This is the accessible text file for GAO report number GAO-06-272 
entitled '2010 Census: Census Bureau Needs to Take Prompt Actions to 
Resolve Long-Standing and Emerging Address and Mapping Challenges' 
which was released on June 15, 2006. 

This text file was formatted by the U.S. Government Accountability 
Office (GAO) to be accessible to users with visual impairments, as part 
of a longer term project to improve GAO products' accessibility. Every 
attempt has been made to maintain the structural and data integrity of 
the original printed product. Accessibility features, such as text 
descriptions of tables, consecutively numbered footnotes placed at the 
end of the file, and the text of agency comment letters, are provided 
but may not exactly duplicate the presentation or format of the printed 
version. The portable document format (PDF) file is an exact electronic 
replica of the printed version. We welcome your feedback. Please E-mail 
your comments regarding the contents or accessibility features of this 
document to Webmaster@gao.gov. 

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. 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. 

United States Government Accountability Office: 

Report to the Congressional Committees: 

GAO: 

June 2006: 

2010 Census: 

Census Bureau Needs to Take Prompt Actions to Resolve Long-standing and 
Emerging Address and Mapping Challenges: 

GAO-06-272: 

GAO Highlights: 

Highlights of GAO-06-272, a report to congressional committees. 

Why GAO Did This Study: 

To conduct a successful census, it is important that the U.S. Census 
Bureau (Bureau) produce the most complete and accurate address file and 
maps for 2010. For this review, GAO’s specific objectives were to 
determine the extent to which (1) the Bureau’s efforts to modernize the 
address file and maps are addressing problems experienced during the 
2000 Census, (2) the Bureau is managing emerging address file and map 
issues, (3) the Bureau is able to collect and transmit address and 
mapping data using mobile computing devices (MCD) equipped with global 
positioning system (GPS) technology, and (4) the Bureau has a plan to 
update the address file and maps in areas affected by hurricanes 
Katrina and Rita. GAO reviewed the Bureau’s progress in modernizing 
both the address file and maps. 

What GAO Found: 

The Bureau’s address and map modernization efforts have progressed in 
some areas. The Bureau is researching how to correct addresses that 
were duplicated, missed, deleted, and incorrectly located on maps. 
However, some deadlines for completing research are not firm, while 
other deadlines that had been set continue to slip. Thus, whether 
research will be completed in enough time to allow the Bureau to 
develop new procedures to improve the 2010 address file is unknown. 
Also, the Bureau has not fully addressed emerging issues. For one such 
issue, the Bureau has acknowledged the compressed time frame for 
completing address canvassing—an operation where census workers walk 
every street in the country to verify addresses and maps—but has not 
reevaluated the associated schedule or staffing workloads. Also, the 
Bureau has allotted only 6 weeks to conduct address canvassing it 
completed in 18 weeks in 2000 and expanded the operation from urban 
areas in 2000 to the entire country in 2010. 

Figure: Mobile Computing Devices for Collecting and Transmitting Field 
Data: 

[See PDF for Image] 

[End of Figure] 

Whether the Bureau can collect and transmit address and mapping data 
using the MCD is unknown. The MCD, tested during 2006 address 
canvassing, was slow and locked up frequently. Bureau officials said 
the MCD’s performance is an issue, but a new MCD to be developed 
through a contract awarded in March 2006 will be reliable. However, the 
MCD will not be tested until the 2008 Dress Rehearsal, and if problems 
emerge, little time will remain to develop, test, and incorporate 
refinements. If after the Dress Rehearsal the MCD is found unreliable, 
the Bureau could face the remote but daunting possibility of reverting 
to the costly paper-based census of 2000. 

Bureau officials do not believe a specific plan is needed to update the 
addresses and maps for areas affected by the hurricanes. Securing a 
count is difficult under normal conditions, and existing procedures may 
insufficient to update addresses and maps after the hurricanes’ 
destruction—made even more difficult as streets, housing, and 
population will be in flux. 

What GAO Recommends: 

GAO recommends that the Secretary of Commerce direct the Bureau to 
mitigate risks in building its address file and maps. Specific actions 
include setting firm dates to complete research and evaluations and 
develop resulting action plans; reevaluating the schedule and staffing 
workloads for conducting address canvassing; and developing plans to 
assess resources needed to update the address file and maps along the 
Gulf Coast. In commenting on a draft of this report, Commerce agreed 
with each of GAO’s three recommendations. 

[Hyperlink, http://www.gao.gov/cgi-bin/getrpt?GAO-06-272]. 

To view the full product, including the scope and methodology, click on 
the link above. For more information, contact Brenda S. Farrell at 
(202) 512-6806 or farrellb@gao.gov. 

[End of Section] 

Contents: 

Letter: 

Results In Brief: 

Background: 

Uncertainties Surround Completion of Ongoing MAF/TIGER Modernization 
Research: 

Emerging Issues Related to Overlapping and Compressed Schedules Pose a 
Risk to MAF/TIGER Modernization Efforts: 

Reliability of MCD to Conduct Address Canvassing Activities is Unknown: 

Bureau Does Not Have a Specific Plan for Updating MAF/TIGER in the 
Aftermath of Hurricanes Katrina and Rita: 

Conclusions: 

Recommendations for Executive Action: 

Agency Comments and Our Evaluation: 

Appendix I: Scope and Methodology: 

Appendix II: Comments from the Department of Commerce: 

Appendix III: GAO Contact and Staff Acknowledgments: 

GAO Contact: 

Acknowledgments: 

Table: 

Table 1: Status of Bureau Efforts to Resolve MAF/TIGER Issues: 

Figures: 

Figure 1: Key Operations Required for a Complete and Accurate MAF/TIGER 
for the 2000 Census: 

Figure 2: Group Homes Can Resemble Housing Units: 

Figure 3: TIGER Map Overlay of an Aerial Photograph: 

Figure 4: Devastation in the Lower Ninth Ward in New Orleans, 
Louisiana: 

United States Government Accountability Office: 

Washington, DC 20548: 

June 15, 2006: 

The Honorable Susan M. Collins: 
Chairman: 
The Honorable Joseph I. Lieberman: 
Ranking Member: 
Committee on Homeland Security and Governmental Affairs: 
United States Senate: 

The Honorable Tom Coburn: 
Chairman: 
The Honorable Thomas R. Carper: 
Ranking Minority Member: 
Subcommittee on Federal Financial Management, Government Information, 
and International Security: 
Committee on Homeland Security and Governmental Affairs: 
United States Senate: 

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

The Honorable Michael R. Turner: 
Chairman: 
The Honorable William Lacy Clay: 
Ranking Minority Member: 
Subcommittee on Federalism and the Census: 
Committee on Government Reform: 
House of Representatives: 

The decennial census is an important, constitutionally mandated 
activity undertaken by the federal government that is complex and 
costly--estimated at $11.3 billion for the 2010 Census. The data that 
the census produces are used to reapportion the seats of the U.S. House 
of Representatives; realign the boundaries of the legislative districts 
of each state; allocate hundreds of billions of dollars in federal 
financial assistance; and provide a social, demographic, and economic 
profile of the nation's people to guide policy decisions at each level 
of government. The U.S. Department of Commerce's Census Bureau (Bureau) 
is responsible for conducting the decennial census, and the success of 
the census depends in large part on the ability of the Bureau to locate 
and deliver questionnaires to every person residing in the United 
States. To successfully accomplish this monumental task, the Bureau 
must maintain accurate address and map information for every location 
where a person could reside. During the 2000 Census, Bureau evaluations 
estimated that of the 116 million housing units in the final census 
count, about 2.3 million housing units were incorrectly included in the 
census and about 2.7 million housing units were missed. 

One of the Bureau's principal objectives for the 2010 Census is 
modernizing the Master Address File (MAF)--the Bureau's repository of 
approximately 130 million addresses to which the Bureau expects to 
deliver census forms for the 2010 Census. The Bureau also works to 
ensure the accuracy of the associated mapping system, the Topologically 
Integrated Geographic Encoding and Referencing (TIGER®) 
database.[Footnote 1] The Bureau hopes to improve the completeness and 
accuracy of MAF/TIGER through a combination of activities, including 
partnering with state and local governments to verify the address lists 
and maps and updating maps to reflect correct geographic features. The 
Bureau will also use satellite-based global positioning system (GPS) 
technology to correctly locate housing units and door-to-door 
canvassing to verify the status of all housing units. The combined cost 
of these efforts is estimated to be about $536 million (nominal 
dollars). 

An important component of the Bureau's attempts to modernize its 
address listing and mapping activities will be the planned use of 
relatively new technology. For the first time, census workers will use 
a GPS-equipped mobile computing device (MCD) to collect data in the 
field, including address and map data. The Bureau anticipates that the 
MCDs will be used in three major census operations, and their 
successful implementation would allow the Bureau to reduce the amount 
of paper used, process data in real time, and improve the quality of 
the information collected. 

Because of the critical importance of complete and accurate address 
lists and maps, under the Comptroller General's statutory authority, we 
reviewed the Bureau's progress in modernizing both MAF and TIGER. As 
agreed with your offices, we are providing this report to you which 
contains information that will be useful for your oversight 
responsibilities of the decennial census. Our specific objectives were 
to determine the extent to which (1) the Bureau's efforts to modernize 
the address file and maps are addressing problems experienced during 
the 2000 Census, (2) the Bureau is managing emerging MAF/TIGER issues, 
(3) the Bureau is able to collect and transmit address and mapping data 
using a MCD that is equipped with GPS technology, and (4) the Bureau 
has a plan to update the address file and maps in areas affected by 
hurricanes Katrina and Rita. 

