2010 Census: Basic Design Has Potential, but Remaining Challenges
Need Prompt Resolution (12-JAN-05, GAO-05-9).			 
                                                                 
A rigorous testing and evaluation program is a critical component
of the census planning process because it helps the U.S. Census  
Bureau (Bureau) assess activities that show promise for a more	 
cost-effective head count. The Bureau conducted a field test in  
2004, and we were asked to (1) assess the soundness of the test  
design and the extent to which the Bureau implemented it	 
consistent with its plans, (2) review the quality of the Bureau's
information technology (IT) security practices, and (3) identify 
initial lessons learned from conducting the test and their	 
implications for future tests and the 2010 Census.		 
-------------------------Indexing Terms------------------------- 
REPORTNUM:   GAO-05-9						        
    ACCNO:   A15656						        
  TITLE:     2010 Census: Basic Design Has Potential, but Remaining   
Challenges Need Prompt Resolution				 
     DATE:   01/12/2005 
  SUBJECT:   Census						 
	     Computer security					 
	     Computers						 
	     Cost control					 
	     Cost effectiveness analysis			 
	     Data collection					 
	     Data integrity					 
	     Information technology				 
	     Operational testing				 
	     Performance measures				 
	     Population statistics				 
	     Statistical data					 
	     2000 Decennial Census				 
	     2010 Decennial Census				 

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO 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.                                         **
**                                                              **
******************************************************************
GAO-05-9

                 United States Government Accountability Office

                     GAO Report to Congressional Requesters

January 2005

                                  2010 CENSUS

Basic Design Has Potential, but Remaining Challenges Need Prompt
Resolution

                                       a

GAO-05-9

[IMG]

January 2005

2010 CENSUS

Basic Design Has Potential, but Remaining Challenges Need Prompt
Resolution

  What GAO Found

The Bureau's design for the 2004 census test addressed important
components of a sound study, and the Bureau generally implemented the test
as planned. For example, the Bureau clearly identified its research
objectives, developed research questions that supported those objectives,
and developed evaluation plans for each of the test's 11 research
questions.

The initial results of the test suggest that while certain new procedures
show promise for improving the cost-effectiveness of the census, the
Bureau will have to first address a number of problems that could
jeopardize a successful head count. For example, enumerators had little
trouble using hand held computers (HHC) to collect household data and
remove late mail returns. The computers could reduce the Bureau's reliance
on paper questionnaires and maps and thus save money. The test results
also suggest that certain refinements the Bureau made to its procedures
for counting dormitories, nursing homes, and other "group quarters" could
help prevent the miscounting of this population group.

The 2004 Census Test Was Conducted in Rural Georgia and Queens, New York

The Bureau conducted a 2004 census test in Queens, New York and three
counties in rural Georgia. The test focused on using HHCs for conducting
nonresponse follow-up interviews-where census workers go door-to-door to
count those households that did not mail back their census forms. The
Bureau also tested new methods for improving coverage, redesigned race and
ethnicity questions, and improved methods for defining and identifying
group quarters. The Bureau established these test objectives to help meet
its goals for improving accuracy, reducing risk, and containing the cost
of the 2010 Census.

Source: U.S. Census Bureau.

Other aspects of the test did not go as smoothly. For example, security
practices for the Bureau's IT systems had weaknesses; the HHCs had
problems transmitting data; questionnaire items designed to improve
coverage and better capture race/ethnicity confused respondents;
enumerators sometimes deviated from prescribed enumeration procedures; and
certain features of the test were not fully operational at the time of the
test, which hampered the Bureau from fully gauging their performance. With
few testing opportunities remaining, it will be important for (1) the
Bureau to find the source of these problems, devise cost-effective
solutions, and integrate refinements before the next field test scheduled
for 2006, and (2) Congress to monitor the Bureau's progress in resolving
these issues.

                 United States Government Accountability Office

Contents

    Letter                                                                  1 
                                         Results in Brief                   2 
                                            Background                      4 
                                      Scope and Methodology                 6 
                             The Census Test Was Generally Sound, but      
                                        Refinements Could                  
                             Produce Better Cost and Performance Data       8 
                         The Bureau Needs to Implement Better IT Security  11 
                                            Practices                      
                           Test Reveals Technical, Training, and Other     
                                      Challenges in Need of                
                                        Prompt Resolution                  13 
                                           Conclusions                     29 
                               Recommendations for Executive Action        30 
                                Agency Comments and Our Evaluation         31 
Appendix                                                                
            Appendix I:      Comments from the Department of Commerce      
                          Table 1: Design for 2004 Census Test Addressed   
    Table                                   Important                      
                                   Components of a Sound Study             

Figures Figure 1:

Figure 2: Figure 3:

Figure 4:

Figure 5:

Figure 6:

Figure 7:

HHC Being Tested for Use in Collecting Data in the
Field 5
Maps of Test Sites in Georgia and New York 7
An Enumerator Using an HHC for Nonresponse
Follow-up 14
Data Transmission Process for Nonresponse
Follow-up 16
New Coverage Questions Were Designed to Ensure a
Complete Count 20
Race and Ethnicity Categories on the HHCs Were
Formatted Differently From the Paper Questionnaires 22
Group Homes Could Resemble Conventional Houses 25

This is a work of the U.S. government and is not subject to copyright
protection in the United States. It may be reproduced and distributed in
its entirety without further permission from GAO. However, because this
work may contain copyrighted images or other material, permission from the
copyright holder may be necessary if you wish to reproduce this material
separately.

A

United States Government Accountability Office Washington, D.C. 20548

January 12, 2005

The Honorable Tom Davis Chairman The Honorable Henry A. Waxman Ranking
Minority Member Committee on Government Reform House of Representatives

The Honorable Adam H. Putnam House of Representatives

The consequences of a poorly planned census are high given the billions of
dollars spent on the enumeration and the importance of collecting quality
data. Therefore, a rigorous testing and evaluation program is a critical
component of the census planning process because it helps the U.S. Census
Bureau (Bureau) assess activities and related information technology (IT)
systems that show promise for a more cost-effective head count. In
preparing for the 2010 Census, the Bureau conducted a field test in 2004
and plans additional tests for 2005 and 2006. It is important that these
early assessments lead to a design that is sufficiently mature so that the
dress rehearsal for the 2010 Census, now planned for 2008, will
demonstrate the feasibility of the various operations and technologies
planned for the decennial under conditions that are as close as possible
to the actual census.

The Bureau designed the 2004 census test to examine the feasibility of
using (1) handheld computers (HHC) for field data collection; (2) new
methods for improving coverage; (3) redesigned race and ethnicity
(Hispanic origin) questions; and (4) improved methods for defining and
identifying nursing homes, prisons, college dormitories, and similar
facilities known collectively as "group quarters." The Bureau established
these test objectives as part of a broader effort to modernize and
re	engineer the 2010 Census. Major goals of this initiative are to improve
the accuracy, reduce the risk, and contain the cost of the 2010 Census,
estimates of which now exceed $11 billion. A rigorous planning and testing
program is critical to this effort.

As agreed with your offices, our objectives for this review were to assess
the soundness of the design of the 2004 census test and the extent to
which the Bureau implemented the test consistent with its plans. We also
agreed to review the quality of its IT security practices, and identify
initial lessons

learned from the test and the implications they have for the Bureau's
future plans.

