to all temporary
enumerator employees for nonresponse follow-up--both those to be hired to
visit the 30 million housing units under the original budget request as
well as those for the additional 16 million housing units added after the
use of sampling for nonresponse follow-up was prohibited by the Census Act.

The bureau provided several explanations for reducing the assumed
enumerator productivity rate from 1.28 in the original budget request to
1.03 in the amended budget request. For example, the bureau believes that
given the low unemployment rate in the United States, hiring temporary
employees to fill hundreds of thousands of new positions is expected to
result in the hiring of less productive workers. In addition, the bureau
believes enumerators will work slower in order to avoid mistakes knowing
that they are being reviewed in a more thorough and formal manner as part
of the additional quality control programs. Also, the bureau assumed an
increase in rework resulting from added quality control programs.

The bureau did not provide any documented internal or external
quantitative analysis or other analysis that supported the original or the
revised productivity rate. Consequently, the 20-percent reduction in
productivity is based on senior management judgments, which the bureau
acknowledges are very conservative.

Because of the increased workload and reduced productivity, the bureau
increased the total number of temporary field positions from the 780,000
estimated in the original budget to 1,350,000. This increase in positions
does not necessarily mean that the bureau is hiring 570,000 additional
people. Many positions exist for only a few weeks; thus some individuals
can be, and are, hired for more than one position. Of the 570,000
increase, about 200,000 positions (35 percent) are for nonresponse follow-
up and 120,000 (21 percent) are for enumeration-related activities such as
counting people at homeless shelters; 220,000 positions (38 percent) are
for added coverage improvement and quality control programs. (The
remaining 6 percent includes various other positions.)

Key Assumption--Additional Advertising
--------------------------------------

The bureau included nearly $72 million for advertising intended to
increase questionnaire responses, including advertising that will be
targeted at hard-to-enumerate communities. This additional advertising was
primarily intended to educate the public on the 2000 census and to
increase the mail response rate for questionnaires and thus reduce the
bureau's workload.

The bureau's original budget request included about $56 million for a paid
advertising program in fiscal year 2000. The bureau intends to use the
additional $72 million for motivational programs intended to increase the
mail-in response rate and for an expanded "Census in the Schools" program.
An example of how the bureau plans to spend this money includes a planned
educational message, delivered several weeks before the data collection
period, reminding residents in the hard-to-enumerate communities about the
benefits of participating in the census process. Also, advertising is to
be targeted to run concurrent with the follow-up visits to nonresponding
households so that enumerators have a better chance of successfully
gaining access to households and completing census forms.

The bureau has no data available to support how much, if any, the
increased advertising will increase the response rate. As a result, the
bureau's assumed average questionnaire response rate of 61 percent in the
original budget request did not increase in the amended budget request.
Thus, the bureau has assumed no cost savings in the form of increased
response rate and resultant reduced workload from the increased
advertising dollars.

Components of the $1.7 Billion Increase Related and Unrelated to the
Sampling Issue

According to the bureau, about $1.6 billion of the $1.7 billion requested
increase is related to the inability to use statistical sampling, and the
remaining $104 million is not related. Unrelated items include costs not
included in the original budget request and revisions of prior estimates.
Examples of items unrelated to the decision include the following:

o   $29 million for leasing of common space,

o   $23 million for long-distance telephone service,

o   $16 million for improving address lists and delivering
  questionnaires, and

o   $10 million for copier paper and map supplies.

We did not conclude whether the bureau properly categorized components of
the $1.7 billion requested increase as related and unrelated. The briefing
slides in appendix I provide a detailed discussion by program activity (or
framework) of the components of the $1.7 billion requested increase.

Process Used to Develop the Fiscal Year 2000 Amended Budget Request

To develop the amended fiscal year 2000 budget request, the bureau used a
model that consisted of an extensive set of interrelated software
spreadsheets. Both the original and amended budget requests were developed
with this cost model, with each estimate being developed independently
using different versions of the cost model. The process the bureau used to
develop the amended budget request involved (1) revising key assumptions
in the cost model supporting the original budget request that were based
on the statistical estimation approach for nonresponse follow-up to
improve accuracy and (2) incorporating new assumptions based on a
traditional census approach. As shown in table 1, the bureau derived the
$1.7 billion requested increase by calculating the net difference between
the original budget request of $2.8 billion and the amended budget request
of $4.5 billion. The output from the bureau's models mathematically agrees
with the original and amended budget requests.