To meet these objectives, we analyzed relevant evaluations from the 
2000 Census and other studies conducted by the Bureau, the Department 
of Commerce Office of Inspector General, and other organizations. We 
also reviewed various documents describing the Bureau's MAF/TIGER 
modernization efforts and interviewed knowledgeable Bureau officials 
about MAF/TIGER, including the Bureau's plans to update MAF/TIGER in 
the wake of hurricanes Katrina and Rita. Further, to obtain a firsthand 
look at how the Bureau's address-building operations and MCDs performed 
in a real-world environment, we observed address canvassing activities 
at the 2006 Census Test sites located at the Cheyenne River American 
Indian Reservation and Tribal Trust Lands in South Dakota and the 
central portion of Travis County, Texas. Address canvassing is an 
operation where census workers walk every street in the country, 
verifying addresses and updating maps. We conducted our work from June 
2005 through April 2006 in accordance with generally accepted 
government auditing standards. Additional information on our scope and 
methodology appears in appendix I. 

Results In Brief: 

While the Bureau's MAF/TIGER modernization efforts have progressed in a 
number of areas, it is not clear if research designed to resolve 
address-related issues from the last census will be completed in 
sufficient time to improve 2010 address-building activities. During the 
2000 Census, the Bureau encountered a number of problems with the MAF, 
including addresses that were duplicated, missed, deleted, and 
incorrectly located on the maps. To address those problems, the Bureau 
has been conducting research and testing some operational changes. For 
example, the Bureau is researching ways to capture missed addresses for 
housing units that were hard to find--often associated with apartments 
in small, multi-unit structures. However, some deadlines for completing 
research are not firm, while other deadlines that have been set 
continue to slip. As a result, it is not known whether the research and 
evaluation efforts currently under way will be completed in sufficient 
time to allow the Bureau to develop new methodologies and procedures 
for improving the MAF by June 2007, the Bureau's announced deadline for 
baselining all program requirements. In addition, one major research 
effort using software to identify duplicate addresses (an estimated 1.4 
million duplicate addresses were removed during the 2000 Census) did 
not work any better at identifying true duplicates than what the Bureau 
already had in place and will not be used in 2010. As a result, 
duplicate addresses may still be a problem for the 2010 MAF, and if not 
detected, can result in reduced accuracy and increased cost. 

As the Bureau has planned for the 2010 Census, issues surrounding the 
schedule of address activities have emerged and have not been fully 
addressed. One such issue revolves around the planning and development 
of the census amid tight and overlapping schedules for updating 
addresses and map files. For example, Bureau officials estimate that 
TIGER maps for 600 to 700 counties of 3,232 counties in the United 
States will not be updated in time to be part of the local update of 
census addresses (LUCA)--the Bureau's program to give local, state, and 
tribal government officials the opportunity to review the address lists 
and maps and suggest corrections. LUCA participation is important 
because local knowledge contributes to a more complete and accurate 
address file, and not having the most current TIGER maps could affect 
the quality of a local government's review. Also, the Bureau has 
compressed the time frame for completing address canvassing--an 
operation where census workers walk every street in the country to 
verify addresses and update maps. The Bureau has allotted 6 weeks for 
verifying the nation's inventory of 116 million housing units, although 
the Bureau took 18 weeks to complete this operation for the 2000 
Census. The time to complete address canvassing is a concern because 
the workload for address canvassing has significantly expanded from 
including only urban areas in 2000 to including the entire country for 
2010. Bureau officials acknowledged the compressed time frame and that, 
in some areas of the country, bad weather could result in more time 
being needed to complete address canvassing. Bureau officials did not 
provide a justification for reducing the amount of time by 12 weeks, 
but did state that they would need to adjust staffing levels to meet 
workload demands. 

The Bureau's ability to collect and transmit address and mapping data 
using the MCD is not known. The performance of these devices is crucial 
to the accurate, timely, and cost-effective completion of address 
listing, nonresponse follow-up, and coverage measurement activities. 
During 2006 testing, the MCD used to collect address and map data was 
slow and locked up frequently. As a result, the Bureau was unable to 
complete address canvassing, even with a 10-day extension. Also, some 
census workers were not always able to get GPS signals for collecting 
coordinates for housing units. Bureau officials have acknowledged that 
the MCD's performance is an issue, but believe that a new version of 
the MCD, to be developed under the Field Data Collection Automation 
(FDCA) contract awarded on March 30, 2006, will be reliable and 
functional. However, because the 2008 Dress Rehearsal will be the first 
time this new MCD will be tested under census-like conditions, it is 
uncertain how effective that MCD will be, and if problems do emerge, 
little time will be left for the contractor to develop and test any 
refinements. Further, if after the dress rehearsal the MCD is found not 
to be reliable, the Bureau could be faced with the remote, but 
daunting, possibility of having to revert to a costly paper-based 
census used in 2000. 

Finally, Bureau officials do not believe they need to have a specific 
plan to update the address and maps files for those areas affected by 
hurricanes Katrina and Rita. Securing a complete count is difficult 
under normal circumstances, and the destruction caused by the 
hurricanes makes it even more difficult because the baseline 
information the Bureau must work with--streets, housing, and the 
population itself--will be in flux for some time to come. Bureau 
officials stated that by 2009, when address fieldwork is set to begin, 
residents will have decided whether to return to the affected region. 
Therefore, they believe that by 2009, they will be in a better position 
to add or delete addresses in the Gulf region affected by the 
hurricanes. However, Bureau officials could not provide support for the 
2009 date, nor have they identified local partners with whom they can 
monitor this situation. Given the magnitude of the area affected and 
the degree of destruction, this approach may not be adequate. As a 
result, the quality of the address and map files could be reduced if 
the Bureau is not prepared to conduct address operations in those areas 
affected by hurricanes Katrina and Rita. 

In conversations with Bureau officials, it became apparent to us that 
they are keenly aware of the existing time constraints and challenges 
detailed in this report. However, the Bureau had not developed risk 
mitigation plans to address these challenges. Our recommendations, 
therefore, are intended to make transparent for Bureau and 
congressional decision makers how those challenges can and should be 
addressed. At a minimum, the Bureau should have a risk-based mitigation 
plan in place that includes specific dates for completing research on 
the address file and an approach for exploring the difficulties the 
Bureau may face in updating MAF/TIGER along the Gulf Coast. Because 
time is running short, it is imperative that the Bureau continue to 
stay focused on identifying and resolving problems to help ensure that 
the most accurate and complete address file and maps are produced for 
the 2010 Census. To facilitate this, we recommend that the Secretary of 
Commerce direct the Bureau to address methodological, timing, and 
procedural improvements to building its address file and maps. Specific 
actions include (1) establishing firm deadlines to complete research, 
testing, and evaluations of the MAF to prevent missed, deleted, or 
duplicate addresses and map errors, and develop an action plan that 
will allow sufficient time to affect the 2010 MAF/TIGER design; (2) 
reevaluating the 2010 address canvassing schedule in areas affected by 
bad weather as well as staffing levels to ensure that the status of all 
housing units are accurately verified throughout the entire country; 
and (3) developing a plan, prior to the start of LUCA in August 2007, 
that will assess whether new procedures, additional resources, or local 
partnerships are needed to update MAF/TIGER along the Gulf Coast for 
areas affected by hurricanes Katrina and Rita. 

On June 2, 2006, the Department of Commerce forwarded written comments 
from the Bureau on a draft of this report. The Bureau agreed with each 
of our three recommendations and also noted actions it was taking to 
address the recommendations. The Bureau's comments also included some 
technical corrections and suggestions where additional context was 
needed, and we revised the report to reflect these comments as 
appropriate. The comments are reprinted in their entirety in appendix 
II. 

Background: 

A complete and accurate address list is the cornerstone of a successful 
census, because it both identifies all households that are to receive a 
census questionnaire and serves as the control mechanism for following 
up with households that fail to respond. If the address list is 
inaccurate, people can be missed, counted more than once, or included 
in the wrong location. MAF is intended to be a complete and current 
list of all addresses and locations where people live or could live. 
The TIGER database is a mapping system that identifies all visible 
geographic features, such as type and location of streets, housing 
units, rivers, and railroads. To link these two separate databases, the 
Bureau assigns every housing unit in the MAF to a specific location in 
the TIGER, a process called "geocoding." 

As shown in figure 1, for the 2000 Census the Bureau's approach to 
building complete and accurate address lists and maps consisted of a 
number of labor-and data-intensive operations that sometimes overlapped 
and were conducted over several years. This effort included 
partnerships with the U.S. Postal Service and other federal agencies; 
state, local, and tribal governments; local planning organizations; the 
private sector; and nongovernmental entities. The Bureau employed 
thousands of temporary census workers to walk every street in the 
country to locate and verify places where people could live. 
Determining this was no simple task as people can reside in cars, 
sheds, illegally converted basements and garages, and similar 
nontraditional and often hidden living arrangements. 

Figure 1: Key Operations Required for a Complete and Accurate MAF/TIGER 
for the 2000 Census: 

[See PDF for image] 

[End of figure] 

For the 2000 Census, the Bureau found that the MAF/TIGER databases were 
less than complete and accurate. Although the number of errors was 
small in proportion to the total number of housing units at the 
national level, the errors could be problematic at lower levels of 
geography for certain purposes for which census data are used, such as 
allocating federal assistance to state and local governments. 