To address these objectives, we reviewed applicable planning, IT, and
other documents, and interviewed knowledgeable Bureau officials
responsible for key operations and computer security. We also made several
visits to the two test sites-an urban location in the northwestern portion
of Queens Borough, New York, and a rural location in south central
Georgia. We conducted our work from November 2003 through November 2004 in
accordance with generally accepted government auditing standards.

Results in Brief The design of the 2004 census test addressed important
components of a sound study, and the Bureau generally implemented the test
as planned. For example, the Bureau clearly identified its test
objectives, developed research questions that supported those objectives,
and developed evaluation plans for each of the test's 11 research
questions.

Still, there are opportunities to improve both the utility of data from
the current test as well as the design of the next field test in 2006.
Combined, these improvements could help inform future budget estimates and
investment and design decisions. For example, the 2004 test could benefit
by analyzing the impact the HHCs and targeted second mailing-an operation
designed to increase the response rate-had on cost savings and
productivity. Similarly, the 2006 test could be improved if the Bureau
developed quantifiable productivity and other performance requirements for
the HHCs and then used the 2006 test to determine whether the devices are
capable of meeting those requirements.

The 2004 test was an important milestone in the 2010 Census life cycle
because it shed light on those operations that show potential for
improving the cost-effectiveness of the decennial head count, as well as
problem areas that could jeopardize the success of the census if not
resolved. For example, the initial test results showed that the HHCs were
effective for conducting interviews and removing late mail returns.
Indeed, most enumerators we observed had little trouble using the
computers for conducting interviews, and they were generally pleased with
the HHC's overall functionality, durability, and screen clarity. Likewise,
the HHCs enabled the Bureau to remove over 7,000 late mail returns from
enumerators' workloads at both test sites, which could help the Bureau
save money by eliminating the need to visit those households that already
mailed back their census questionnaires. The test results also suggest
that

certain refinements the Bureau made to its procedures for counting group
quarters-namely integrating its housing unit and group quarters address
lists-could help prevent the miscounting of this population group.

Other aspects of the test did not go as smoothly and point to areas on
which the Bureau should focus as it looks toward the future. For example:

o  The Bureau's IT security practices had weaknesses;

o  Technical and training difficulties caused HHC transmission problems;

o  The HHC's mapping function was slow to load and was thus little used;

o Questionnaire items designed to improve coverage and better determine
race/ethnicity were awkward for census workers to ask and confusing for
respondents, which could affect data quality;

o Census workers sometimes deviated from prescribed enumeration
procedures, which could impair the reliability of the data;

o Enumerators had difficulties finding the physical locations of specific
group quarters; and

o Certain features of the test were not fully operational at the time of
the test, which hampered the Bureau from gauging their true performance.

The 2006 field test is the Bureau's last opportunity to assess its basic
design for the census before conducting a dress rehearsal in 2008. At that
time, the Bureau plans to demonstrate the entire design under conditions
that mirror the census. Any changes to the design made after the dress
rehearsal could entail considerable risk as they would not be properly
tested. Thus, it will be important for the Bureau to exhaustively assess
the results of the 2004 test, diagnose and remedy any shortcomings, and
thoroughly road test refinements in 2006.

To facilitate this, we recommend that the Secretary of Commerce direct the
Bureau to address the technical, methodological, training, and procedural
shortcomings revealed during the 2004 test. Specific actions include
enhancing the Bureaus' IT security practices, improving the reliability of
HHC transmissions, and taking a more strategic approach to training
enumerators. The Bureau should also regularly update Congress on the
progress it is making in addressing these and any other challenges, as
well

as the extent to which it is on track for meeting the overall goals of the
2010 Census.

The Under Secretary for Economic Affairs forwarded written comments from
the Bureau on a draft of this report (see app. I). The Bureau generally
agreed with seven of our nine recommendations-those dealing with improving
IT security practices, the reliability of the HHCs, training, testing, and
enumeration procedures-and reported it was already taking a number of
steps to address our concerns. However, the Bureau took exception to our
recommendations to (1) analyze the impact that HHCs and a targeted second
mailing had on cost savings and other Bureau objectives, and (2) define
specific, measurable performance requirements for the HHCs and other
census-taking activities. Because the HHCs and certain other operations
are critical for containing costs and achieving other Bureau goals for the
2010 Census, it will be essential for the Bureau to gauge their impact and
determine whether they can meet the Bureau's demanding requirements. As a
result, we believe these recommendations still apply.

BackgroundCongress, GAO, the Department of Commerce Inspector General, and
even the Bureau itself have all noted that the 2000 Census was marked by
poor planning, which unnecessarily added to the cost, risk, and
controversy of the national head count. In January 2003, we named the 2010
Census a major performance and accountability challenge because of our
growing concern over the numerous obstacles to a cost-effective
enumeration as well as its escalating price tag.1 More recently, we
reported that while the Bureau's preparations for the 2010 Census appeared
to be further along than at a similar point during the planning cycle for
the 2000 Census, considerable risks and uncertainties remained.2 Thus, it
is imperative that the Bureau adequately test the various components of
its design for the 2010 Census.

A rigorous testing program provides at least four major benefits. First,
testing allows the Bureau to refine procedures aimed at addressing

1 GAO, Major Management Challenges and Program Risks: Department of
Commerce, GAO-03-97 (Washington, D.C.: January 2003).

2 GAO, 2010 Census: Cost and Design Issues Need to Be Addressed Soon,
GAO-04-37 (Washington, D.C.: Jan. 15, 2004).

problems encountered in past censuses. During the 2000 Census, for
example, group quarters were sometimes counted more than once or counted
in the wrong location; the wording of the race and ethnicity question
confused some respondents, which in some cases resulted in lower quality
data; and following up with nonrespondents proved to be costly and
labor-intensive. A second benefit is that sound testing can assess the
feasibility of new procedures and technologies, such as HHCs (see fig. 1),
that have never before been used in a decennial census.

Figure 1: HHC Being Tested for Use in Collecting Data in the Field

Source: U.S. Census Bureau.

Third, a rigorous testing program helps instill a comfort level among
members of Congress and other stakeholders that the Bureau (1) has chosen
the optimal design given various trade-offs and constraints and (2) has
identified and addressed potential risks and will be able to successfully
execute its plan. Such confidence building, developed through regular
updates and open lines of communication, is essential for continuing
congressional support and funding.

And finally, proper testing early in the decade will help the Bureau to
conduct a dress rehearsal in 2008 that fully assesses all aspects of the
census design under realistic conditions. Because of various late
requirement changes, certain procedures that were added after the 1998
dress rehearsal for the 2000 Census were not properly tested.

Scope and Methodology

As agreed with your offices, our objectives for this report were to (1)
assess the soundness of the Bureau's design for the 2004 censustest and
whether the Bureau implemented the test consistent with its plans, (2)
review the quality of the Bureau's IT security practices, and (3) identify
initial lessons learned from conducting the test and their implications
for the 2010 Census.