Table****Helvetica:x11****1:    Comparison of the Bureau's Original and
                                Amended Budget Request 

-------------------------------------------------------------------------
| Dollars in millions                :         :           :            |
|-----------------------------------------------------------------------|
| Program activity (or framework)    : Original:   Amended : Requested  |
|                                    :  budget :    budget :  increase  |
|                                    : request :   request :            |
|-----------------------------------------------------------------------|
| 1. Program development and         :     $28 :       $31 :        $3  |
| management                         :         :           :            |
|-----------------------------------------------------------------------|
| 2. Data content and products       :     195 :       195 :         0  |
|-----------------------------------------------------------------------|
| 3. Field data collection and       :   2,023 :     3,474 :     1,451  |
| support systems                    :         :           :            |
|-----------------------------------------------------------------------|
| 4. Address list development        :      34 :        44 :        10  |
|-----------------------------------------------------------------------|
| 5. Automated data processing and   :     341 :       477 :       136  |
|     telecommunications support     :         :           :            |
|-----------------------------------------------------------------------|
| 6. Testing, evaluation, and dress  :      21 :        21 :         0  |
| rehearsal                          :         :           :            |
|-----------------------------------------------------------------------|
| 7. Puerto Rico, Virgin Islands,    :      37 :        71 :        34  |
| and Pacific Areas                  :         :           :            |
|-----------------------------------------------------------------------|
| 8. Marketing, communications, and  :     111 :       200 :        89  |
| partnerships                       :         :           :            |
|-----------------------------------------------------------------------|
| Total requested amount             :  $2,790 :    $4,513 :    $1,723  |
-------------------------------------------------------------------------

Source: GAO analysis of unaudited bureau data.

The bureau's budget estimation model is a set of 16 spreadsheets with
thousands of separate formulas that generate estimates for specific census
program activities. The formulas generate outputs for items such as
salaries and benefits, travel, and advertising based on hundreds of
different assumptions. The model serves as a collection point for field
costs as well as headquarters and other costs estimated outside the model.
Of the
$4.5 billion amended budget request, about $1.05 billion (23 percent) was
calculated outside the model. This $1.05 billion includes costs for
headquarters activities and contracts. The assumptions are developed by
program managers and are generally based on either third party evidence,
such as independent studies, or senior management's judgment.

Agency Comments and Our Evaluation

We provided a draft of this report to the Department of Commerce for
comment. As requested, the Director of the Bureau of the Census provided
written comments on behalf of the department in 2 days. (See appendix II.)
We appreciate the bureau's rapid response to the draft and its overall
cooperation and timely responses to our data requests. In commenting on a
draft of this report, the bureau concurred with the facts as presented in
the report and provided its perspective on four matters, which we address
below.

First, the bureau stated that in addition to serving primarily as the
provider of apportionment counts, the 2000 census has three other major
components of unique usefulness to the American public-providing data to
be used for legislative redistricting, federal housing surveys, and
distribution of federal funds. We agree with the bureau that these other
components of the 2000 census are important. However, the purpose of our
report was to analyze the bureau's overall amended budget request with a
focus on the $1.7 billion requested increase. Given that objective, our
report focused on activities that resulted in the substantial increase in
the bureau's original fiscal year 2000 budget request. As our analysis
shows, the net $1.7 billion requested increase was primarily related to
the bureau's assumptions of increased workload, reduced employee
productivity, and increased advertising. We did not see any substantial
cost increases related to redistricting, federal housing surveys, and
distribution of federal funds.

Second, the bureau noted that the current plan for the 2000 census
contains components of both a traditional and a sampling methodology. In
addition, the bureau noted that beginning in November 1997, it devoted a
substantial effort in planning (on a dual track basis) for a census based
entirely on a traditional methodology. Based on our review of the bureau's
fiscal year 2000 amended budget request, we agree that the 2000 census
includes components of both traditional and sampling methodologies.
However, the increase in the estimated cost of the 2000 census is driven
primarily by factors relating to a traditional census. The $1.7 billion
requested increase in the bureau's fiscal year 2000 budget is net of a
decrease of over
$200 million related to the Accuracy and Coverage Evaluation (ACE)
program. The purpose of ACE in the amended budget request is primarily to
estimate the population for purposes such as redistricting by sampling
households from an address list developed independently from the list used
to perform the census. Under the bureau's original budget request, the
results of ACE (formerly Integrated Coverage Measurement) were to have
been statistically combined with the results of enumeration to provide a
single, integrated set of population counts. Evaluating the bureau's
planning for a "dual track" census was beyond the scope of our work. 