According to Bureau evaluations conducted after the 2000 Census, the 
final census count contained approximately 116 million housing units. 
However, the address file used to conduct the 2000 Census also 
contained a number of errors.[Footnote 2] Bureau evaluations estimate 
that there were: 

* 0.7 million duplicate addresses, 

* 1.6 million vacant housing units misclassified as occupied, 

* 1.4 million housing units not included, 

* 1.3 million housing units improperly deleted, and: 

* 5.6 million housing units incorrectly located on census maps. 

In light of these and other problems, the Bureau made enhancing the 
MAF/TIGER one of three critical components to support the 2010 Census. 
The other two components are replacing the long form questionnaire with 
the American Community Survey[Footnote 3] and conducting a short-form- 
only decennial census that is supported by early research and testing. 

For the 2010 Census, the Bureau is making extensive use of contractors 
to provide a number of mission-critical functions and technologies. One 
of the technologies to be provided by a contractor is the MCD. Under a 
contract awarded on March 30, 2006, a new MCD will be developed for the 
2008 Dress Rehearsal. To date, the Bureau has tested two models of the 
MCD--one during the 2004 Census Test and another during the 2006 Census 
Test. In January 2005, we reported that the MCD used during the 2004 
Census Test to collect nonresponse follow-up data experienced problems 
transmitting, and the mapping feature was slow. Consistent with our 
recommendations, the Bureau took steps to improve the dependability of 
transmissions and correct the speed of the mapping feature.[Footnote 4] 

Because of the critical role of contractors to help carry out the 2010 
Census, we conducted a review of major acquisitions for the 2010 
Census. [Footnote 5] In that report issued in May 2006, we highlighted 
the tight time frames the FDCA contractor has for developing and 
implementing systems to support the upcoming 2008 Dress Rehearsal and 
recommended that the Bureau ensure that all systems are fully 
functional and ready to be assessed in time for the Dress Rehearsal. In 
addition, on March 1, 2006, we testified on the status of the FDCA 
project. [Footnote 6] In that testimony, we discussed the need for the 
Bureau to: 

* validate and approve a baseline set of operational requirements for 
the FDCA contract, because if not, the FDCA project would be at risk of 
having changes to requirements, potentially affecting its ambitious 
development and implementation schedule; 

* implement an effective risk management process that identifies, 
prioritizes, and tracks project risks; and: 

* select detailed performance measures for tracking the contractor's 
work. 

In response to our work, the Bureau stated that they plan to complete 
these activities as soon as possible. 

Uncertainties Surround Completion of Ongoing MAF/TIGER Modernization 
Research: 

While the Bureau's MAF/TIGER modernization efforts have progressed in a 
number of areas, uncertainties and risks remain in dealing with address-
related problems that affected the 2000 Census. Currently it is not 
known whether ongoing research to resolve those problems will be 
completed in sufficient time to allow the Bureau to develop new 
methodologies and procedures for improving the MAF by June 2007--the 
Bureau's announced deadline for baselining all program requirements. 
One significant cause for this uncertainty is that some deadlines for 
completing research do not have firm dates, while other deadlines that 
have been set continue to slip. In addition, one major research effort 
using software to identify duplicate addresses (an estimated 1.4 
million duplicate addresses were removed during the 2000 Census) did 
not work any better at identifying true duplicates than what the Bureau 
already had in place and will not be used in 2010. As a result, 
duplicate addresses may still be a problem for the 2010 MAF, and to the 
extent they are not detected, can result in reduced accuracy and 
increased cost. 

During the 2000 Census, the Bureau encountered a number of problems 
with the MAF including (1) missed addresses, where the Bureau failed to 
include addresses in the MAF; (2) improperly deleted addresses, where 
the Bureau removed otherwise valid addresses from the MAF; (3) 
duplicate addresses, with two or more addresses for the same housing 
unit; and (4) geocoding errors, where addresses were improperly located 
on a census map.[Footnote 7] All of the errors affect the quality of 
census data. When detected, the errors can increase the cost of the 
census to the extent they result in rework. Moreover, these errors are 
associated with a variety of living arrangements and addresses, 
including small, multi-unit dwellings; dormitories, prisons, and other 
group living facilities, known collectively as "group quarters," as 
well as hidden housing units, such as converted basement apartments. As 
shown in table 1, to address those problems the Bureau has been 
conducting research and making some operational changes. 

Table 1: Status of Bureau Efforts to Resolve MAF/TIGER Issues: 

Types of errors identified in 2000: Missed addresses; 
Primary reasons for those errors: Some housing units are difficult to 
identify; 
Type of dwelling most likely affected: Small multi-unit structures; 
Actions taken by Bureau: Testing new method to identify clusters of 
small multi-units; 
Status of current effort: Testing to be completed by end of 2006. 

Types of errors identified in 2000: Improperly deleted addresses; 
Primary reasons for those errors: Varied; 
Type of dwelling most likely affected: Varied; 
Actions taken by Bureau: Tested new method for verifying the status of 
all housing units marked as deleted in 2006 address canvassing testing; 
Status of current effort: Evaluation was due January 2006 and that date 
has been moved to April 2006. The evaluation was not available at the 
time of this review. 

Types of errors identified in 2000: Duplicate addresses; 
Primary reasons for those errors: Redundancy and overlap in the address 
list building process; Address lists were created separately for group 
quarters and housing units. and some addresses were listed on both 
lists; 
Type of dwelling most likely affected: Housing unit with a city-style 
address. (e.g., 123 Main Street); Group quarters; 
Actions taken by Bureau: Tested address-matching software in 2004; 
Tested procedures during 2004 and 2006 Census Tests to integrate group 
quarters and housing unit address lists; 
Status of current effort: Results indicate matching software is not 
ready for 2010 Census. 2004 evaluation indicated progress is being made 
for integrating address lists. Evaluation of 2006 testing was due May 
2006 but was not available at the time of this review. 

Types of errors identified in 2000: Geocoding errors; 
Primary reasons for those errors: Maps not accurate; 
Type of dwelling most likely affected: Varied; 
Actions taken by Bureau: Collected GPS coordinates for housing units in 
the 2004 and 2006 tests; 
Hired contractor to update maps; 
Status of current effort: 2004 test results indicate that workers only 
used GPS 55 percent of the time. Evaluation was due January 2006 and 
that date has been moved to April 2006. The evaluation was not 
available at the time of this review. Contractor updating maps and will 
be finished in April 2008. 

Source: GAO analysis of Census Bureau data. 

[End of table] 

Research to Identify Hidden Housing Units Is Progressing, but 
Completion Date Is Uncertain: 

Although research to find hidden housing units holds promise for a more 
accurate census, whether the results will be delivered in time to be 
useful for the 2010 Census is uncertain. While Bureau officials do not 
have a firm date for completing this research, they do estimate it will 
be completed by the end of 2006. 

According to Bureau evaluations, approximately 1.4 million housing 
units were missed in the 2000 Census. Missed addresses often result 
when temporary census workers do not recognize that particular 
structures, such as tool sheds, are being used as residences. Addresses 
can also be missed when census workers fail to detect hidden housing 
units, such as basement apartments, within what appear to be single 
housing units. This is especially true for urban areas, where row 
houses have been converted into several different apartments. If an 
address is not in the MAF, its residents are less likely to be included 
in the census. 

In May 2003, Bureau staff met with the New York City Planning 
Department to discuss and observe the address problems associated with 
small multi-unit structures in Queens, New York. After the visit, the 
Bureau concluded that delivering questionnaires to small multi-unit 
structures was a problem that needed to be addressed. In response, the 
Bureau is using the MAF to identify urban areas, including Baltimore, 
an area west of Chicago, and counties in New Jersey, where small multi- 
unit dwellings exist, fitting the description of those that were 
missed.[Footnote 8] According to Bureau officials, to accurately 
identify and count these missed housing units, the Bureau would use 
update/enumerate procedures--where census workers update the address 
list and conduct interviews to collect census data--instead of using 
mailout/mailback procedures, where census forms are mailed to the 
housing units. Update/enumerate procedures are more labor-intensive and 
costly than mailout/mailback procedures. 

In reviewing the research plan on small multi-unit structures,we found 
no milestones for completing this research. Bureau officials could not 
provide a firm completion date, but estimated that the research would 
be completed by the end of 2006. Without clear milestones for 
completing this research and action plans based on research results, it 
is uncertain whether the Bureau will have sufficient time to develop a 
methodology for identifying all the problematic locations across the 
country where update/enumerate methodology would be necessary and to 
inform decision makers on the cost of converting these areas from 
mailout/mailback procedures to update/enumerate procedures. 

Research to Prevent Valid Addresses from Being Deleted Is Ongoing, but 
Completion Date Has Slipped: 

The Bureau has tested new procedures to validate whether an address 
initially marked "delete" should be removed from the address file. 
However, the results from that testing, due January 2006, were delayed 
until April 2006, and were not available at the time of this review. 

For the 2000 Census, the Bureau found that it had mistakenly deleted 
1.3 million existing housing units from the address file used to 
conduct the census. In some instances, this occurred when the Bureau 
deleted an address that the U.S. Postal Service had coded as a business 
address, although people were living at that address. According to a 
Bureau evaluation, when this happens, the Bureau relies on census 
workers to find and add back those units. Bureau officials stated that 
identifying residential housing units is difficult for some structures, 
such as apartments in businesses. 