To assess the soundness of the design we reviewed pertinent documents that
described the Bureau's test and evaluation plans. We systematically rated
the Bureau's approach using a checklist of design elements that, based on
our review of program evaluation literature, are relevant to a sound study
plan. For example, we reviewed the Bureau's approach to determine, among
other things, (1) how clearly the Bureau presented research objectives,
(2) whether research questions matched the research objectives, and (3)
the appropriateness of the data collection strategy for reaching the
intended sample population. As part of our assessment of the Bureau's test
design, we also reviewed evaluations of the prior decennial census to
determine the degree to which the new operations being tested addressed
problematic aspects of the 2000 Census. However, we did not assess the
Bureau's criteria in selecting its objectives for the 2004 census test.

To determine if the Bureau implemented the test consistent with its plans,
we made multiple site visits to local census offices in Thomasville,
Georgia; and Queens Borough, New York. During these visits, we interviewed
local census office mangers and staff, observed various data collection
activities, and attended weeklong enumerator training. We observed a total
of 20 enumerators as they completed their daily nonresponse follow-up
assignments-half of these were in southern Georgia, in the counties of
Thomas, Colquitt, and Tift, and half were in Queens (see fig. 2 for maps
of the test site areas). The results of these observations are not
necessarily representative of the larger universe of enumerators.

              Figure 2: Maps of Test Sites in Georgia and New York

                                                      NY Northwest Queens, NY 
                                   GA Colquitt, Thomas, and Tift Counties, GA 

Source: GAO and U.S. Census Bureau.

To evaluate the quality of the Bureau's IT security practices, we assessed
risk management documentation associated with IT systems and major
applications for the 2004 census test. We based our determination on
applicable legal requirements, Bureau policy, and leading practices
described in our executive guide for information security management.3

3GAO, Executive Guide: Information Security Management-Learning from
Leading Organizations, GAO/AIMD-98-68 (Washington, D.C.: May 1998).

We also interviewed key Bureau officials associated with computer
security.

To identify lessons learned from the 2004 census test, we met with
officials from the Bureau's Decennial Management Division regarding
overall test plans and with officials from its Technologies Management
Office about using HHCs. Bureau officials and census workers from both
test locations also provided suggestions on improving census operations.

We requested comments on a draft of this report from the Secretary of
Commerce. On December 20, 2004, the Under Secretary for Economic Affairs,
Department of Commerce, forwarded written comments from the Bureau (see
app. I). We address these comments in the "Agency Comments and Our
Evaluation" section at the end of this report.

The Census Test Was Generally Sound, but Refinements Could Produce Better
Cost and Performance Data

The Bureau designed a sound census test and generally implemented it as
planned. However, in looking ahead, the Bureau's planning and investment
decisions could benefit from analyzing (1) the degree to which HHCs
contributed to the Bureau's cost containment goal and (2) the results of
the targeted second mailing, an operation designed to increase
participation by sending a follow-up questionnaire to nonresponding
households. Future tests could also be more informative if the Bureau
developed quantifiable productivity and other performance requirements for
the HHCs and then used the 2006 test to determine whether the devices are
capable of meeting those requirements. Collectively, these refinements
could provide Bureau officials with better information to guide its IT and
other design decisions, as well as refine future census tests.

The Bureau Developed a The design of the 2004 census test contained many
components of a sound

Sound Test Designstudy (see table 1). For example, the Bureau identified
test objectives, designed related research questions, and described a data
collection strategy appropriate for a field test. The Bureau also
developed evaluation plans for each of the test's 11 research questions,
and explained how stakeholders were involved with the design, as well as
how lessons learned from past studies were incorporated.

Table 1: Design for 2004 Census Test Addressed Important Components of a
Sound Study

Components of a
sound study Planned components of the 2004 census test

Clearly stated The objectives for the test concerned the feasibility of
using: objectives  o  HHCs for field data collection;

o  new methods for improving coverage;

o  redesigned race and ethnicity questions; and

o  improved methods for defining and identifying group quarters.

Research questions  o  Each of the 11 key research questions can be linked
to one of linked to objectives and the four objectives. rationale for site
    o  Two sites-in Queens and south central Georgia-were selection provided
selected based on test requirements.

Data collection strategy  o  Census Day was April 1, 2004, for the test.
thoroughly  o  Mode of data collection was the paper version of the short
form
documented to be mailed back by household.

o  Nonrespondent data were collected during personal interviews using
HHCs.

Input from stakeholders  o  Various research and development planning
groups were and lessons learned convened to develop 2004 test objectives.
considered in  o  The Bureau's Decennial Advisory Board and the National
developing test Academy of Science were informed of test plans. objectives
    o  Lessons learned from the 2000 Census were considered in

                        developing 2004 test objectives.

Design had data  o  Separate evaluations to answer key research questions
were analysis plan developed.

o  Evaluation plans recognized limitations. For example, the introduction
of a new data collection mode and new questions may make comparison to
2000 data difficult.

                Source: GAO analysis of U.S. Census Bureau data.

Additional Analysis Would Provide Better Data on the Impact of Key Test
Components

Although the Bureau plans to evaluate various aspects of the 2004 test, it
does not currently plan to assess the impact that the HHCs and targeted
second mailing had on cost savings and productivity. According to the
Bureau, the census test was focused more on determining the feasibility of
using the HHCs and less on the devices' ability to save money. Likewise,
the Bureau said it is not assessing the impact of the targeted second
mailing because the operation is not one of its four test objectives for
improving (1) field data collection using the HHC, (2) the coverage of
undercounted groups, (3) questions about race and ethnicity, and (4)
methods for defining special places and group quarters.

These decisions might be shortsighted, however, in that the Bureau
included the HHCs and targeted second mailing in the 2010 Census design,
in part, to reduce staff, improve productivity, and control costs. For
example, Bureau studies have shown that sending out replacement
questionnaires could yield a gain in overall response of 7 to 10 percent
from households that do not respond to the initial census mailing, and
thus generate significant cost savings by eliminating the need for census
workers to obtain those responses via personal visits. Thus, information
on the degree to which the HHCs and second mailing contribute to these key
goals could help inform future budget estimates, investment and design
decisions, as well as help refine future census tests.

Moreover, the feasibility of a targeted second mailing is an open
question, as the Bureau has never before included this operation as part
of a decennial census. Although a second mailing was part of the original
design for the 2000 Census, the Bureau had abandoned it because it was
found to be logistically unworkable. A Bureau official said that the
second mailing was included in the 2004 test only to facilitate the
enumeration process, and it would be better tested in a larger scale
operation such as the 2008 dress rehearsal. However, we believe that it
would be more prudent to assess the second mailing earlier in the census
cycle, such as, during the 2006 test so that its basic feasibility could
be assessed, any refinements could be evaluated in subsequent tests, and
the impact on savings could be estimated more accurately.

Future Tests Could Be Improved

While the design of the 2004 test was generally sound, refinements could
strengthen the next field test in 2006. Opportunities for improvement
exist in at least two areas: ensuring that (1) the HHCs can meet the
demanding requirements of field data collection and (2) management of the
local census offices mirrors an actual enumeration as much as possible.