Third, the bureau provided its perspective on the reasons why the full-
cycle cost per housing unit of the 2000 census is projected to nearly
double when compared to the 1990 census. The bureau pointed to such
factors as normal inflationary increases, infrastructure, federal wage
increases, declining response rate, paid advertising, postage, and
information technology. It is important to note that our analysis
eliminates the impact of general inflation and the growth in housing units
over the last 4 decades. Thus, our comparison of full-cycle cost per
housing unit is an "apples to apples" comparison. The purpose of showing
this comparison is to provide a historical perspective on the real
increase in the full-cycle cost per housing unit of the decennial census.
Providing an analysis of the reasons for the substantial increase of the
projected cost of the 2000 census was beyond the scope of our review. We
agree with the bureau that factors such as a declining mail response rate
would increase the relative costs of the census. However, it is important
to point out that the bureau was unable to demonstrate whether the
benefits of certain activities, such as increased advertising and
additional programs, justify their cost. In addition, the bureau could not
say whether a 2000 census costing an estimated $56 per housing unit (in
constant 1998 dollars) will result in a census that is more accurate than
1990.

Finally, the bureau stated that it believes it is prudent to assume a
temporary enumerator employee productivity rate of 1.03 housing units per
hour. The bureau cited the immovable calendar and the need to hire a large
temporary labor force, the size of which is a direct result of assumed
lower productivity. The bureau believes our use of "very conservative"
leaves an unwarranted impression of management choices made at one end of
a spectrum, which could be shifted to the other end of the spectrum
without risk. Based on our analysis, we continue to conclude that the
bureau's productivity assumption of 1.03 is based on senior management's
"very conservative" judgment as was represented to us by bureau staff
during our review. What we mean by "very conservative" is that the bureau
is planning for a worst case scenario. As we reported, the bureau did not
provide any documented internal or external quantitative analysis or other
analysis to support the initial or revised productivity rates. As such,
our only possible conclusion is that the 1.03 was developed based on
management's judgment, which is what bureau staff represented to us during
our review. Because the reported productivity rate from the 1990 census
was 1.56 and the assumption used in the original budget was 1.28, we
conclude that the 1.03 assumption used in the amended budget request,
which covers not only the additional 16 million housing units now planned
to be enumerated but also the 30 million housing units to be enumerated in
the original budget, is very conservative--or a "worst case scenario."
Should the bureau's productivity assumption prove to be more or less than
1.03 housing units per hour, the bureau's fiscal year 2000 amended budget
request could differ substantially from the bureau's needs.

We are sending copies of this report to Senator Fred Thompson, Senator
Joseph Lieberman, Senator Judd Gregg, Senator Ernest Hollings,
Representative Carolyn Maloney, Representative Dan Burton, Representative
Henry Waxman, Representative Harold Rogers, and Representative
Jos****ITCCentury Book:x8e**** Serrano in their capacities as Chairs or
Ranking Minority Members of Senate and House Committees and Subcommittees.
We are also sending copies to the Honorable William M. Daley, Secretary of
Commerce; the Honorable Kenneth Prewitt, Director of the Bureau of the
Census; the Honorable Jacob J. Lew, Director of the Office of Management
and Budget; and other interested parties. Copies will also be made
available to others upon request.

If you have any questions on matters discussed in this letter, please
contact me at (202) 512-3406. Other contacts and key contributors to this
report are listed in appendix III.

Sincerely yours,

*****************

*****************

Gregory D. Kutz
Associate Director
Governmentwide Accounting and Financial
Management Issues

--------------------------------------
/Footnote1/-^The bureau refers to this activity as "nonresponse follow-up."
/Footnote2/-^13 U.S.C. 195.
/Footnote3/-^The number of positions does not equal the number of staff
  required. Some positions last only a few weeks, and one person can fill
  more than one position.
/Footnote4/-^Referred to as a traditional census.
/Footnote5/-^The bureau's budget includes eight program activities, which
  are also referred to by the bureau as "frameworks."

BRIEFING ON THE ANALYSIS OF THE CENSUS BUREAU'S FISCAL YEAR 2000 AMENDED
BUDGET REQUEST
===========================================================================

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

COMMENTS FROM THE DEPARTMENT OF COMMERCE
========================================

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

*****************

GAO CONTACT AND STAFF ACKNOWLEDGEMENTS
======================================

GAO Contact

Scott McNulty, (202) 512-9184

Acknowledgements

In addition to the contact named above, the following individuals made key
contributions to this report: Cindy Brown-Barnes, Joan Hawkins, Theresa
Patrizio, Mike Vu, and Sophia Harrison.

(901793)

*** End of document. ***