The Bureau would also delete an address if no census form was returned 
from the unit and if two other census operations determined that the 
address should be deleted. A Bureau evaluation found that this process 
identified and removed 8.3 million nonexistent addresses; however, 
about 653,000[Footnote 9] of those addresses were valid and should not 
have been deleted. The evaluation does not provide an explanation for 
why these valid addresses were deleted or what could be done in the 
future to prevent valid addresses from being removed. Concerned that 
valid addresses were deleted, the Bureau, for the 2006 Census Test of 
address canvassing, tested a new follow-up quality check procedure 
designed to verify the status of all addresses that were identified as 
"delete" during the address canvassing operation. The 2000 Census did 
have a follow-up operation, but not one specifically for all deleted 
addresses during the canvassing operation. By building this quality 
control operation into the address canvassing operation, the Bureau 
hopes to prevent valid addresses from getting inadvertently deleted. An 
assessment report of address procedures that were tested in 2005 as 
part of the 2006 address canvassing operation was to be completed by 
January 2006. However, the deadline for this assessment slipped until 
the end of April 2006, and was not available at the time of this 
review. 

Research Efforts on Duplicate Addresses Have Mixed Results: 

The Bureau has taken actions to prevent duplicate addresses. However, 
one research effort to identify duplicates using software was found to 
be ineffective because approximately 10 percent of the time the 
software would incorrectly identify a valid address as a duplicate 
address, and as a result, this software will not be used in 2010. 
According to Commerce officials, it is their philosophy to favor the 
inclusion of addresses in the census process over the exclusion of 
addresses. Nevertheless, preventing duplicate addresses in the MAF 
saves the Bureau from having to make unnecessary and expensive follow- 
up visits to households already surveyed. Furthermore, preventing 
duplicate responses also enhances the accuracy of the data. 

Bureau studies initially estimated that during the 2000 Census, about 
2.4 million duplicate addresses existed in the MAF. The problem was so 
significant that in the summer of 2000, the Bureau initiated a special 
follow-up operation[Footnote 10] to identify and remove duplicate 
addresses. Research from this special operation confirmed that 1.4 
million addresses were duplicates, and the Bureau removed those 
addresses from the census. However, the operation was not able to 
determine with certainty whether the remaining 1 million addresses were 
duplicates. As a result, according to Commerce officials, the 1 million 
addresses were not removed from the census because those addresses were 
believed to be a combination of apartment mix-ups and misdelivery of 
questionnaires, and not duplicates. Had the Bureau identified these 1.4 
million housing units before nonresponse follow-up had occurred, it 
could have saved $39.7 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-up[Footnote 11]). Even after the special operation 
to remove duplicates was completed, the Bureau still estimated that 
approximately 0.7 million duplicates remained in the MAF in error. 

According to Bureau officials, duplicate addresses resulted from the 
multiple operations used to build the MAF. While the redundancy of 
having multiple address-building operations helps produce a more 
complete and accurate address list because more opportunities exist for 
an address to be added to the MAF, any variations in city-style 
addresses, which are addresses with house numbers and street names, 
could produce a duplicate entry. For example, the Postal Service, which 
is the source of many addresses in the MAF, might refer to an address 
in its database as 123 Waterway Point. A census worker in another 
address operation might record that address as 123 South Waterway 
Point. If not detected, two addresses would remain in the MAF for this 
single residence. To help resolve this problem, in 2004, the Bureau 
tested whether it could detect duplicate addresses in the MAF by using 
computerized matching software to link variations in street addresses. 
In test results, the Bureau found that 90 percent of the potential 
duplicates identified by the process of "probablistic matching" were 
actual duplicates, while 10 percent were valid addresses. Because the 
number of false duplicates was significantly high, the Bureau decided 
against incorporating this approach into its plans for 2010 and planned 
no further testing of the software. As a result of not being able to 
use this software, duplicate addresses may still be a problem for the 
2010 MAF, and duplicate addresses that are not detected can reduce 
accuracy and increase costs. 

At the same time, the Bureau has made some progress toward preventing 
duplicates. The Bureau is testing new methods to resolve difficulties 
in distinguishing group quarters (which include dormitories, prisons, 
group homes, and nursing homes) from housing units, such as single- 
family homes and apartments. In the 2000 Census, the Bureau used 
different operations and compiled separate address lists for group 
quarters and housing units. Group quarters are sometimes difficult for 
census workers to identify because they often look the same as 
conventional housing units (see fig. 2). As a result, these homes were 
sometimes counted twice during the 2000 Census--once as a group quarter 
and once as a housing unit. 

Figure 2: Group Homes Can Resemble Housing Units: 

[See PDF for image] 

[End of figure] 

One approach to help prevent duplicates that the Bureau tested during 
the 2004 and 2006 Census Tests is integrating the two address lists and 
then verifying potential group quarters on that list. Evaluation 
results from the 2004 testing showed progress was being made for 
integrating the address lists. The operational assessment report on the 
2006 group quarters testing validation/advance visit operation that 
occurred in 2005, as a part of the address canvassing operation for the 
2006 Census Test, was expected by May 30, 2006, and was not available 
at the time of this review. 

Mixed Progress Is Being Made to Properly Identify and Locate Housing 
Units on TIGER Maps: 

The Bureau is using a contractor to update its TIGER maps and intends 
to use GPS technology to locate every housing unit across the country 
precisely. Collectively, these two efforts are designed to avoid the 
geocoding errors of the 2000 Census, when residences were sometimes 
counted in the wrong census block. However, progress can be hindered if 
technical problems associated with the GPS continue. 

Bureau evaluations estimated that in 2000, of the nation's 
approximately 116 million housing units, 5.6 million (about 4.8 
percent) housing units in the country were counted in the wrong 
locations. Resolving geocoding errors will be important, as census data 
are used to redraw congressional lines and allocate federal assistance 
and state funding. For example, in June 2005, we reported that Soledad, 
California, lost more then $140,000 in state revenue when a geocoding 
error caused over 11,000 Soledad residents to be miscounted in two 
nearby cities.[Footnote 12] 

Geocoding errors are partly attributable to inaccuracies in the TIGER 
maps that census workers use to verify the locations of residences. As 
shown in figure 3, roads and other features on TIGER maps did not 
always reflect their true geographic locations. 

Figure 3: TIGER Map Overlay of an Aerial Photograph: 

[See PDF for image] 

[End of figure] 

To help improve TIGER maps, in June 2002, the Bureau awarded an 8-year, 
$200 million contract to correct in TIGER the location of every street, 
boundary, and other map feature so that they are aligned with their 
true geographic locations, among other contractual tasks. This work is 
to be completed on a county-by-county schedule. According to Bureau 
officials, as of March 2006, nearly 1,700 county maps have been 
completed, with about another 1,600 to be completed by April 2008. 

In conjunction with updating TIGER, the Bureau, as part of its 2010 
address canvassing operations, plans to have census workers capture the 
exact location of every structure on the address list by using GPS 
receivers. This approach has the potential to resolve the cause of many 
geocoding errors; however, as we discuss later in this report, when 
this operation was tested as part of the 2006 Census Test, the GPS 
receiver did not always operate properly, leaving some housing units 
without a GPS coordinate to determine their locations. As part of the 
address canvassing operational assessment report, the Bureau will 
provide the number and type of map spots collected (GPS, manual, or 
attached multi-unit). This report, initially due in January 2006, has 
been delayed and was not available at the time of our review. 

Testing GPS coordinates was a part of the 2004 Census Test, and 
evaluations showed that workers only used the GPS receiver to capture 
the location of housing units 55 percent of time. The evaluation, 
however, did not address why census workers did not use the GPS 
receiver. 

Emerging Issues Related to Overlapping and Compressed Schedules Pose a 
Risk to MAF/TIGER Modernization Efforts: 

As the Bureau has planned for the 2010 Census, issues surrounding the 
schedule of address activities have emerged and have not been fully 
addressed. One key challenge in conducting the 2010 Census is the 
Bureau's ability to keep the myriad of census activities on track amid 
tight and overlapping schedules for updating addresses and maps. For 
example, in planning the various 2010 address list activities, Bureau 
officials estimate that TIGER maps for 600 to 700 counties (out of 
3,232 counties in the United States) will not be updated in time to be 
part of the local update of census addresses (LUCA)--a program through 
which the Bureau gives local, state, and tribal government officials 
the opportunity to review and suggest corrections to the address lists 
and maps for their jurisdictions.[Footnote 13] LUCA is to begin in 
August 2007, when, according to the current schedule, the Bureau will 
still have to update 368 counties in 2008 alone. Because all updates 
will not have been completed, some counties will not have the most 
current maps to review, but instead will be given the most recent maps 
the Bureau has available. According to Bureau officials, some maps have 
been updated for the American Community Survey, but others have not 
been updated since the 2000 Census. LUCA participation is important 
because local knowledge contributes to a more complete and accurate 
address file. Not having the most current TIGER maps could affect the 
quality of a local government's review. The Bureau is aware of the 
overlapping schedules, but officials stated that they need to start 
LUCA in 2007 in order to complete the operation in time for address 
canvassing--an operation where census workers walk every street in the 
country to verify addresses and update maps. Further, Commerce 
officials stated that the primary focus of the LUCA program is to 
review and update the address list and not to review and update maps; 
therefore, not having the improved maps should not affect the ability 
of LUCA participants to add or make corrections to the census address 
list. We, however, believe that improved maps would help LUCA 
participants to provide more accurate address data. 