With respect to the HHCs, because they replace the paper version of the
nonresponse follow-up questionnaire the devices mustfunction effectively.
Further, this test was the first time the Bureau used the HHCs under
census-like conditions so their functionality in an operational
environment was unknown. Bureau officials have acknowledged that for the
2004 test they had no predefined indicators of success or failure other
than if there was a complete breakdown the test would be halted. This is a
very low standard. Now that the Bureau has demonstrated the basic
functionality of the computers, it should next focus on determining the
specific performance requirements for the HHCs and assess whether the
devices

are capable of meeting them. For example, the Bureau needs productivity
benchmarks for the number of interviews per hour and per day that is
expected per census worker. Durability measures, such as how many devices
were repaired or replaced, should be considered as well. Assessing whether
the HHCs can meet the requirements of nonresponse follow-up will help
inform future design and investment decisions for whether or not to
include the devices in the 2010 design.

Ensuring that key positions in the local census offices are filled from
the same labor pool as they would be in an actual decennial census could
also enhance future census tests. Such was not the case during the 2004
test when, according to the Bureau, because of difficulties finding
qualified applicants, it used an experienced career census employee to
manage the overall day-to-day operations of the local census office at the
Queens test site. Another career employee occupied the office's regional
technician slot, whose responsibilities included providing technical and
administrative guidance to the local census office manager. In the actual
census, the Bureau would fill these and other positions with temporary
employees recruited from local labor markets. However, because the Bureau
staffed these positions with individuals already familiar with census
operations and who had ties to personnel at the Bureau's headquarters, the
Queens test may not have been realistic and the test results could be
somewhat skewed.

The Bureau Needs to Implement Better IT Security Practices

The Bureau operated a number of IT systems in order to transmit, manage,
and process data for the test. The equipment was located at various places
including the Bureau's headquarters in Suitland, Maryland; its National
Processing Center in Jeffersonville, Indiana; a computer facility in
Bowie, Maryland; as well as the New York and Georgia test sites.

Under Title 13 of the U.S. Code, the Bureau must protect from disclosure
the data it collects about individuals and establishments. Thus, the
Bureau's IT network must support both the test's telecommunications and
data processing requirements, as well as safeguard the confidentiality and
integrity of respondents' information.

The Federal Information Security Management Act of 2002 (FISMA) requires
each agency to develop, document, and implement an agency-wide information
security program for the IT systems that supports its operations.4
Although the Bureau took a number of steps to implement IT security over
the systems used for the test, based on available information, the Bureau
did not meet several of FISMA's key requirements. As a result, the Bureau
could not ensure that the systems supporting the test were properly
protected againstintrusion or unauthorized disclosureofsensitive
information. For example:

o IT inventory was not complete. FISMA requires an inventory of major
information systems and interfaces. The Bureau did not have a complete
inventory that showed all applications and general support IT systems
associated with the test. Without such information, the Bureau could not
ensure that security was effectively implemented for all of its systems
used in the test, including proper risk assessments, adequate security
plans, and effectively designed security controls.

o There was not sufficient evidence that the Bureau assessed all of the
devices used in the test for vulnerabilities, or that it corrected
previously identified problems. FISMA requires that agencies test and
evaluate the effectiveness of information security policies, procedures,
and practices for each system at least annually and that agencies have a
process for remediating any identified security weaknesses. Since the
Bureau could not provide us with a complete inventory of all network
components used in the test, we could not determine if the Bureau's tests
and evaluations were complete. Moreover, there was not always evidence
about whether the Bureau had corrected past problems or documented reasons
for not correcting them. As a result, the Bureau did not have adequate
assurance that the security of systems used in the 2004 census test was
adequately tested and evaluated or that identified weaknesses were
corrected on a timely basis.

o Assessments were not consistent. FISMA requires agencies to assess the
risks that could result from the unauthorized access, use, disclosure,
disruption, modification, or destruction of information or information
systems. Although the Bureau performed risk assessments

4 Federal Information Security Management Act of 2002, Title III,
E-Government Act of 2002, Pub. L. No. 107-347 (Dec. 17, 2002).

for some of the IT components used in the 2004 census test, the
documentation was not consistent. For example, documentation of
information sensitivity risks (high, medium, and low) for confidentiality,
integrity, and availability of information were not consistent and did not
always follow Bureau policy. In addition, documents showed different
numbers of file servers, firewalls, and even different names of devices.
Without complete and consistent risk assessment documentation, the Bureau
had limited assurance that it properly understood the security risks
associated with the test.

o The Bureau did not always follow its own risk policies. FISMA requires
the implementation of policies and procedures to prevent and/or mitigate
security risks. Although Bureau policies allowed for the waiver of
security policies, if appropriate, we noted that such policies were not
always followed. For example, a waiver for the test of certain password
policies was not properly documented and other system documents were not
properly updated to reflect the waiver. As a result, the risk assessment
for the 2004 census test did not properly identify the related risks and
did not identify any compensating controls to reduce the risk to an
acceptable level.

As the Bureau plans future tests and the census itself, it will be
important for it to strengthen its IT security risk management practices,
ensuring they fully adhere to FISMA requirements and its own IT security
policies.

Test Reveals Technical, Training, and Other Challenges in Need of Prompt
Resolution

The 2004 test suggests that while certain census initiatives have
potential, formidable challenges remain. For example, the HHCs show
promise in that enumerators were successful in using them to collect data
from nonrespondents and remove late mail returns. Still, they were not
street ready as they experienced transmission and memory overload
problems. Likewise, automated maps were difficult to use, certain
questionnaire items confused respondents, and enumerators did not always
follow interview protocols. These problems shed light on issues in need of
the Bureau's attention as it develops solutions and incorporates
refinements for additional testing in the years ahead.

HHCs Were Effective for The Bureau purchased 1,212 HHCs for the test at a
total cost of about $1.5 Conducting Interviews and million. The devices
were sent directly to the two test sites packaged in Removing Late Mail
Returns kits that included a battery, AC adaptor, and modem card for
transmitting

data via the telephone. The HHCs were also equipped with a Global
Positioning System (GPS), a satellite-based navigational system to help
enumerators locate street addresses. The Bureau anticipates the HHCs will
allow it to eliminate the millions of paper questionnaires and maps that
enumerators need when following up with nonrespondents, thereby improving
their efficiency and reducing overall costs.

Because the Bureau had never used HHCs in the decennial census, an
important goal of the test was to see whether enumerators could use them
for interviewing nonrespondents (see fig. 3). Most workers we observed had
little trouble using the device to complete the interviews. In fact, most
said they were pleased with the HHC's overall functionality, durability,
screen clarity, and the ability to toggle between the questionnaire and
taking GPS coordinates.

Figure 3: An Enumerator Using an HHC for Nonresponse Follow-up

Source: U.S. Census Bureau.

Another important function of the HHC was removing late mail returns from
each enumerator's assignment area(s). Between the Georgia and Queens test
sites, over 7,000 late mail returns were removed, reducing the total
nonresponse follow-up workload by nearly 6 percent.