The census schedule will be a challenge for address canvassing in 2010. 
The Bureau has allotted 6 weeks for census workers to verify the 
nation's inventory of approximately 116 million housing units. This 
translates into a completion rate of over 2.75 million housing units 
every day. The challenge in maintaining this schedule can be seen in 
the fact that for the 2000 Census, the Bureau took 18 weeks just to 
canvass "city-style" address areas, which are localities where the U.S. 
Postal Service uses house-number and street-name addresses for most 
mail delivery. However, a Bureau official could not explain why the 
schedule had been shortened by 12 weeks, compared to the 2000 Census. 

Although Bureau officials agreed that more time will be needed to 
conduct the address canvassing operation, especially in the northern 
sections of the country where bad weather can hinder those operations, 
they have not reevaluated the schedule. A Bureau official stated that 
the Bureau would need to assess staffing levels to ensure it will be 
able to meet workload demands. Meeting the demands of the shortened 
time frame for completing address canvassing is a concern because the 
workload for address canvassing has significantly expanded from 
including only urban areas in 2000 to including the entire country for 
2010. Furthermore, in the summer of 2005, when address canvassing was 
conducted for the 2006 test, the Bureau was unable to finish in 6 weeks 
because of problems with the new MCD and GPS technology. In its 
comments to a draft of this report, Commerce officials said it would 
work to expand the address canvassing schedule to ensure that it can be 
done without having a negative impact on other critical decennial 
activities. 

Reliability of MCD to Conduct Address Canvassing Activities is Unknown: 

The Bureau's ability to collect and transmit address and mapping data 
using the MCD is not known. The performance of these devices is crucial 
to the accurate, timely, and cost-effective completion of address 
listing, nonresponse follow-up, and coverage measurement activities. 
During 2006 testing, the MCD used to collect address and map data was 
slow and locked up frequently. As a result, the Bureau was unable to 
complete address canvassing, even with a 10-day extension. Also, some 
census workers were not always able to get GPS signals for collecting 
coordinates for housing units. Bureau officials have acknowledged that 
the MCD's performance is an issue but believe that a new version of the 
MCD, to be developed under the Field Data Collection Automation (FDCA) 
contract awarded on March 30, 2006, will be reliable and functional. 
However, because the 2008 Dress Rehearsal will be the first time this 
new MCD will be tested under census-like conditions, it is uncertain 
how effective that MCD will be, and if problems do emerge, little time 
will be left for the contractor to develop, test, and incorporate any 
refinements. Moreover, if after the Dress Rehearsal the MCD is found to 
be unreliable, the Bureau could be faced with the remote, but daunting 
possibility of having to revert to the costly paper-based census used 
in 2000. 

Bureau Is Unable to Complete Address Canvassing Operation Because of 
Technical Difficulties with the MCD: 

During the address canvassing operation, the technical problems with 
the MCDs were so significant that the operation did not finish as 
scheduled.[Footnote 14] The 6 week operation was expected to run 
through September 2, 2005, but had to be extended by 10 days (through 
Sept. 12, 2005). However, the Bureau was still unable to finish the 
operation, leaving six assignment areas in Travis County, Texas and 
four assignment areas at the Cheyenne River Reservation, South Dakota 
not canvassed. 

To conduct address canvassing, each MCD was loaded with address 
information and maps and was also equipped with GPS. Census workers 
were trained to locate every structure in their assignment area, as 
well as to compare the locations of housing units to address and map 
data on the MCD and update the data accordingly. They also were 
instructed to capture each housing unit's GPS coordinates. However, 
workers we observed and interviewed had problems updating address and 
map data as well as collecting GPS coordinates, largely because the 
device's software and GPS receiver were unstable. For example, we 
observed census workers unable to complete their planned assignments 
for the day because it took too long to complete address and map 
updates, as the device was slow to pull up and exit address registers, 
accept the data entered by the worker, and link a map spot to addresses 
for multi-unit structures. Furthermore, the devices would often lock 
up, requiring workers to reboot them. 

Census workers also experienced problems with the GPS receiver acquired 
by the Bureau. Some workers had problems getting a signal, but even 
when a signal was available, the GPS receiver was slow to locate 
assignment areas and provide coordinates for map spots. Bureau 
officials were not certain why the Bureau's equipment was unreliable, 
but provided several possible explanations: (1) the software, hardware, 
or both did not function properly, (2) GPS units were not correctly 
inserted into the device, and (3) too few satellites were available for 
capturing coordinates. Given the importance of GPS to collecting 
precise coordinates for housing units, it will be important for the 
Bureau to understand and correct the source of the problems that 
affected the reliability of the GPS. 

Going into address canvassing, the Bureau was aware that the MCDs had 
software problems and delayed the address canvassing operation by a 
month to try to resolve them. The Bureau was unable to resolve the 
problems, but wanted to test the feasibility of the MCD and decided to 
go forward with the operation with the goal of learning as much as 
possible. For the 2008 Dress Rehearsal, the Bureau plans to test a new 
MCD that is being developed under the FDCA contract. However, less than 
a year remains for the contractor to develop the MCD that will be used 
in April 2007 for the canvassing operation of the 2008 Dress Rehearsal. 
In a May 2006 report,[Footnote 15] we reported on the tight time frames 
to develop the MCD and recommended that systems being developed or 
provided by contractors for the 2010 Census--including the MCD--be 
fully functional and ready to be assessed as part of the 2008 Dress 
Rehearsal. In commenting on a draft of this report, Commerce noted that 
the Bureau designed the FDCA acquisitions strategy to reduce risks 
related to cost, schedule and performance, stating that the Bureau 
required offerors to develop and demonstrate a working prototype for 
address canvassing. Nevertheless, because the previous two MCD models 
had performance problems, the introduction of a new MCD adds another 
level of risk to the success of the 2010 Census. 

Bureau Does Not Have a Specific Plan for Updating MAF/TIGER in the 
Aftermath of Hurricanes Katrina and Rita: 

The Bureau does not have a plan to update the MAF/TIGER for areas 
affected by hurricanes Katrina and Rita. On August 29, 2005, Hurricane 
Katrina devastated the coastal communities of Louisiana, Mississippi, 
and Alabama. A few weeks later, Hurricane Rita plowed through the 
border areas of Texas and Louisiana. Damage was widespread. In the wake 
of Katrina, for example, the Red Cross estimated that nearly 525,000 
people were displaced. Their homes were declared uninhabitable, and 
streets, bridges, and other landmarks were destroyed. Approximately 
90,000 square miles were affected overall and, as shown in figure 4, 
entire communities were obliterated. 

Figure 4: Devastation in the Lower Ninth Ward in New Orleans, 
Louisiana: 

[See PDF for image] 

[End of figure] 

The task of updating MAF/TIGER for 2010 to reflect these changes will 
be a formidable one, as much has changed since the 2000 Census. For the 
2010 Census, locating housing units and the people who reside in them 
will be critical to counting the population of places hit by the 
hurricanes, especially since it is estimated that hundreds of thousands 
of people have--either temporarily or permanently-- migrated to other 
areas of the country. To ensure an accurate count, it will be important 
for the Bureau to have accurate maps and an updated address file for 
the 2010 Census in those areas affected by hurricanes Katrina and Rita. 

Bureau officials do not believe a specific plan is needed to update the 
address and map files for those areas affected by hurricanes Katrina 
and Rita. Although Census Day is still several years away, preliminary 
activities, such as operations for building the MAF, have to occur 
sooner. Consequently, a key question is whether the Bureau's existing 
operations are adequate for capturing the dramatic changes to roads and 
other geographic features caused by the hurricanes, or whether the 
Bureau needs to develop enhanced or additional procedures before August 
2007 when LUCA is scheduled to begin. For example, new housing and 
street construction in the areas affected by the hurricanes could 
require more frequent updates of the Bureau's address file and maps. 
Also, local governments' participation in LUCA might be affected 
because of the loss of key personnel, information systems, or records 
needed to verify the Bureau's address lists and maps. Further, the 
Bureau has not identified local partners with whom it can monitor this 
situation. 

The Bureau's short-term strategy for dealing with the effect of the 
hurricanes on MAF/TIGER is to see who returns and whether communities 
decide to rebuild. Bureau officials stated that by 2009, as census 
workers prepare to go out in the field for address canvassing for the 
2010 Census, residents will have decided whether to return to the 
region. The Bureau believes that by then it will be in a better 
position to add or delete addresses for areas in the Gulf region 
affected by the hurricanes. However, Bureau officials could not provide 
us with information on the basis of their conclusion that by 2009, most 
affected persons will have made final decisions about whether they are 
returning to the region. This approach may not be adequate, given the 
magnitude of the area, population, and infrastructure affected. 
Therefore, it would be prudent for the Bureau to begin assessing 
whether new procedures will be necessary, determining whether 
additional resources may be needed, and identifying whether local 
partners will be available to assist the Bureau in its effort to update 
address and map data, as well as in other census-taking activities. In 
its comments on a draft of this report, Commerce officials stated that 
there was a team working on how to reflect the impact of the hurricanes 
in the MAF and that they were aware of the sensitive nature of working 
with local officials on using data that had not been updated since the 
catastrophe. 

Securing a complete count, a difficult task under normal circumstances, 
could face additional hurdles along the Gulf Coast, in large part 
because the baseline the Bureau will be working with-- streets, 
housing, and the population itself--will be in flux for some time to 
come. According to Bureau officials, different parts of the agency work 
on hurricane-related issues at different times, but no formal body has 
been created to deal with the hurricanes' impact on the 2010 Census. 

Conclusions: 

The success of the 2010 Census relies on an accurate and complete MAF, 
and the Bureau has taken steps to improve the MAF. For example, many of 
the problems identified in the 2000 Census are being addressed through 
sequential address list building, the collection of GPS coordinates, 
and the verifications of deleted addresses. However, several key 
challenges and sources of uncertainty remain. The management of some of 
the Bureau's research efforts to resolve problems from the 2000 Census 
are negatively affected by a lack of specific end dates for that 
research or because those end dates have slipped. Also, one research 
effort to prevent duplicate addresses was found to be ineffective and 
was abandoned altogether. Time to complete research and take the 
appropriate resulting action is of the essence, as the Bureau has 
announced that all design features should be baselined by June 2007. If 
long-standing problems are not resolved, the address file could 
experience the same problems with missed and incorrectly included 
housing units as it did in the 2000 Census. 

The Bureau must also manage the planning and development of the census 
amid tight and overlapping schedules. In our view, changing milestones 
to complete MAF research, the Bureau's tight development schedule for 
the MCD, and the interdependence of the various address activities 
could affect the Bureau's ability to develop a fully functional set of 
address-building operations that can be tested along with other census 
operations during the 2008 Dress Rehearsal--the Bureau's last 
opportunity to assess MAF/TIGER under near census-like conditions. If 
the MCDs do not function as planned in the Dress Rehearsal, little time 
will remain for the Bureau to develop, test, and incorporate any 
refinements. This uncertainty places the accuracy and completeness of 
data collected using the MCD at risk. 

Because the MCD has not yet been developed, it will be important for 
the Bureau to closely monitor the contractor's progress for developing 
the MCD. In May 2006, we reported on the tight time frames to develop 
the MCD and recommended that systems being developed or provided by 
contractors for the 2010 Census--including the MCD--be fully functional 
and ready to be assessed as part of the 2008 Dress Rehearsal. However, 
if after the Dress Rehearsal the MCD is found to be unreliable, the 
Bureau could be faced with the remote but daunting possibility of 
having to revert to the costly paper-based census used in 2000. 

Finally, the destruction and chaos caused by hurricanes Katrina and 
Rita underscore the nation's vulnerability to all types of hazards and 
highlight how important it is for government agencies to consider 
emergency preparedness and continuity of operations as part of their 
planning. However, the immediate concern for the 2010 Census is that 
the Bureau has no plan for how it will successfully update MAF/TIGER in 
the affected hurricane zone. If problems updating the address file and 
maps do occur, the census count in those areas affected by hurricanes 
Katrina and Rita could be inaccurate or incomplete. 

In conversations with Bureau officials, it became apparent to us that 
they are keenly aware of the existing time constraints and challenges 
detailed above. However, the Bureau had not developed risk mitigation 
plans to address these challenges. Our recommendations, therefore, are 
intended to make transparent for Bureau managers and congressional 
decision makers how those challenges can and should be addressed. At a 
minimum, the Bureau should have a risk-based mitigation plan in place 
that includes specific dates for completing research on the address 
file and an approach for exploring the difficulties that the Bureau may 
face updating MAF/TIGER along the Gulf Coast. Because time is running 
short, it is imperative that the Bureau continue to stay focused on 
identifying and resolving problems to ensure that the most accurate and 
complete address file and maps are produced for the 2010 Census. 

Recommendations for Executive Action: 

To mitigate potential risks facing the Bureau as it plans for 2010 and 
to ensure a more complete and accurate address file for the 2010 
Census, we recommend that the Secretary of Commerce direct the U.S. 
Census Bureau to take the following three actions: 

* Establish firm deadlines to complete research, testing, and 
evaluations of the MAF to prevent missed, deleted, and duplicate 
addresses, as well as map errors, and develop an action plan that will 
allow sufficient time for the Bureau to revise or establish 
methodologies and procedures for building the 2010 MAF. 

* Reevaluate the 2010 address canvassing schedule in areas affected by 
bad weather, as well as staffing levels, to ensure that the status of 
all housing units are accurately verified throughout the entire 
country. 

* Develop a plan, prior to the start of LUCA in August 2007, that will 
assess whether new procedures, additional resources, or local 
partnerships may be required to update the MAF/TIGER databases for 
areas along the Gulf Coast affected by hurricanes Katrina and Rita. 

Agency Comments and Our Evaluation: 

On June 2, 2006, the Department of Commerce forwarded written comments 
from the Bureau on a draft of this report. The Bureau agreed with each 
of our three recommendations and also noted actions it was taking to 
address the recommendations. The Bureau's comments also included some 
technical corrections and suggestions where additional context was 
needed, and we revised the report to reflect these comments as 
appropriate. The comments are reprinted in their entirety in appendix 
II. 

In responding to the first recommendation to develop an action plan 
that will allow sufficient time to revise or establish methodologies or 
procedures for building the 2010 MAF, the Bureau stated that it would 
revise its action plan to reflect final milestones for research to be 
completed in time for the 2010 Census. Regarding the second 
recommendation to reevaluate the 2010 address canvassing schedule, as 
well as its staffing, the Bureau stated that this will be a challenge 
but that it is committed towards developing a new schedule. Finally, in 
addressing our third recommendation to develop a plan to assess whether 
new procedures, additional resources or local partnerships may be 
required to update the MAF/TIGER databases for areas affected by 
hurricanes Katrina and Rita, the Bureau stated that it was working on a 
proposal for additional work in the areas affected by hurricanes 
Katrina and Rita. The Bureau also noted that conducting additional work 
will be subject to obtaining funding. 

We are sending copies of this report to other interested congressional 
committees, the Secretary of Commerce, and the Director of the U.S. 
Census Bureau. Copies will be made available to others upon request. 
This report will also be available at no charge on GAO's Web site at 
http://www.gao.gov. 

Please contact me at (202) 512-6806 or farrellb@gao.gov if you have any 
questions about this report. Contact points for our Office of 
Congressional Relations and Public Affairs may be found on the last 
page of this report. Key contributors to this report are listed in 
appendix III. 

Signed by: 

Brenda S. Farrell, Acting Director: 
Strategic Issues: 

[End of section] 

Appendix I: Scope and Methodology: 

To determine the extent to which the Bureau's MAF/TIGER modernization 
efforts are addressing problems experienced during the 2000 Census, we 
reviewed pertinent documents, including evaluations of the 2000 Census 
conducted by GAO, the Bureau, the National Academy of Sciences, and the 
Department of Commerce's Office of Inspector General. To determine the 
status of those efforts, we also interviewed cognizant Bureau officials 
in the Geography Division and Decennial Management Division responsible 
for implementing the modernization efforts. To assess the extent to 
which past problems were being addressed, we compared the Bureau's 
current efforts--including, but not limited to, the 2010 LUCA draft 
plan, 2004 and 2006 test plans, other research efforts, and TIGER 
improvement documents--to problems identified in evaluations of the 
2000 Census conducted by GAO, the Bureau, the National Academy of 
Sciences, and the Department of Commerce's Office of Inspector General. 

We reviewed the MAF/TIGER contract that was awarded in June 2002 to 
update the street and geographic features for the TIGER maps, as well 
as monthly earned-value management system (EVMS) cost and performance 
reports, to determine whether the deliverable schedule for the contract 
was on time and on budget. We did not independently verify the accuracy 
of the data contained in the EVMS cost and performance reports, but we 
did obtain a certification from the contractor that its EVMS was 
adequate to provide timely and accurate data from the Defense Logistics 
Agency. 

To determine the extent to which the Bureau is managing emerging MAF/ 
TIGER issues, we focused on planning documents that described proposed 
2010 plans. Specific documents we reviewed included the 2010 LUCA draft 
proposal, 2010 Census decision memorandums, and Bureau papers from 
National Academy of Sciences and Census Advisory Committee meetings. We 
also reviewed and compared the timeline for conducting 2000 Census 
address operations to the proposed plan for conducting 2010 Census 
address operations. We interviewed officials from the Bureau's 
Geography Division and the Decennial Management Division on the 2010 
plans, 2010 time lines, current status of work, and areas of concern.  

To assess the extent to which the Bureau is able to collect and 
transmit address data using new, GPS-enabled mobile computing devices, 
we made site visits to census offices on the Cheyenne River 
Reservation, South Dakota, and in Travis County, Texas, when we 
observed the address canvassing operation conducted during the summer 
of 2005 as part of the 2006 Census Test. During these site visits, we 
also interviewed local and regional census managers and staff, observed 
address data collection activities using the MCD, and attended census 
worker training sessions. We observed and interviewed a total of 38 
census workers (16 in South Dakota and 22 in Texas) about the address 
canvassing operation and the use of the MCD to collect address data. 
However, the results of these observations are not necessarily 
representative of the larger universe of census workers. After our 
visits, we discussed our observations with the Bureau's Technology 
Management Office, Field Division, Geography Division, and Decennial 
Management Division. 

Finally, to determine the extent to which the Bureau has a plan to 
update the address file and maps in areas impacted by hurricanes 
Katrina and Rita, we interviewed Bureau top management officials. 
Specifically, we discussed whether the Bureau had taken any steps to 
assess the difficulties it may encounter as it attempts to update the 
address file and maps and count persons affected by hurricanes Katrina 
and Rita. We conducted our work from June 2005 through April 2006 in 
accordance with generally accepted government auditing standards. 

[End of section] 

Appendix II: Comments from the Department of Commerce: 

United States Department Of Commerce: 
The Under Secretary for Economic Affairs: 
Washington, D.C. 20230: 

June 2, 2006

Ms. Brenda S. Farrell: 
Acting Director: 
Strategic Issues: 
United States Government Accountability Office: 
Washington, DC 20548: 

Dear Ms. Farrell: 

The U.S. Department of Commerce appreciates the opportunity to comment 
on the United States Government Accountability Office's draft report 
entitled 2010 Census: Census Bureau Needs to Take Prompt Actions to 
Resolve Long-standing and Emerging Address and Mapping Challenges (GAO- 
06-272). I enclose the Department's comments on this report. 