The ability to remove late mail returns from the Bureau's nonresponse
follow-up workload could help save money in that it could eliminate the
need for enumerators to make expensive follow-up visits to households that
return their questionnaires after the mail-back deadline. Had the Bureau
possessed this capability during the 2000 Census, it could have eliminated
the need to visit nearly 773,000 late-responding households and saved an
estimated $22 million (based on our estimate that a 1-percentage point
increase in workload could add at least $34 million in direct salary,
benefits, and travel costs to the price tag of nonresponse follow-up5).
Because of the Bureau's experience in 2000, in our 2002 report on best
practices for more cost-effective nonresponse follow-up, we recommended,
and the Bureau agreed, that it should develop options that could purge
late mail returns from its nonresponse follow-up workload.6

Technical and Training Difficulties Caused HHC Transmission Problems

Each day, enumerators were to transmit completed nonresponse follow-up
cases to headquarters and receive assignments, software uploads, or both
via a telephone modem (see fig. 4 for a flowchart describing the file
transmission process). However, the majority of workers we interviewed had
problems doing so, in large part because of technical reasons or because
the Bureau's training did not adequately prepare them for the complexity
of the transmission procedure, which was a multistep process involving the
connection of a battery pack, cables, and other components. As reliable
transmissions are crucial to the success of nonresponse follow	up, it will
be important for the Bureau to resolve these issues so that the HHCs can
be reevaluated in 2006.

5 GAO, 2000 Census: Contingency Planning Needed to Address Risks That Pose
a Threat to a Successful Census, GAO/GGD-00-06 (Washington, D.C.: Dec. 14,
1999).

6 GAO, 2000 Census: Best Practices and Lessons Learned for More
Cost-Effective Nonresponse Follow-up, GAO-02-196 (Washington, D.C.: Feb.
11, 2002).

Figure 4: Data Transmission Process for Nonresponse Follow-up

Source: GAO analysis of U.S. Census data.

Difficulties began during training when the first transmission was
supposed to occur and continued through the remainder of the test. During
that first transmission, the Bureau needed to upload a number of software
upgrades along with each census worker's first assignment. Many of these
transmissions failed because of the volume of data involved. Thus, without
cases, the trainees could not complete an important section of on-the-job
training. The Bureau acknowledged that these initial problems could have
been avoided if the final version of software had been installed on the
devices prior to their distribution at training.

Transmission problems persisted throughout nonresponse follow-up.
According to the Bureau, during the first 2 weeks of this operation,
successful data transmission occurred 80 percent of the time once a
connection was made. However, a number of enumerators never even
established a connection because of bad phone lines, incorrect passwords,
and improper setup of their modems. Other transmission problems were due
to the local telecommunication infrastructure at both test sites. For
example, in Georgia, older phone lines could not always handle
transmissions, while in Queens, apartment intercoms that used phone lines
sometimes interrupted connections.

Further, while the transmission rate ultimately increased to 95 percent--
roughly the maximum allowed by the technology-that level is still short of
the performance level needed for 2010. During the 2000 Census, a 95
percent success rate would have resulted in the failure to transmit around
30,000 completed questionnaires each day.

During the test, the Bureau also had to contend with census workers who
were "living off the grid"; that is, they only used cellular phones and
lacked landlines to transmit and receive data from their homes. While
individuals could make alternative arrangements, such as usinganeighbor's
telephone, an increasing number of people nationwide in the coming years
might give up their landline service to rely on cellular phones, which
could be problematic for the Bureau. Bureau officials have noted that all
these transmission problems need to be addressed before 2010.

HHCs experienced memory overloads if too many assignment areas were loaded
onto them. An assignment area typically contains 40 housing units or cases
that are assigned to an enumerator for nonresponse follow-up. The design
was to have an entire assignment area transmitted to the HHC even when as
few as one case needed follow-up. However, some enumerators' HHCs became
overloaded with too much data, as cases had to be reassigned due to staff
turnover, a larger-than-expected number of refusals, and reassignments
resulting from language problems. As such, when HHCs became overloaded
they would crash and enumerators had to reconfigure them at the local
census office, which made them less productive. To the Bureau's credit,
during the test, it was able to work out a solution to avoid overloads by
assigning individual cases instead of the entire assignment area to a
census worker's HHC.

Another problem that surfaced during the test was that the HHC's mapping
feature was difficult to use. To contain costs and increase efficiency,
the

Bureau expects to replace paper maps with the electronic maps loaded on
the HHCs for 2010. However, during the test, enumerators reported that
they did not always use the mapping function because it ran slowly and did
not provide sufficient information. Instead, they relied on local maps or
city directories, and one worker explained that she found it easier to use
an Internet mapping service on her home computer to prepare for her route.

Without the Bureau's maps, enumerators might not properly determine
whether a housing unit was located in the Bureau's geographic database.
This verification is important for ensuring that housing units and the
people who reside in them are in the correct census block, as local and
state jurisdictions use census population figures for congressional
redistricting and allocating federal funds.

Enumerators were also unable to use the HHCs' "go back" function to edit
questionnaires beyond a certain point in the interview. In some cases,
this led to the collection of incorrect data. For example, we observed one
worker complete half an interview, and then discover that the respondent
was providing information on a different residence. After the census
worker entered the number of residents and their names, the "go back"
function was no longer available and as a result that data could not be
deleted or edited. Instead, the worker added information in the "notes
section" to explain that the interview had taken place at the wrong
household. However, Bureau officials told us that they had not planned to
review or evaluate these notes and were not aware that such address
mix	ups had been documented in the notes section.

To the extent address mix-ups and other inconsistencies occur and are not
considered during data processing, accuracy could be compromised. In
earlier censuses when the Bureau used paper questionnaires, if workers
made mistakes, they could simply erase them or record the information on
new forms. As mistakes are inevitable, it will be important for the Bureau
to ensure that the HHCs allow enumerators to edit information, while still
maintaining the integrity of the data.

Bureau Needs to Review We found that questions designed to improve
coverage and better Format of Coverage determine race and ethnicity were
awkward for enumerators to ask and Improvement and confusing for
respondents to answer. Consequently, enumerators

sometimes did not read the questions exactly as worded, which
couldRace/Ethnicity Questions adversely affect the reliability of the data
collected for these items, as well as the Bureau's ability to evaluate the
impact of the revised questions. Our

observations also highlight the importance of ensuring that workers are
trained to follow interview protocols; this issue will be discussed later
in this report.

Coverage Improvement While the Bureau attempts to count everyone during a
census, inevitably some people are missed and others are counted more than
once. To help ensure that the Bureau properly counts people where they
live, the Bureau revised and assessed its residency rules for the 2004
census test. For example, under the residence rules, college students
should be counted at their campus addresses if they live and stay there
most of the time. The Bureau also added two new coverage questions aimed
at identifying household residents who might have been missed or counted
in error (see fig. 5 for coverage questions).

Figure 5: New Coverage Questions Were Designed to Ensure a Complete Count

                          Source: U.S. Census Bureau.

  Enumerators were to show respondents flashcards with the residence rules to
    obtain the number of people living or staying in the housing unit and to

read the two coverage questions. However, during our field visits we noted
that they did not consistently use the flashcards, preferring to summarize
them instead. Likewise, enumerators did not always ask the new coverage
questions as written, sometimes abbreviating or skipping them altogether.
A frequent comment from the workers we spoke with was that the two new
coverage questions were awkward because the questions seemed redundant.
Indeed, one census worker said that he asked the overcount and undercount
questions more times than not, but if people were in a hurry, he did not
ask the questions. During one of these hurried interviews, we observed
that the census worker did not ask the questions and simply marked "no"
for the response.

Race and Ethnicity QuestionsCollecting reliable race and ethnicity data is
an extremely difficult task. Both characteristics are subjective, which
makes accurate measurement problematic. In 2003, the Bureau tested seven
different options for formatting the race and ethnic questions, and
selected what it thought was the optimal approach to field test in 2004.
The Bureau planned to examine respondent reaction to the new race and
Hispanic origin questions by comparing responses collected using the paper
questionnaire to responses recorded on the HHCs during nonresponse
follow-up.

One change the Bureau planned to analyze was the removal of the "some
other race" write-in option from the questionnaire. In 2000, the Bureau
found that when given this option, respondents would check off "some other
race," but did not always write in what their race was. Thus, in the 2004
test, the Bureau wanted to assess respondents' reaction to the removal of
the "some other race" write-in option. Specifically, the Bureau wanted to
see whether respondents would skip the item or select from one of the
other options given.

However, we found that the Bureau formatted the race question on the paper
questionnaire differently from the question on the HHC. As shown in figure
6, on the paper version, there is not a category for another race other
than those categories listed, thus forcing respondents to select a
category or skip the question entirely.

This contrasts with the HHCs where, if respondents do not fit into one of
the five race categories, the questionnaire format allows them to provide
an "other" response and enumerators can record their answers. In fact, the
HHC requires enumerators to record a response to the race question and
will not allow the interview to continue until a response is entered. As a
result, the data recorded by the two questionnaire formats are not

comparable as they could produce different data depending on the data
collection mode.

 Figure 6: Race and Ethnicity Categories on the HHCs Were Formatted Differently
                         From the Paper Questionnaires

Source: U.S. Census Bureau.

According to the Bureau, it formatted the paper version of the race
question differently from the HHC version because it considered the
"other" response option on the HHC a respondent comment and not a write-in
response. Nevertheless, if the Bureau's purpose is to measure respondent
reaction to eliminating the write-in option, it is uncertain what
conclusions the Bureau will be able to draw given that this option, even
though in the form of a comment, is still available to the respondent
during the nonresponse follow-up interview.

As was the case with the coverage measurement question, enumerators at
both test locations did not always follow proper interview procedures
because they felt the questions were awkward to ask and confused
respondents. For example, some workers did not use the flashcards designed
to guide respondents in selecting categories for their race and ethnicity
and to ensure data consistency. One census worker said that rather than
use the flashcards or ask the questions, he might "eyeball" the race and
ethnicity. Another worker said that most people laughed at the Spanish,
Hispanic, or Latino origin question and she had complaints about the
wording of this question. A third census worker noted that he was "loose
with the questions" because he could pose them better. Like lapses to the
coverage improvement procedures for the 2004 census test, deviating from
the interview procedures for the new race and ethnicity questions may
affect the reliability of the data and the validity of the Bureau's
conclusions concerning respondent reaction to these questions.

Since the 2004 census test, the 2005 Consolidated Appropriations Act7
required that the Bureau include "some other race" as a category when
collecting census data on race identification. Consequently, the Bureau
said it will include this category on all future census tests and the 2010
Census itself. Thus, while research into eliminating the "some other race"
category is now moot, it will still be important for the Bureau to have
similar formats for the HHCs and paper questionnaires so that similar data
can be captured across modes. Likewise, it will be important for the
wording of those questions to be clear and for enumerators to follow
proper procedures during interviews.

New Procedures Should Help Reduce Duplicate Enumerations of Group Quarter
Residents, but Other Challenges Remain

As noted previously, under its residence rules, the Bureau enumerates
people where they live and stay most of the time. To facilitate the count,
the Bureau divides residential dwellings into two types: housing units,
such as single-family homes and apartments, and group quarters, which
include dormitories, prisons, and nursing homes.

The Bureau tested new group quarters procedures in 2004 that were designed
to address the difficulties the Bureau had in trying to identify and count
this population group during the 2000 Census. For example,

7 Consolidated Appropriations Act, 2005, Pub. L. No. 108-447, Div. B,
Title II, Dec. 8, 2004.

communities reported instances where prison inmates were counted in the
wrong county and residents of college dormitories were counted twice.

One refinement the Bureau made was integrating its housing unit and group
quarter address lists in an effort to avoid counting them once as group
quarters and again as housing units, a common source of error during the
2000 Census. Census workers were then sent out to verify whether the
dwellings were in fact group quarters and, if so, to classify the type of
group quarter using a revised "other living quarters facility"
questionnaire.

A single address list could, in concept, help reduce the duplicate
counting that previously occurred when the lists were separate. Likewise,
we observed that census workers had no problems using the revised facility
questionnaire and accompanying flashcard that allowed the respondent to
select the appropriate type of living facility. This new procedure
addresses some of the definitional problems by shifting the responsibility
for defining the group quarter type from the Bureau to the respondent, who
is in a better position to know about the dwelling.

Another change tested in 2004 was the classification of group homes, which
in 2000 was a part of the group quarter inventory. Group homes are
sometimes difficult for census workers to spot because they often look the
same as conventional housing units (see fig. 7). As a result, they were
sometimes counted twice during the 2000 Census-once as a group quarter,
and once as a housing unit. For the 2004 test, the Bureau decided to treat
group homes as housing units and include them in the housing unit list.

Figure 7: Group Homes Could Resemble Conventional Houses

Source: GAO, 2004.

Early indications from the Bureau suggest that including group homes as
housing units, whereby they receive a short-form questionnaire in the
mail, may not work. According to the Bureau, the format of the short form
is not well suited to group home residents. For example, the questionnaire
asks for the "name of one of the people living or staying here who owns or
rents this place." Since the state or an agency typically owns group
homes, these instructions do not apply. The Bureau stated that it plans to
reassess how it will identify and count people living in group homes.

We identified other problems with the Bureau's group quarters validation
operation during the 2004 census test. For example, we were told that
census workers were provided maps of the areas they were assigned but
needed maps for adjoining areas so that they could more accurately locate
the physical location of the group quarters. In Georgia, where workers
used address data from the 2000 Census, the crew leader explained that
approximately one-third of all the addresses provided were incorrectly
spotted on maps and had to be redone. They also lacked up-to-date

instructions-for example, they did not know that they were to correct
addresses rather than just delete them if the addresses were wrong.
Further, census workers said that scenarios in the manual and classroom
training were based on perfect situations; thus, they did not provide
adequate training for atypical settings or when problems arose.

The Bureau Should Rethink Its Approach to Training Enumerators

The success of the census is directly linked to the Bureau's ability to
train enumerators to do their jobs effectively. This is a tremendous task
given the hundreds of thousands of enumerators the Bureau needs to hire
and train in just a few weeks. Further, enumerators are temporary
employees, often with little or no prior census experience, and are
expected, after just a few days of training, to do their jobs with minimal
supervision, under sometimes difficult and dangerous conditions. Moreover,
the individuals who train enumerators-crew leaders-are often recent hires
themselves, with little, if any, experience as instructors. Overall, few,
if any, organizations face the training challenges that confront the
Bureau with each decennial population count.

To train the 1,100 enumerators who conducted nonresponse follow-up for the
2004 test, the Bureau employed essentially the same approach it has used
since the 1970 Census: crew leaders read material word-for-word from a
training manual to a class of 15 to 20 students. The notable exception was
that in transitioning from a paper questionnaire to the HHCs, the Bureau
lengthened the training time from 3 days to 5 days. However, given the
demographic and technological changes that have taken place since 1970,
the Bureau might want to explore alternatives to this rigid approach.