Sincerely, 

Signed by: 

Elizabeth R. Anderson: 

Enclosure: 

Ms. Brenda S. Farrell: 

Commerce Control Number 06-001709: 
Census Control Number 49503: 
DMD:TAngueira:PWhite:5/19/06: 
CQAS:Review:lmh:5/19/06;tdw:5/22/06;kem:5/22/06: 
CQAS:Final:nth:5/24/06: 

cc: 

ES, US/EA, CQAS (3), A. Moxam, T. Johnson, M. Raines, N. Gordon, R. 
Swartz, J. Waite, J. Taylor, T. Mesenbourg, H. Hogan: 

U.S. Department of Commerce Comments on the United States Government 
Accountability Office Draft Report Entitled 2010 Census: 
Census Bureau Needs to Take Prompt Actions to Resolve Long-standing and 
Emerging Address and Mapping Challenges (GAO-06-272) May 2006: 

The U.S. Census Bureau generally agrees with the recommendations in 
this report, but has some concerns and comments about various 
statements and conclusions. 

Regarding the recommendations that begin on page 32: 

GAO Recommendation: "Establish firm deadlines to complete research, 
testing, and evaluations of the MAF to prevent missed, deleted, and 
duplicate addresses, as well as map errors, and develop an action plan 
which will allow sufficient time for the Bureau to revise or establish 
methodologies and procedures for building the 2010 MAF." 

Census Bureau Response: We agree with this recommendation. Throughout 
this testing cycle, we have maintained a Coverage Improvement Action 
Plan. In this plan are milestones for conducting research related to 
improving coverage of housing units in the census. Although this action 
plan may not be at the level of detail that GAO is suggesting, it has 
served as well in this decade. We have updated it periodically to 
reflect results from tests and changes in priorities. Now that the 
research is winding down, we agree that we should revise this action 
plan again to reflect final milestones for any research we plan to 
complete in time for the 2010 Census. 

GAO Recommendation: "Reevaluate the 2010 address canvassing schedule in 
areas affected by bad weather, as well as its staffing, to ensure that 
the status of all housing units are accurately verified throughout the 
entire [seems to be a word missing . probably `country']." 

Census Bureau Response: We agree with this recommendation. We have 
started working on the scheduling challenges associated with extending 
the Address Canvassing operation to allow for waves. We are not at the 
point where we can commit to a new schedule, but we are committed to 
working toward a new schedule. 

GAO Recommendation: "Develop a plan, prior to the start of LUCA in 
August 2007, that will assess whether new procedures, additional 
resources, or local partnerships may be required to update the MAF/ 
TIGER databases for areas affected by hurricanes Katrina and Rita." 

Census Bureau Response: We agree with this recommendation. We should 
have a proposal for any additional work in the areas affected by 
Hurricanes Katrina and Rita in the near future. Of course, doing 
additional work will be subject to obtaining additional funding. 

Our specific comments and concerns about the report are as follows: 

Highlights page - Here, and throughout this report, there are 
statements about testing the Hand Held Computers (HHCs) in 2005. In 
order to avoid confusion with our 2005 National Census Test (which did 
not include any address listing or field interviewing), these 
statements would be clearer if they referred to this testing as part of 
the Address Canvassing operation for the 2006 Census Test. 

Page 2 - The approximate housing unit count in 2000 was 116 million. 
Currently, we estimate there will be close to 130 million housing units 
to be enumerated in the 2010 Census. 

Page 4 - The research project regarding small multiunit structures is 
about avoiding missed units, but is also about duplicate enumerations 
caused by confusion during mail delivery, during the non-response 
follow-up (NRFU), or during the coverage follow-up operation. If unit 
designations are unclear, the Census Bureau will not always be able to 
determine which questionnaire was delivered to a particular unit, so it 
will not be certain which units to visit during NRFU. This sort of 
situation can result in some units being enumerated twice and some not 
at all. 

Page 4 - Concerning the issue of our study of software to identify 
duplicate addresses, we do not think it is accurate to conclude it did 
not work. Rather, we found that different types of software did not 
work any better at identifying true duplicates than what we already 
have in place. Because of the risks of excluding valid addresses by 
eliminating duplicates, we will always take a very careful approach to 
deleting records from the address list based on matching. While it is 
true that duplicate addresses may still be a problem for 2010, a number 
of other improvements are tackling that same issue, such as sequential 
address list operations, Global Positioning System (GPS) coordinates on 
structures and maps, and integration of the address lists for housing 
units and group quarters. Additionally, the 1.4 million duplicate 
addresses deleted from Census 2000 that are referred to did not result 
entirely from the lack of this type of address matching. 

Page 5 - In paragraph 1, the report states that the Census Bureau 
completed Address Canvassing in 18 weeks in 2000, but on the 
"Highlights" page up front, it states 8 weeks (which may just be a 
typo). Overall, it is not clear whether GAO is comparing the 2010 
Census Address Canvassing operation to just the Census 2000 Block 
Canvassing operation which lasted 18 weeks, or to the Census 2000 
Address Listing operation which lasted 18 weeks, or to both (which 
lasted 36 weeks combined). 

Page 6 - We will not be identifying housing units as vacant, occupied, 
or under construction during our 2009 activities. We only will be 
updating the address list by adding or deleting addresses. 

Page 6 - We believe the discussion about updating the maps and lists in 
the Katrina/Rita areas is not clear. First of all, we need to create 
contingency plans for any disaster that impacts our list at any time, 
not just this one. Our research to date on the Katrina/Rita areas has 
shown that the address list has been impacted much more severely than 
the maps. We believe our Address Canvassing operation for the 2010 
Census can deal with the impacts on the address file and maps, though 
of course we need to be prepared for more extensive updating in these 
areas. In addition, we are aware of the sensitive nature of working 
with local officials using data that have not been updated since the 
catastrophe. We are in the process of addressing this problem on a 
number of levels. 

Page 9 - The United States Postal Service (LISPS) discussion in this 
report should refer specifically to their Delivery Sequence File (DSF) 
because there are many other data products that the LISPS provides to 
us. We received and processed DSFs through April 2000 for Census 2000, 
including a special extract in February 2000 that included only added 
units. Later DSF adds were enumerated as part of the Coverage 
Improvement Followup operation (CIFU) conducted from July 6 through 
August 22, 2000, so it should be listed as well. In addition, the list 
should include List/Enumerate areas and remote Alaska. The Local Update 
of Census Addresses (LUCA) time line includes the dates for the 
separate additional effort to get newly constructed units added to the 
list (and enumerated in CIFU), but New Construction was a separate 
operation that should be mentioned as such in this section if the time 
line is going to include its dates. 

Page 10 - The report sometimes refers to the Master Address File (MAF) 
and Decennial Master Address File (DMAF) as though they are 
interchangeable names for the same thing, and we believe this is 
confusing some of the issues you are discussing. Nothing was deleted 
from the MAF, and the MAF contains units that were not on the DMAF 
(which is an extract from the MAF created for conducting the census). 
When the report discusses estimates of misclassifications, this should 
refer to the DMAF, not the MAR: 

Page 16 - The example given for how the Census Bureau could have 
mistakenly deleted an address would better be described as an example 
of how any address could be on the MAF but missing from the census. 
That is, if the LISPS DSF listed a unit as commercial (nonresidential), 
and no contradictory information was received from any of our other 
address list development efforts, the address remained on the MAF but 
was never added to the DMAF. It is not accurate to say that the Census 
Bureau deleted this type of unit. 

Page 17 - The report implies that we did not verify deletes in 2000 and 
that the 2006 delete verification is new. This is not accurate. For 
Census 2000, delete verification was conducted during a separate 
operation which followed the operation that identified the potential 
delete. For example, addresses classified as delete during Block 
Canvassing were verified during the: 

1998 LUCA field verification. Similarly, addresses identified as 
deletes during NRFU were verified during a separate operation (CIFU), 
which did not begin until NRFU was completed. With the use of 
automation, we now are able to identify the deletes quickly and send 
them back for verification immediately. We believe that doing this 
concurrently with the operation (in effect, as a Quality Control 
operation) will improve effectiveness and quality because the 
verification will happen closer in time to Census Day. 

Pages 17 to 20 - The automated detection of duplicates is largely 
dependent on the amount and quality of the information being matched. 
Our philosophy has been to "favor" the inclusion of addresses in the 
census process over the exclusion of addresses. The "unplanned" housing 
unit unduplication operation that was done in Census 2000 was 
successful at confirming additional duplicates (1.4 million) in large 
part because it had the added benefit of decennial person information 
to consider during the match. For example, an automated process may not 
be able to absolutely determine if 123 Waterway Point is different from 
123 South Waterway Point, but when we include as part of the matching 
process that a person named John Doe, aged 24, lived at both addresses, 
we can be more confident in calling it a duplicate address. The report 
indicates we would save money if this unduplication were done prior to 
NRFU, but the benefit of the person information is not available before 
NRFU and, therefore, we could not get similar results if this operation 
were conducted prior to NRFU. 

Page 18 - Partly in response to the lessons we learned about 
duplication during Census 2000, we are designing our address list 
development operations to be sequential in 2010. Also, this paragraph 
fails to mention that we maintained one million cases in the census 
universe because we believed them to be apartment mix-ups/misdelivery 
cases (that should not, therefore, be removed from the census), rather 
than just cases where we were uncertain of their existence. 