As noted earlier, during nonresponse follow-up, enumeratorsexperienced a
variety of problems that could be mitigated through improved training. The
problems included difficulties setting up equipment to transmit and
download data; failure to read the coverage and race/ethnicity questions
exactly as worded; and not properly using the flashcards, which were
designed to help respondents answer specific questions.

Most of the shortcomings related to training that we observed during the
test were not new. In fact, the Bureau had identified these and a number
of other training weaknesses in its evaluation of the 2000 Census, but it
is clear they have not been fully resolved. Thus, as the Bureau plans for
the 2010 Census, it will be important for it to resolve long-standing
training problems as well as address new training issues, such as how best
to teach

enumerators to use the HHCs and their associated automated processes. Our
observations of the test point to specific options the Bureau might want
to explore. They include (1) placing greater emphasis on the importance of
following prescribed interview procedures and reading questions exactly as
worded; (2) supplementing verbatim, uniform training with modules geared
toward addressing the particular enumeration challenges that census
workers are likely to encounter at specific locales; and (3) training on
how to deal with atypical situations or respondent reluctance.

To help evaluate its future training needs, the Bureau hired a contractor
to review the training for the 2004 test and recommend actions for
improving it. From GAO's work on assessing agencies' training and
development efforts, we have developed a framework that can also help in
this regard.8 Though too detailed to discuss at length in this report,
highlights of the framework, and how they could be applied to census
training, include:

1.performing proper front-end analysis to help ensure that the Bureau's
enumerator training is aligned with the skill and competencies needed to
meet its field data collection requirements and work processes and that
the Bureau leverages best practices and lessons learned from training
enumerators and from past experience;

2. identifying specific training initiatives that in conjunction with
other strategies, improve enumerators' performance and help the Bureau
meet its goal of collecting high-quality data from nonrespondents;

3.ensuring effective and efficient delivery of training that reinforces
new and needed competencies, skills, and behaviors without being wedded to
past, and perhaps outmoded, methods; and

4.evaluating the training to ensure it is addressing known skill and
competency weaknesses through such measures as assessing participant
reactions and changes in enumerators' skill levels and behaviors.

8 GAO, Human Capital: A Guide for Assessing Strategic Training and
Development Efforts in the Federal Government, GAO-04-546G (Washington,
D.C.: March 2004).

Readiness Will Be Critical for Future Tests

Several key features of the 2004 test were not test ready; that is, they
were not fully functional or mature when they were employed at the test
sites. This is a serious shortcoming because it hampered the Bureau from
fully evaluating and refining the various census-taking procedures that
will be used in subsequent tests and the actual census in 2010. Further,
to the extent these features were integrated with other operations, it
impeded the Bureau from fully assessing those associated activities as
well.

Our work, and that of the Department of Commerce Inspector General,9
identified the following areas where the Bureau needed to be more prepared
going into the test:

o The HHCs crashed, in part, because earlier testing did not identify
software defects that caused the download of more data to the HHCs than
their memory cards could hold.

o Transmission failures occurred during enumerator training, in part,
because the HHCs were shipped without the latest version of needed
software. Although the Bureau ultimately provided the latest software
after several weeks, the upgraded version was unavailable for training
field operations supervisors and crew leaders and for the initial
enumerator training.

o According to the Department of Commerce Inspector General, the Bureau
finalized the requirements for the new group quarter definitions too late
for inclusion in group quarters training manuals. Consequently, the
training lacked certain key instructions, such as how to categorize group
homes.

The Bureau experienced other glitches during the test that with better
preliminary testing or on-site dry runs, might have been detected and
possibly addressed before the test started. These included the slow
start	up of the HHC's mapping function, and the tendency for apartment
house intercoms to interrupt transmissions.

An important objective of any type of test is to identify what is working
and where improvements are needed. Thus, it should not be surprising, and,
in

9U.S. Department of Commerce, Office of Inspector General, Improving Our
Measure of America: What the 2004 Census Test Can Teach Us in Planning for
the 2010 Decennial Census, OIG-16949 (Washington, D.C.: September 2004).

fact, should be expected and commended, that shortcomings were found with
some of the various activities and systems assessed during the 2004 test.
We believe that the deficiency is not the existence of problems; rather it
is the fact that several components were incomplete or still under
development going into the test, which made it difficult for the Bureau to
gauge their full potential. The Bureau had a similar experience in the
dress rehearsal for the 2000 Census, when, because a number of new
features were not test ready, the Bureau said it could not fully test them
with any degree of assurance as to how they would affect the head count.

Because of the tight time frames and deadlines of the census, the Bureau
needs to make the most of its limited testing opportunities. Thus, as the
Bureau plans for the next field test in 2006 and the 2008 dress rehearsal,
it will be important for the Bureau to ensure the various census
operations are fully functional at the time of the test so they can be
properly evaluated.

ConclusionsThe Bureau is well aware that a successful enumeration hinges
on early research, development, testing, and evaluation of all aspects of
the census design. This is particularly true for the 2010 Census for
which, under its current plan, the Bureau will be relying on HHCs and
other methods and technologies that (1) have never been used in earlier
censuses and (2) are mission critical. Consequently, the 2004 test was an
important milestone in the 2010 life cycle because it demonstrated the
fundamental feasibility of the Bureau's basic design and allows the Bureau
to advance to the next and more mature phase of planning and development.

Nevertheless, while the test revealed no fatal flaws in the Bureau's
approach, the results highlighted serious technical, training,
methodological, and procedural difficulties that the Bureau will need to
resolve. Since one of the purposes of testing is to determine the
operational feasibility of the census design, it is not surprising that
problems surfaced. However, looking toward the future, it will be critical
for the Bureau to diagnose the source of these challenges, devise
cost	effective solutions, and integrate refinements and fixes in time to
be assessed during the next field test scheduled for 2006. It will also be
important for Congress to monitor the Bureau's progress as it works to
resolve these issues.

Recommendations for Executive Action

To facilitate effective census planning and development, and to help the
Bureau achieve its key goals for the census-reduce risks, improve
accuracy, and contain costs, we recommend that the Secretary of Commerce
direct the Bureau to take the following eight actions:

o Analyze the impact that HHCs and the targeted second mailing had on cost
savings and other Bureau objectives.

o Ensure the Bureau's IT security practices are in full compliance with
applicable requirements, such as the FISMA, as well as its own internal
policies.

o Enhance the reliability and functionality of HHCs by, among other
actions, (1) improving the dependability of transmissions, (2) exploring
the ability to speed up the mapping feature, (3) eliminating the causes of
crashes, and (4) making it easier for enumerators to edit questionnaires.

o Define specific, measurable performance requirements for the HHCs and
other census-taking activities that address such important measures as
productivity, cost savings, reliability, durability, and test their
ability to meet those requirements in 2006.

o Review and test the wording and formatting of the coverage and
race/ethnicity questions to make them less confusing to respondents and
thus help ensure the collection of better quality data, and ensure they
are formatted the same way on both the HHC and paper versions of the
census form.

o Develop a more strategic approach to training by ensuring the curriculum
and instructional techniques (1) are aligned with the skills and
competencies needed to meet the Bureau's data collection requirements and
methodology and (2) address challenges identified in the 2004 test and
previous censuses.

o Revisit group quarter procedures to ensure they allow the Bureau to best
locate and count this population group.

o Ensure that all systems and other census-taking functions are as mature
as possible and test ready prior to their deployment for the 2006 test, in
part by conducting small-scale, interim tests under the various

conditions and environments the Bureau is likely to encounter during the
test and actual enumeration.

Further, to ensure the transparency of the census-planning process and
facilitate Congressional monitoring, we also recommend that the Secretary
of Commerce direct the Bureau to regularly update Congress on the progress
it is making in addressing these and any other challenges, as well as the
extent to which the Bureau is on track for meeting the overall goals of
the 2010 Census.

Agency Comments and Our Evaluation

The Under Secretary for Economic Affairs at the Department of Commerce
forwarded us written comments from the Census Bureau on a draft of this
report on December 20, 2004, which are reprinted in appendix I. The Bureau
noted that the 2004 test was its first opportunity to assess a number of
the new methods and technologies under development for 2010, and
emphasized the importance of a sustained, multiyear planning, testing, and
development program to its census modernization effort.

The Bureau generally agreed with seven of our nine recommendations, and
described the steps it was taking to address our concerns. The Bureau also
provided additionalcontext and clarifyinglanguage and we haveadded this
information to the report where appropriate.

Specifically, the Bureau generally agreed with our recommendations
relating to improving IT security practices, the reliability of the HHCs,
training, testing, and enumeration procedures-and reported it was already
taking a number of steps to address our concerns. We commend the Bureau
for recognizing the risks and challenges that lie ahead and taking action
to address them. We will continue to monitor the Bureau's progress in
resolving these issues and update Congress on a regular basis.

At the same time, the Bureau took exception to our recommendations to (1)
analyze the impact that HHCs and targeted second mailings had on cost
savings and other Bureau objectives, and (2) define specific, measurable
performance requirements for the HHCs and other census-taking activities
and test their ability to meet those requirements in 2006. With respect to
the first recommendation, the Bureau noted that it did not establish
cost	savings and other impacts as test objectives, in part, because the
Bureau believes that the national sample mail test that it conducted in
2003 provided a better method for determining the boost in response rates
that could accrue from a second mailing. The Bureau maintains that
analyzing

the impact of the second mailing would provide it with no more information
beyond what it has already established from the 2003 test and would be of
little value.

We believe this recommendation still applies because it will be important
for the Bureau to assess the impact of the targeted second mailing on
other Bureau objectives. As we noted in the report, the Bureau included
the HHCs and targeted second mailing in the 2010 Census design, in part,
to reduce staff, improve productivity, and control costs. Further, as we
also note in the report, the feasibility of a targeted second mailing is
an open question. Thus, information on the degree to which the HHCs and
second mailing contribute to these key goals could help inform future
budget estimates, investment and design decisions, as well as help refine
future census tests. In short, the purpose of the analysis we recommend
would not be to see whether these features of the 2010 Census will produce
cost	savings, but the extent of those savings and the impact on other
Bureau objectives.

With respect to the second recommendation, the Bureau noted that it had
"baseline assumptions" about productivity, cost-savings, and other
measures for the 2004 Census test and that a key objective of the test was
to gather information to help refine these assumptions. According to the
Bureau, this will also be a key objective of the 2006 Census Test,
although its performance goal will not be whether it meets specific
measures. Instead, the Bureau intends to focus on successfully collecting
information to further refine those assumptions. As a result, the Bureau
believes the 2006 test will not be a failure if HHC productivity is not
achieved, but that it will be a failure if productivity data are not
collected.

The Bureau's position is inconsistent with our recommendation which we
believe still applies. As noted in the report, we call on the Bureau to
define measurable performance requirements for the HHCs as well as take
the next step and assess whether the HHCs can meet those requirements as
part of the 2006 test. This information is essential because it will help
the Bureau gauge whether HHCs can meet its field data collection needs in
2010. Should the HHCs fail to meet these pre-specified performance
requirements during the 2006 test, the Bureau would need to rethink how it
employs these devices in 2010.

As agreed with your offices, unless you release its contents earlier, we
plan no further distribution of this report until 30 days from its date.
At that

time, we will send copies of this report to the Secretary of Commerce and
the Director of the U.S. Census Bureau. Copies will be made available to
others on request. This report will also be available at no charge on
GAO's home page at http://gao.gov. Please contact me at (202) 512-6806 or
[email protected] or Robert Goldenkoff, Assistant Director, at (202)
512	2757 or [email protected] if you have any questions. Key
contributors to this report were Tom Beall, David Bobruff, Betty Clark,
Robert Dacey, Richard Donaldson, Elena Lipson, Ronald La Due Lake, Robert
Parker, Lisa Pearson, and William Wadsworth.

Patricia A. Dalton Director Strategic Issues

Appendix I

Comments from the Department of Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

Appendix I
Comments from the Department of
Commerce

GAO's Mission	The Government Accountability Office, the audit, evaluation
and investigative arm of Congress, exists to support Congress in meeting
its constitutional responsibilities and to help improve the performance
and accountability of the federal government for the American people. GAO
examines the use of public funds; evaluates federal programs and policies;
and provides analyses, recommendations, and other assistance to help
Congress make informed oversight, policy, and funding decisions. GAO's
commitment to good government is reflected in its core values of
accountability, integrity, and reliability.

Obtaining Copies of The fastest and easiest way to obtain copies of GAO
documents at no cost

is through GAO's Web site (www.gao.gov). Each weekday, GAO postsGAO
Reports and newly released reports, testimony, and correspondence on its
Web site. To Testimony have GAO e-mail you a list of newly posted products
every afternoon, go to

www.gao.gov and select "Subscribe to Updates."

Order by Mail or Phone	The first copy of each printed report is free.
Additional copies are $2 each. A check or money order should be made out
to the Superintendent of Documents. GAO also accepts VISA and Mastercard.
Orders for 100 or more copies mailed to a single address are discounted 25
percent. Orders should be sent to:

U.S. Government Accountability Office 441 G Street NW, Room LM Washington,
D.C. 20548

To order by Phone:	Voice: (202) 512-6000 TDD: (202) 512-2537 Fax: (202)
512-6061

To Report Fraud, Contact:
Waste, and Abuse in Web site: www.gao.gov/fraudnet/fraudnet.htm

E-mail: [email protected] Programs Automated answering system: (800)
424-5454 or (202) 512-7470

Congressional	Gloria Jarmon, Managing Director, [email protected] (202)
512-4400 U.S. Government Accountability Office, 441 G Street NW, Room 7125

Relations Washington, D.C. 20548

Public Affairs	Susan Becker, Acting Manager, [email protected] (202)
512-4800 U.S. Government Accountability Office, 441 G Street NW, Room 7149
Washington, D.C. 20548

                               Presorted Standard
                              Postage & Fees Paid
                                      GAO
                                Permit No. GI00

United States
Government Accountability Office
Washington, D.C. 20548-0001

Official Business
Penalty for Private Use $300

Address Service Requested
*** End of document. ***