Pages 23 to 24 - The draft report expresses a concern about some areas 
not having been completed through the MAF/TIGER Accuracy Improvement 
Program (MTAIP, which focuses primarily on improving the positional 
accuracy of the roads in TIGER) prior to the start of the LUCA program: 
"Because all updates will not have been completed, some counties will 
not have the most current maps to review but instead will be given the 
most recent maps the Bureau has available.... Not having the most 
current TIGER maps could affect the quality of a local government's 
review." 

The primary focus of the LUCA program is the review and update of the 
address list, not the review and update of the associated maps. While 
it is true that the LUCA maps for some communities will not reflect the 
road positional accuracy improvements and road updates that MTAIP will 
yield, this in no way interferes with the ability of the affected LUCA 
participants to add to and correct the census address list. The MTAIP, 
from its inception, was planned to be completed before the Address 
Canvassing operation and is on schedule to meet that objective. 

Pages 24 to 27 - The statements about the problems we had with the HHC 
for Address Canvassing do not provide sufficient context. Over a year 
ago, we stated publicly (on more than one occasion) that we were having 
difficulty developing the automated instrument and applications for 
address canvassing. At that time, we also made it clear that we would 
soon have to decide if enough could be learned by going forward-even 
with an imperfect instrument-or whether we would have to cancel the 
test. We later decided to go forward with a goal of learning as much as 
we could and shared that decision publicly. Therefore, while many of 
the problems and issues you raise in your report certainly did arise 
during Address Canvassing, we are disappointed that the report provides 
none of this background or context for the general reader of the 
report. 

Page 24 (and on Page 5) - The Census Bureau agrees that the six-week 
period of time initially allotted for conducting Address Canvassing in 
the 2010 Census was very ambitious. This issue has been under 
discussion for quite a while, and those discussions continue. As noted 
in this report, some of the reasons for lengthening the operation are 
related to staffing and weather concerns, but other reasons relate to 
the management effort in the regional offices that would be necessary 
to successfully manage and complete an operation of this size in six 
weeks. 

For example, because Address Canvassing is the first field operation of 
the decennial census, an expanded period of time for conducting the 
operation would give the regional offices a critically needed 
opportunity to train staff, such as area managers and Local Census 
Office managers, on a flow basis and improve on what was done in 
previous waves, rather than trying to accomplish this task all at once 
in a short time frame. 

We are committed to working on expanding the length of the Address 
Canvassing operation and are currently reviewing the schedule to ensure 
that this can be done without a negative impact on other critical 
decennial activities. The proposal currently being considered is to 
conduct Address Canvassing in three six-week waves, similar to Census 
2000. The final wave likely would be primarily for areas that are prone 
to severe weather conditions, which make it difficult for field staff 
to perform their jobs earlier in the year. 

Pages 24 to 27 - Regarding the Field Data Collection Automation (FDCA) 
contractor's ability to be ready for the Address Canvassing operation, 
the Census Bureau designed the FDCA acquisition strategy to reduce 
risks related to cost, schedule, and performance. For example, we 
required offerors to develop and demonstrate a working prototype for 
Address Canvassing. The acquisition strategy also provided for a series 
of technical exchange meetings with each offeror over a period of three 
months. As a result of these meetings, offerors had the opportunity to 
learn details associated with the field data collection business 
processes. The Census Bureau reviewed the completed prototypes in 
January 2006, prior to contract award. 

On March 30, 2006, the Census Bureau awarded the FDCA contract to the 
Harris Corporation. The Harris Team developed the prototype system with 
the final solution in mind and plans to refine it, based on prototype 
lessons learned and based on the requirements decomposition activity 
that must occur between now and the end of the baseline planning phase 
in June 2006. The software development for Address Canvassing will 
occur between June and December 2006. The build cycle includes 
incremental coding and testing of functionality so that we have 
visibility into the evolving product, as well as extensive usability 
and load testing. 

Page 29 - As mentioned in the discussion about page 6 above, we will 
not be identifying vacant, occupied, and under-construction units in 
advance of the census. Also, we expect that by 2009, the situation on 
the ground will resemble what will exist in 2010 much more closely than 
what it would be if we were to make corrections to the list at this 
point in time. There is, in fact, a team working on the question of how 
to reflect the impact of the hurricanes in the MAF. Final 
recommendations should be out soon. 

Page 30 - Many problems identified in Census 2000 are being tackled on 
a variety of fronts, from sequential address list-building operations 
to the collection of GPS coordinates for streets and structures and the 
recycling of deletes to be validated concurrently with the NRFU 
operation. This will simplify our efforts to create complete address 
lists; reduce geocoding error, which contributes to a variety of 
coverage problems; reduce units deleted in error; and reduce 
duplication. 

[End of section] 

Appendix III: GAO Contact and Staff Acknowledgments: 

GAO Contact: 

Brenda S. Farrell, (202) 512-6806 or farrellb@gao.gov: 

Acknowledgments: 

In addition to the contact named above, Carlos Hazera, Assistant 
Director; Sheranda Smith Campbell; Betty Clark; Tim DiNapoli; Robert 
Goldenkoff; Shirley Hwang; Sonya Phillips; Lisa Pearson; Ilona Pesti; 
and Brendan St. Amant made key contributions to this report. 

FOOTNOTES 

[1] TIGER is a registered trademark of the U.S. Census Bureau. 

[2] The address file used to conduct the decennial census is referred 
to as the decennial master address file or DMAF. In this report we 
refer to the address file as the master address file (MAF). 


[3] The American Community Survey (ACS) will contain the same questions 
as the long form, but will be mailed monthly to an annual sample of 3 
million housing units. With the smaller sample, the ACS is designed to 
provide the same information at the same level of geographic detail as 
the long form by means of continuous measurement methodology in which 
survey responses will be accumulated over time. 

[4] GAO, 2010 Census: Basic Design Has Potential, but Remaining 
Challenges Need Prompt Resolution, GAO-05-9 (Washington, D.C.: Jan. 12, 
2005). 

[5] GAO, 2010 Census: Census Bureau Generally Follows Selected Leading 
Acquisition Planning Practices, but Continued Management Attention Is 
Needed to Help Ensure Success, GAO-06-277 (Washington, D.C.: May 18, 
2006). 

[6] GAO, Census Bureau: Important Activities for Improving Management 
of Key 2010 Decennial Acquisitions Remain to be Done, GAO-06-444T 
(Washington, D.C.: Mar. 1, 2006). 

[7] Another type of MAF error identified by the Bureau is 
misclassifying a housing unit as occupied when it is vacant. However, 
our focus is on whether an address has been properly captured in the 
MAF and not the occupancy status of the address. Therefore, we do not 
discuss occupancy errors in this report. 

[8] The research project regarding small multiunit structures is also 
about avoiding duplicate enumerations caused by confusion during mail 
delivery, as well as follow-up operations. 

[9] These approximately 653,000 valid addresses that were deleted are a 
subset of the 1.3 million addresses mistakenly deleted. 

[10] An unduplication operation in the summer of 2000 was implemented 
to identify and remove duplicate addresses. This operation was not a 
part of the original 2000 Census plan, but was considered necessary. 

[11] GAO, 2000 Census: Contingency Planning Needed to Address Risks 
That Pose a Threat to a Successful Census, GAO/GGD-00-6 (Washington, 
D.C.: Dec. 14, 1999). 

[12] GAO, Data Quality: Improvements to Count Correction Efforts Could 
Produce More Accurate Census Data, GAO-05-463 (Washington, D.C.: June 
20, 2005). 

[13] LUCA is an example of how the Bureau partners with external 
entities, tapping into their knowledge of local populations and housing 
conditions in order to secure a more complete count. In the Census 
Address List Improvement Act of 1994, Pub. L. No. 103-430, codified as 
13 U.S.C. § 16, Congress required the Bureau to develop a local address 
review program to give local and tribal governments greater input into 
the Bureau's address list development process. 

[14] As noted earlier in this report, during the 2004 Census Test the 
Bureau also experienced problems with the MCD (different model than the 
one used in the 2006 Census Test) used to collect nonresponse follow-up 
data. Specifically, that MCD had difficulties transmitting work and was 
slow to load maps. 

[15] GAO-06-277. 

GAO's Mission: 

The Government Accountability Office, the 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 GAO Reports and Testimony: 

The fastest and easiest way to obtain copies of GAO documents at no 
cost is through the Internet. GAO's Web site ( www.gao.gov ) contains 
abstracts and full-text files of current reports and testimony and an 
expanding archive of older products. The Web site features a search 
engine to help you locate documents using key words and phrases. You 
can print these documents in their entirety, including charts and other 
graphics. 

Each day, GAO issues a list of newly released reports, testimony, and 
correspondence. GAO posts this list, known as "Today's Reports," on its 
Web site daily. The list contains links to the full-text document 
files. To have GAO e-mail this list to you every afternoon, go to 
www.gao.gov and select "Subscribe to e-mail alerts" under the "Order 
GAO Products" heading. 

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, Waste, and Abuse in Federal Programs: 

Contact: 

Web site: www.gao.gov/fraudnet/fraudnet.htm 

E-mail: fraudnet@gao.gov 

Automated answering system: (800) 424-5454 or (202) 512-7470: 

Public Affairs: 

Jeff Nelligan, managing director, 

NelliganJ@gao.gov 

(202) 512-4800 

U.S. Government Accountability Office, 

441 G Street NW, Room 7149 

Washington, D.C. 20548: