2000 Census: Update on Essential Operations (Testimony, 03/14/2000,
GAO/T-GGD/AIMD-00-119).

This report updates the status of key census-taking operations. GAO
discusses developments in such essential areas as (1) outreach and
promotion, (2) field follow-up operations, and (3) data capture. GAO
also discusses steps that the Bureau has taken to help ensure that the
census questionnaires do not contain the same misprint in mailing
addresses found in the 120 million advance notification letters sent out
recently.

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

 REPORTNUM:  T-GGD/AIMD-00-119
     TITLE:  2000 Census: Update on Essential Operations
      DATE:  03/14/2000
   SUBJECT:  Census
	     Population statistics
	     Surveys
	     Government information dissemination
	     Public relations
	     Federal/state relations
	     Data collection
IDENTIFIER:  Census Bureau Questionnaire Assistance Center Program
	     Census Bureau Be Counted Program
	     Census Bureau Complete Count Committee Program
	     Census in Schools Program
	     Census Bureau Data Capture System 2000

******************************************************************
** This file contains an ASCII representation of the text of a  **
** GAO report.  This text was extracted from a PDF file.        **
** Delineations within the text indicating chapter titles,      **
** headings, and bullets have not been preserved, and in some   **
** cases heading text has been incorrectly merged into          **
** body text in the adjacent column.  Graphic images have       **
** not been reproduced, but figure captions are included.       **
** Tables are included, but column deliniations have not been   **
** preserved.                                                   **
**                                                              **
** Please see the PDF (Portable Document Format) file, when     **
** available, for a complete electronic file of the printed     **
** document's contents.                                         **
**                                                              **
** A printed copy of this report may be obtained from the GAO   **
** Document Distribution Center.  For further details, please   **
** send an e-mail message to:                                   **
**                                                              **
**                                            **
**                                                              **
** with the message 'info' in the body.                         **
******************************************************************

GAO/T-GGD/AIMD-00-119

2000 CENSUS: Update on Essential Operations (GAO/T-GGD/AIMD-00-119)
2000 CENSUS Update on Essential Operations
Statement of J. Christopher Mihm Associate Director, Federal Management
and Workforce Issues General Government Division
United States General Accounting Office
GAO Testimony Before the Subcommittee on the Census
Committee on Government Reform House of Representatives
For Release on Delivery 2: 00 p. m. EST Tuesday March 14, 2000
GAO/ T- GGD/ AIMD- 00- 119
Statement 2000 Census: Update on Essential Operations
Page 1 GAO/ T- GGD/ AIMD- 00- 119
Mr. Chairman, Mrs. Maloney, and Members of the Subcommittee: I am pleased to be here to provide an update on the status of key censustaking operations. When we last testified before the Subcommittee in midFebruary, we noted that although the Bureau of the Census has put forth a tremendous effort to help ensure as complete and accurate a count of the population as possible, certain essential census- taking operations still faced formidable challenges. 1 My statement today focuses on developments that have occurred since then regarding such essential activities as (1) outreach and promotion, (2) field follow- up operations, and (3) data capture. In addition, I will discuss the steps the Bureau has taken to ensure that the census questionnaires do not contain the same misprint in the mailing addresses that was contained in the 120 million advance notification letters that went out over the last few weeks.
In discussing these developments, I want to stress as I did at last month's hearing three themes in particular. First, it is important to go beyond national- level data when gauging the progress of the census. Key census operations are locally implemented; thus, national trends, although useful for providing an overall perspective, can mask successes or challenges occurring at the regional and local levels.
Second, many of the operational challenges that the Bureau faces are interrelated and their effects are often cumulative. Indeed, experience from the 1990 Census demonstrates how problems with one operation can trigger a chain of events that could ultimately affect the accuracy and cost of census data. On the other hand, a particularly successful outreach and promotion effort could also have a ripple effect that results in better quality data.
Third, cooperation is key to a successful census. Federal agencies, state and local governments, private and nonprofit organizations, and perhaps most important the public at large all have vital roles to play in ensuring a quality census. The Bureau cannot be expected to conduct a successful census on its own.
March is one of the most crucial time periods in the decade- long census cycle. Over the course of this month, the Bureau is to deliver census questionnaires to the vast majority of the nation's 120 million households and begin processing millions of completed questionnaires at its 4 data capture centers located across the country. In addition, outreach and
1 2000 Census: Status of Key Operations (GAO/ T- GGD/ AIMD- 00- 91, Feb. 15, 2000).
Statement 2000 Census: Update on Essential Operations
Page 2 GAO/ T- GGD/ AIMD- 00- 119
promotion efforts will be at their greatest intensity; the Bureau's temporary workforce will approach peak levels; and certain coverage improvement programs, such as those aimed at non- English- speaking and other hard- to- count populations, are to get under way. In short, the success of the operations that are launched this month will have a major impact on the results of the census.
My remarks today are based on our ongoing monitoring of the 2000 Census on behalf of the Subcommittee. Specific information was obtained through interviews with Bureau officials in headquarters, regional, and local census offices and analysis of Bureau documents and data. To obtain a local perspective of how the census is being conducted, we interviewed participants in the Bureau's local partnership program in each of 9 different localities. These individuals often local government employees are to help the Bureau promote the census and recruit employees. We selected the localities because, among other reasons, they have large concentrations of hard- to- enumerate populations. The partners' views are not projectable to the nation as a whole.
A high mail response rate is key to a successful census because it helps the Bureau obtain more accurate data and reduce what has been an errorprone and costly nonresponse follow- up workload. To help boost public participation in the census, the Bureau has instituted both a local and national outreach and promotion campaign. The local effort is largely based on partnerships with various governmental and nongovernmental organizations across the country, while the national initiative centers on a paid advertising campaign developed by a consortium of private sector advertising agencies, and includes an initiative to promote the census through the nation's schools. Both programs face the twin challenges of raising the public's awareness of the census and motivating people to respond.
In a noteworthy development over past censuses, for 2000, the Bureau has augmented its partnership efforts to include almost 90,000 organizations across the country ranging from state and local governments, community groups, and businesses. The Bureau's partnership program stems from its recognition that without the assistance and support from members of local communities, the message that participating in the census is important and confidential will not reach everyone-- particularly those in hard- to- count areas. The Bureau hopes that local people who are trusted by the communities they represent can promote the census and persuade everyone to respond. Bureau partnership specialists are to work with Local Outreach and
Promotion Efforts Are Proceeding With Mixed Results
Local Partners are Promoting the Census to Varying Degrees
Statement 2000 Census: Update on Essential Operations
Page 3 GAO/ T- GGD/ AIMD- 00- 119
local groups to help them initiate and sustain local outreach and promotion initiatives.
A key component of the Bureau's partnership program are Complete Count Committees, which consist of local government, religious, media, and other community leaders. The committees are to promote the census by sponsoring community events and placing articles in local newspapers, among other grassroots initiatives.
However, the committees differ regarding the resources they have available to promote the census. For example, one Complete Count Committee, the Philadelphia Census 2000 Project, has a $426,000 budget and full- time staff. Others, such as the Laredo, Texas, Complete Count Committee, have no budget and are staffed part time by government officials with other areas of responsibility.
Some committees are very active. The Chicago Complete Count Committee, for example, has arranged for 1.5 million census posters to be distributed through the Chicago Sun Times. The back of the posters lists 35 different local telephone numbers that people can call to obtain census information in such languages as Arabic, Korean, Polish, and Russian. Complete Count Committees are also helping the Bureau count hard- toenumerate populations. For example, the Cuyahoga County, Ohio, Complete Count Committee printed maps showing municipalities in which hard- to- count areas were located so they could target their outreach efforts accordingly. The Navajo Nation's committee allocated $20,000 for its own billboard advertising campaign and other promotional materials. They also translated the census long- and short- form questionnaires into their native tongue for use by census enumerators.
However, this level of effort is not uniform. For example, a representative of the Erie County, New York, Complete Count Committee said the committee has no activities planned at this time.
Regarding the support the Bureau is providing Complete Count Committees, the situation appears to be mixed. On the one hand, committee representatives we contacted were generally pleased with the assistance and guidance they were receiving from the Bureau's partnership staff. The local committee representatives were also satisfied with the quality and quantity of English- language materials they have received from the Bureau. On the other hand, several committee representatives we spoke with said that the amount of foreign language materials were insufficient. For example, a Texas Complete Count Committee with a large
Statement 2000 Census: Update on Essential Operations
Page 4 GAO/ T- GGD/ AIMD- 00- 119
Spanish- speaking population said that it has not been able to obtain sample Spanish questionnaires to train people assigned as Questionnaire Assistance Center staff. Other committee representatives were concerned about the lack of Asian language materials. To address this problem, local census staff and partners often arrange for the translation of Bureau materials into languages needed by their communities.
In addition to the partnership program, in 1997, the Bureau hired a consortium of private- sector advertising agencies led by Young & Rubicam to develop an extensive paid advertising program to promote the census. As part of this initiative, Scholastic, Inc., was hired to develop a package of educational material and a marketing plan to promote the census in the nation's schools. Known as the Census in Schools program, the Bureau hopes to increase response rates by encouraging students to remind their parents or guardians to complete their census forms.
During the spring of 1999, the Bureau sent information about the Census in Schools Program to all elementary school teachers and secondary social studies and math teachers located in hard- to- count areas of the country, or about 40 percent of all schools. Teachers could return an enclosed form and request teacher kits containing maps and other informational material. In September 1999, principals, department chairs, and district curriculum coordinators in the remaining schools were to receive a fall recruitment package including a sample teacher kit to distribute along with information for ordering additional kits. To date, the Bureau has filled orders for about 1.5 million teacher kits for elementary, middle and high schools.
However, problems occurred in distributing the materials. The fall recruitment package was not mailed until December, rather than September as originally planned. As a result, schools in this group that wanted to participate in the program did not start receiving materials until about January 2000. Further, Bureau officials reported that orders for some materials, especially for kindergarten through 4 th grade teachers, were not filled between October 1999 and mid- December 1999 due in part to delays in printing materials for the fall recruitment package and the need to reprint teaching kits.
To address the backlog of orders resulting from the printing delays and to expedite the fulfillment of new orders, the Bureau changed its distribution method in January. Since then, new orders for Census in Schools materials are taking between 2 and 4 weeks to be filled, according to Bureau officials responsible for the program. Census in Schools Program
Seen as Effective But Hampered by Distribution Delays
Statement 2000 Census: Update on Essential Operations
Page 5 GAO/ T- GGD/ AIMD- 00- 119
The Complete Count Committee and local census office representatives we spoke to thought the idea of promoting the census through schools was a good one and commended the Bureau and Scholastic on the high quality of the Census in Schools materials. However, several representatives were affected by the distribution backlog. One committee representative we spoke to in early- March reported that the Census in Schools materials arrived too late to be effectively integrated into the curriculum. Another Complete Count Committee was so concerned that Census in Schools materials would not come in time that, according to a committee representative, it spent over $400 printing the material from the Bureau's Internet site. The Census Bureau's Philadelphia office also reported that they printed and distributed Census in Schools materials on their own due to the distribution backlog.
To help ensure a complete and accurate count of the nation's population, the Bureau conducts a nationwide field follow- up operation in which enumerators visit and attempt to collect census information from each nonresponding household. In addition, the Bureau has developed certain coverage improvement initiatives aimed at improving the count of hard- toenumerate population groups.
The Bureau's staffing needs are enormous. To conduct its peak field operations, and to cover for the possibility of high turnover rates, the Bureau estimates it will need to recruit nearly 2.4 million qualified applicants for census employment. To qualify for census employment, candidates must pass a basic skills test and a personal background check.
The Bureau's update/ leave operation, which began March 3, was the first test of the Bureau's ability to staff its peak field activities. Under the update/ leave operation, enumerators are to deliver questionnaires to some 24 million housing units located mainly in small towns and rural areas where address systems have less geographic structure.
Over 70,000 enumerator and other staff are now in the field conducting update/ leave, and the Bureau does not anticipate that it will experience any significant problems staffing this operation. Nevertheless, while the Bureau met its update/ leave staffing needs, and has a nationwide pool of 1.7 million qualified applicants for subsequent operations, it understands that it still faces a formidable recruiting challenge. Over the remaining weeks, the Bureau needs to recruit an additional 700,000 qualified applicants to meet its April 19th goal of 2.4 million qualified applicants to conduct nonresponse follow- up operations. Experience from the 1990 Census suggests that if the Bureau encounters staffing shortages, it could Field Follow- up Efforts
Face Challenges The Bureau Is Taking Action to Address Staffing Challenges
Statement 2000 Census: Update on Essential Operations
Page 6 GAO/ T- GGD/ AIMD- 00- 119
delay the completion of nonresponse follow- up in some locations and compromise data quality.
According to the Census Bureau, as of March 2, the latest date for which data were available, the Bureau set a benchmark to recruit about 71 percent of the 2.4 million qualified applicants needed. Nationally, the Bureau was ahead of this objective, having achieved 74 percent of its recruiting goal. However, as I previously noted, national data can mask local challenges; in fact, 7 of the Bureau's 12 regional offices fell short of the 71- percent benchmark. The current shortfalls ranged from between 4and 14- percentage points. As shown in figure 1, 6 of the regions were also below the Bureau's 52- percent benchmark as of February 9, when the Bureau initially provided us with data.
Statement 2000 Census: Update on Essential Operations
Page 7 GAO/ T- GGD/ AIMD- 00- 119
Source: GAO Analysis of U. S. Census Bureau data.
At the local level, 270 of the Bureau's 511 local census offices fell below the Bureau's March 2 benchmark of 71- percent. 2 Of these 270, 22 had recruited fewer than half of the qualified applicants that the Bureau estimated it needed as of March 2nd.
2 Our analysis did not include nine local census offices in Puerto Rico.
Figure 1: Regional Offices Are Progressing Toward Their Recruiting Goals
Statement 2000 Census: Update on Essential Operations
Page 8 GAO/ T- GGD/ AIMD- 00- 119
Factors contributing to the Bureau's successes and shortfalls vary widely across the country. For example, managers at a Phoenix local census office told us that they were competing for employees with the Bureau's nearby Census Data Capture Center, which has indoor work and higher wages. In contrast, the Phoenix local census office offers applicants the lower- paid job of enumerating people in extreme heat. Other local census offices, such as those in the western and southern parts of Chicago, are better able to meet their staffing goals compared to the northern and central Chicago local census offices, which are located in areas with lower unemployment levels. Additionally, according to local and regional Bureau officials we contacted, local census offices that are conducting the update/ leave operation have been more successful at meeting their recruiting goals because, among other reasons, they can offer immediate jobs.
The regional and local census offices continue to aggressively recruit applicants to ensure that their recruiting goals will be met. For example, the Bureau is working with communities to set up testing sites at community events, meeting with leaders of local civic organizations to understand how to best recruit their members, and placing recruitment advertising in community newspapers to recruit individuals to work within their own neighborhoods.
Moreover, since November 1999, the Bureau has been working with state governments to obtain exemptions so that individuals receiving Temporary Assistance to Needy Families, Medicaid, and selected other types of assistance would not have their benefits reduced when earning temporary census income. Such exemptions could expand the applicant pool by making census employment more desirable to these individuals. According to the Bureau, 36 states and the Virgin Islands have granted an exemption for 1 or more of these programs as of March 6, 2000.
Bureau headquarters has also developed a list of actions that its regional offices can take to help local census offices that are not reaching their recruiting goals. These actions, which vary by the severity of the shortfall, range from saturating neighborhoods with recruiting advertising and targeted postcard mailings in the less severe instances, to raising pay rates and sending in special recruiting teams in the most dire cases. We found that several census regional and local offices are implementing some of these actions. For example, the Atlanta and several other regional offices are mailing postcards to targeted zip codes that local census offices have flagged as hard- to- recruit areas. In addition, regional offices have authorized pay rate increases for some local census offices. In February
Statement 2000 Census: Update on Essential Operations
Page 9 GAO/ T- GGD/ AIMD- 00- 119
2000, for example, the Manassas, Virginia, local census office raised pay rates from $11.25 to $12.75 an hour and the Placerville, California, office raised enumerator pay rates from $11.00 to $13.00 an hour.
The Bureau has included several coverage improvement programs in the 2000 Census that are aimed at increasing the count of various hard- toenumerate groups. The hard- to- enumerate include minority groups, migrant and undocumented workers, the homeless, and those with little or no English- language skills, among others. Such individuals are often missed during the Bureau's initial enumeration and nonresponse follow- up operation.
Two coverage improvement programs we have been examining are the Bureau's walk- in Questionnaire Assistance Centers and Be Counted Program. Questionnaire Assistance Centers are to help people especially those with little or no English- speaking ability complete their census questionnaires by providing assistance in various languages on a walk- in basis. Assistance guides in 37 languages that are designed to help people complete their English language forms, as well as large- print English guides, are to be available at all assistance centers. Assistance guides in 12 additional languages will be available at selected centers on the basis of specific community needs. The centers, which are located in various public locations, such as community centers and libraries have been open since March 8 when questionnaires were first distributed, and are to close on April 14, 2000, about 2 weeks before the beginning of nonresponse follow- up. The Bureau is staffing the centers with a mix of volunteers and paid employees.
The Be Counted Program is designed to count people who believe they did not receive a census questionnaire or who were otherwise not included in the census. The program also allows people who had no usual residence on Census Day, such as transients, migrants, or seasonal farm workers to be included in the census. Be Counted questionnaires essentially a modified short- form are to be available in Questionnaire Assistance Centers as well as in other public locations for people to complete and return to the Bureau. The program is to run from March 31 until April 11, 2000.
The Bureau anticipated setting up as many as 66,895 Be Counted locations (an average of about one Be Counted site for each of the nation's census tracts). Of these, 34, 725 were to be stand- alone sites, and the remainder were to be located in each of the 32,170 Questionnaire Assistance Centers the Bureau expected to establish. Key Coverage Improvement
Programs Appear on Track
Statement 2000 Census: Update on Essential Operations
Page 10 GAO/ T- GGD/ AIMD- 00- 119
Bureau data as of March 1 indicate that the Bureau has established 17,209 stand- alone Be Counted sites and 29,161 Questionnaire Assistance Centers. Table 1 shows how these numbers compare to the Bureau's projections for each census region.
Census region Projected number
of Questionnaire Assistance
Centers Actual number of
Questionnaire Assistance
Centers Percentage of projected number
Projected number of Be Counted
sites Actual number of Be Counted sites
Percentage of projected number
Atlanta 969 1,109 114% 854 643 75% Boston 1,400 1,905 136 925 1,310 142 Charlotte 1,544 871 56 6,905 702 10 Chicago 12,000 8,395 70 3,900 997 26 Dallas 1,500 2,922 195 2,000 836 42 Denver 1,500 1,476 98 2,000 1,425 71 Detroit 1,455 1,136 78 2,000 899 45 Kansas City 2,414 1,346 56 1,521 1,166 77 Los Angeles 3,000 2,231 74 3,000 2,700 90 New York 2,000 1,887 94 3,900 1,716 44 Philadelphia 1,088 1,956 180 2,220 1,575 71 Seattle 3,300 3,927 119 5,500 3,240 59
Total 32,170 29,161 91% 34,725 17,209 50% Source: U. S. Census Bureau data as of March 1, 2000.
Although short of the Bureau's projections overall, several regions are close to or are exceeding the Bureau's expectations, and the combined number of sites 46,370 is still substantial. Indeed, by comparison, there are about 12,600 McDonald's restaurants in the United States. As we discussed in our February report, in addition to the number of sites, other factors will be critical to the effectiveness of the two programs. 3 The factors include (1) working with local partners to identify sites with locations and schedules that best meet the needs of targeted groups and are adequately publicized; (2) ensuring that sites have street- level visibility so that targeted groups are able to find them; (3) monitoring usage so that people will be able to find forms and obtain assistance when and where they are supposed to be available; and (4) in the case of Questionnaire Assistance Centers, making sure staff are available with appropriate foreign language skills.
In implementing the two programs, the Bureau appears to be taking steps consistent with these factors. For example, officials from the Los Angeles
3 2000 Census: Actions Taken to Improve the Be Counted and Questionnaire Assistance Center Programs (

GAO/GGD-00-47
, Feb. 25, 2000).
Table 1: Status of Questionnaire Assistance Centers and Be Counted Sites by Census Region
Statement 2000 Census: Update on Essential Operations
Page 11 GAO/ T- GGD/ AIMD- 00- 119
Regional Census Center have worked with the Los Angeles Unified School District to open Questionnaire Assistance sites at public schools in hardto- enumerate areas. In addition, the Charlotte Regional Census Center has worked with the consortium of Historically Black Colleges and Universities within the region to recruit student volunteers to work in the Questionnaire Assistance Centers. The census work satisfies a public service graduation requirement. Similarly, in Detroit, regional officials noted how they were giving priority to hiring assistance center staff with foreign language skills.
In February, we testified that the Bureau faced formidable challenges in ensuring the operational readiness of its four data capture centers (DCC), as much remained to be done to complete and test DCS 2000 4 within a very short time frame. In particular, we noted that when we concluded our review in early January, measures of DCS 2000's quality and maturity were mixed. Moreover, we added that as of February, the results of important test events had not yet been reported, last minute changes to DCS 2000's software and hardware configurations were still being made to address a recently discovered operational problem, and critical test events yet to be conducted would likely identify new problems.
As of today, the Bureau reports that it has implemented the near- term DCS 2000 system changes and completed all scheduled test events, including an operational test at each DCC and a test of the four centers operating concurrently. Further, as of today, the DCCs have been operating for about 1 week, the first 2 days of which the DCCs checked in an average of 117,525 questionnaires. This workload represents about 8 percent of the daily workload expected later this month when questionnaire processing is expected to reach the production level of about 1. 5 million questionnaires processed per day. Thus, actual operations do not yet demonstrate the DCC's readiness to operate at expected production- level workloads. Additionally, the information we have seen on actual operations does not address whether recent changes to DCS 2000 are functioning correctly. For this and other reasons discussed below, we remain uncertain about the centers' readiness to meet the full production- level workload anticipated to begin in the next 2 weeks.
4 Specifically, each DCC will rely on DCS 2000 to check in questionnaires and determine which households have responded, scan the questionnaires to produce an electronic image, optically read handwritten marks and writings from the imaged questionnaires, and convert the data into files that will be transmitted to Bureau headquarters for tabulation and analysis. Production- Level
Operational Readiness of Data Capture Centers Remains Uncertain
Statement 2000 Census: Update on Essential Operations
Page 12 GAO/ T- GGD/ AIMD- 00- 119
To ensure DCC readiness, the Bureau and its contractors conducted a series of operational tests at each of the four DCCs. These tests were conducted in an incremental fashion, whereby improvements and lessons learned were incorporated in succeeding tests. These tests were designed to assess a DCC's ability to perform key operations, such as sorting, checking- in, and scanning questionnaires at required production levels. Although the four operational tests have been concluded, the Bureau's report on the results of its operational test at the Jeffersonville DCC, performed during December 1999 through February 2000, is not yet available. However, the Jeffersonville site manager told us that this operational test, although less structured than the other DCC operational tests and focused on qualitative observations rather than quantitative measures, was a success.
Although the DCCs perform the same basic data capture operations, test results from Jeffersonville are significant because actual operations, and thus center operational capability and readiness, are very much people dependent. Therefore, it is important to know the Jeffersonville operational test results to determine if the problems identified in the Pomona DCC operational test 5 have been successfully addressed at the Jeffersonville DCC, as they were at the Phoenix DCC. Moreover, the Jeffersonville DCC is unique because unlike the other DCCs, which are operated by contractor personnel, Jeffersonville is being operated by Bureau personnel, supplemented with temporary workers. Bureau officials told us that the operating procedures developed by the DCC operations contractor were provided to the Jeffersonville DCC, but that the Bureau's own staff in Jeffersonville are not required to use these procedures and, in fact, can create their own. Additionally, the Jeffersonville workload mix is different from the other DCCs, encompassing other forms, such as Be Counted and foreign language forms, which the other centers do not process.
As a result of the DCC operational tests, the Bureau realized that the keying rate for key from image (KFI) was not high enough to meet its master schedule for completing Census 2000 and delivering the apportionment counts by December 31, 2000, as required by law. To resolve this dilemma, the Bureau adopted a two- pass approach to data capture operations. During the first pass from March 6, 2000, to August
5 The Pomona DCC operational test, conducted in October and November 1999, showed that the center was not able to sort or scan the number of questionnaires expected during peak data capture operations. The test also demonstrated that the DCC could not achieve the keying rate necessary to handle the expected key from image workload, which entails manually entering data because handwritten characters or marks were difficult for the system to recognize. Jeffersonville DCC
Operational Test Recent DCS 2000 Software Changes
Statement 2000 Census: Update on Essential Operations
Page 13 GAO/ T- GGD/ AIMD- 00- 119
31, 2000 the DCCs will capture the data necessary to determine the apportionment counts, referred to as 100- percent data. 6 During the second pass from September 11, 2000, to November 15, 2000 the Bureau will capture the remaining data from the long forms, known as sample data, which include the detailed social, economic, and housing information collected for a sample of living quarters in the United States. Because the keyers will focus first on the 100- percent data instead of working on both the 100- percent data, and the sample data simultaneously, the Bureau expects to be able to complete the 100 percent data capture by August 31, 2000, providing enough time to produce the apportionment counts by the end of December.
To implement the two- pass data capture solution, two sets of changes or releases were required to DCS 2000. The first release, designed to support the first pass, was completed in early February, according to the Bureau. This work involved modifying DCS 2000 software to write the long form images to a mass storage unit and to not present the sample data to keyers. This release was not in place in time to be included as part of operational tests at the individual DCCs. However, it was in place for the final four- site operational test, which I discuss shortly. Nevertheless, while Bureau officials have stated the changes were successfully implemented, the Bureau has not yet provided us with the procedures used for testing these changes and the results. As a result, we do not have the information needed to know with any certainty whether these software changes are performing as intended.
To help prepare for the actual data capture operations during Census 2000, the Bureau and its contractors conducted a final operational test on February 22 to 25, 2000. This test was important because it involved production- level workloads for all four DCCs, centralized operations, and Bureau headquarters simultaneously. However, the test was limited because it did not include all DCC operations, such as the DCC's ability to sort, check- in, and prepare questionnaires for processing. Also, most of the questionnaires used in the test had machine printed, rather than handwritten responses, with the same answers on each questionnaire, which would simplify KFI demands. Because of these limitations and because, as previously stated, the individual DCC operational tests were incremental, some data capture operations have not been verified with a production- level workload at all DCCs. Specifically, neither the Baltimore DCC nor the Pomona DCC have successfully conducted operational tests
6 The 100- percent data are the population and housing information collected for all living quarters in the United States and includes name, sex, and race of each person living in a household. Four- site Operational Test
Statement 2000 Census: Update on Essential Operations
Page 14 GAO/ T- GGD/ AIMD- 00- 119
of their abilities to support a production load for sorting or document preparation.
The Bureau has characterized the four- site test as successful, however, it has not yet provided us with the results. Assessing the results is still important because the tests did evaluate some important data capture operations, such as the DCC's ability to scan production- level workloads of questionnaires and transmit the resulting data to headquarters.
As previously noted, the Bureau still plans on making a second set of modifications to DCS 2000 to support its new data capture approach. This work will involve modifying the system to retrieve the images of the approximately 22 million sample data questionnaires from the mass storage unit and present those requiring action to keyers. The Bureau plans to complete development and system- level test activities for this modification by July 31, 2000.
Although these changes are necessary for the Bureau to implement its approach to meeting its schedule for apportionment counts, they will require the DCS 2000 development contractor to divert personnel from the DCCs and the central system support facility to ongoing DCS 2000 development. The DCS 2000 development contractor identified the diversion of personnel as a risk that could adversely impact first pass data capture operations. In addition, the Bureau plans to suspend operations at one yet- to- be selected DCC in August 2000 to run an operational test of the second pass DCS 2000 software release. This will require the Bureau to redistribute this DCC's workload among the other three DCCs.
In addition to the previously mentioned challenges, a new, unforeseen issue surfaced late last month when it was discovered that an extra digit was erroneously inserted in front of the house numbers used to mail advance notification letters. Despite this error, the U. S. Postal Service found that the postal barcodes were correctly printed, which, according to a postal official, helped assure the letters were properly delivered. The advance letter is part of the Bureau's multiple mailing strategy, which is designed to boost response rates by increasing the number of mail contacts the Bureau has with the public. The advance letter notifies people that they will soon receive a census form, describes why it is important to respond, and provides an opportunity to request a questionnaire in Spanish, Chinese, Korean, Vietnamese, or Tagalog. Ongoing DCS 2000 Changes
Bureau Confident That Address Error Is Limited to Advance Notice Letters
Statement 2000 Census: Update on Essential Operations
Page 15 GAO/ T- GGD/ AIMD- 00- 119
According to the Bureau, the misprint occurred during printing and addressing by a private contractor. Bureau and Government Printing Office officials believe that the source of the error was a glitch in the contractor's printing program because, on the basis of Postal Service and Bureau information, the error occurred on all of the letters.
According to Bureau officials, quality assurance procedures developed for printing the advance letters and other census documents, including census questionnaires, were not designed to check the accuracy of the printed addresses. Regarding census questionnaires, for example, the quality assurance procedures were meant to identify printing problems that could affect the data capture process, such as type size and color defects.
Once they learned of the error, Bureau officials said that both the Bureau and contractors responsible for printing census questionnaires checked a sample of each of the six types of questionnaires with preprinted addresses to ensure they did not contain the same misprint contained in the advance letters. According to those officials, the samples came from the Bureau's earlier quality assurance samples and from the contractors' plant floors, and included the update/ leave short- form questionnaire printed by the same contractor that printed the advance letters. Specifically, Bureau and contractor staff:
 visually inspected printed addresses for the same error identified in the advance letters,
 matched printed addresses against addresses contained in the census address file,
 compared printed addresses against addresses contained in the Postal Service address files located on the Postal Service Internet site, and/ or
 compared address information read from the Postal Service barcodes against the addresses printed on the questionnaires.
Bureau officials also said they checked the addresses on a limited number of reminder cards, including a visual inspection and comparison of the addresses against the Postal Service file.
Although documentation of the results of these checks is not available, they gave Bureau officials confidence that the address error identified in the advance letter is not present in census questionnaires and reminder cards. In addition, according to Bureau officials, in the course of these checks, no other types of errors were detected. We independently reviewed a very small sample of questionnaires at four post offices in
Statement 2000 Census: Update on Essential Operations
Page 16 GAO/ T- GGD/ AIMD- 00- 119
Northern Virginia and found they did not contain the same misprint present in the advance letters.
Mr. Chairman, with Census Day just over 2 weeks away, I would like to end by briefly returning to one of the themes I have highlighted today public cooperation. The mail response rate the ultimate measure of cooperation-- will be central to determining the overall accuracy and cost of the census. The national response rate will therefore provide a first indication of the success of the 2000 Census. A high mail response rate will reduce the Bureau's follow- up workload and, accordingly, relieve some of the staffing and schedule pressures the Bureau confronts. On behalf of the Subcommittee, we will track these and other operational data and continue to monitor the progress of the census.
Mr. Chairman, this concludes my prepared statement. I would be pleased to respond to any questions you or other Members of the Subcommittee may have.
Contact and Acknowledgments
For further information regarding this testimony, please contact J. Christopher Mihm at (202) 512- 8676 or Randolph C. Hite on (202) 5126240. Individuals making key contributions to this testimony included Thomas Beall, Mark Bird, Cristina Chaplain, Garry Durfey, Deborah Eichhorn, Robert Goldenkoff, Richard Hung, Lily Kim, Victoria Miller, Vicky L. Miller, Ty Mitchell, and Lynn Wasielewski.
Page 17 GAO/ T- GGD/ AIMD- 00- 119
Page 18 GAO/ T- GGD/ AIMD- 00- 119
Page 19 GAO/ T- GGD/ AIMD- 00- 119
Viewing GAO Reports on the Internet For information on how to access GAO reports on the INTERNET, send e- mail message with info in the body to:
info@ www. gao. gov or visit GAO's World Wide Web Home Page at: http:// www. gao. gov Reporting Fraud, Waste, and Abuse in Federal Programs To contact GAO's Fraud Hotline use: Web site: http:// www. gao. gov/ fraudnet/ fraudnet. htm E- Mail: fraudnet@ gao. gov Telephone: 1- 800- 424- 5454 (automated answering system)
(410555)
2000 CENSUS: Update on Essential Operations (GAO/T-GGD/AIMD-00-
119) 2000 CENSUS Update on Essential Operations Statement of J.
Christopher Mihm Associate Director, Federal Management and
Workforce Issues General Government Division United States General
Accounting Office GAO Testimony Before the Subcommittee on the
Census Committee on Government Reform House of Representatives For
Release on Delivery 2: 00 p. m. EST Tuesday March 14, 2000 GAO/ T-
GGD/ AIMD- 00- 119 Statement 2000 Census: Update on Essential
Operations Page 1 GAO/ T- GGD/ AIMD- 00- 119 Mr. Chairman, Mrs.
Maloney, and Members of the Subcommittee: I am pleased to be here
to provide an update on the status of key censustaking operations.
When we last testified before the Subcommittee in midFebruary, we
noted that although the Bureau of the Census has put forth a
tremendous effort to help ensure as complete and accurate a count
of the population as possible, certain essential census- taking
operations still faced formidable challenges. 1 My statement today
focuses on developments that have occurred since then regarding
such essential activities as (1) outreach and promotion, (2) field
follow- up operations, and (3) data capture. In addition, I will
discuss the steps the Bureau has taken to ensure that the census
questionnaires do not contain the same misprint in the mailing
addresses that was contained in the 120 million advance
notification letters that went out over the last few weeks. In
discussing these developments, I want to stress as I did at last
month's hearing three themes in particular. First, it is important
to go beyond national- level data when gauging the progress of the
census. Key census operations are locally implemented; thus,
national trends, although useful for providing an overall
perspective, can mask successes or challenges occurring at the
regional and local levels. Second, many of the operational
challenges that the Bureau faces are interrelated and their
effects are often cumulative. Indeed, experience from the 1990
Census demonstrates how problems with one operation can trigger a
chain of events that could ultimately affect the accuracy and cost
of census data. On the other hand, a particularly successful
outreach and promotion effort could also have a ripple effect that
results in better quality data. Third, cooperation is key to a
successful census. Federal agencies, state and local governments,
private and nonprofit organizations, and perhaps most important
the public at large all have vital roles to play in ensuring a
quality census. The Bureau cannot be expected to conduct a
successful census on its own. March is one of the most crucial
time periods in the decade- long census cycle. Over the course of
this month, the Bureau is to deliver census questionnaires to the
vast majority of the nation's 120 million households and begin
processing millions of completed questionnaires at its 4 data
capture centers located across the country. In addition, outreach
and 1 2000 Census: Status of Key Operations (GAO/ T- GGD/ AIMD-
00- 91, Feb. 15, 2000). Statement 2000 Census: Update on Essential
Operations Page 2 GAO/ T- GGD/ AIMD- 00- 119 promotion efforts
will be at their greatest intensity; the Bureau's temporary
workforce will approach peak levels; and certain coverage
improvement programs, such as those aimed at non- English-
speaking and other hard- to- count populations, are to get under
way. In short, the success of the operations that are launched
this month will have a major impact on the results of the census.
My remarks today are based on our ongoing monitoring of the 2000
Census on behalf of the Subcommittee. Specific information was
obtained through interviews with Bureau officials in headquarters,
regional, and local census offices and analysis of Bureau
documents and data. To obtain a local perspective of how the
census is being conducted, we interviewed participants in the
Bureau's local partnership program in each of 9 different
localities. These individuals often local government employees are
to help the Bureau promote the census and recruit employees. We
selected the localities because, among other reasons, they have
large concentrations of hard- to- enumerate populations. The
partners' views are not projectable to the nation as a whole. A
high mail response rate is key to a successful census because it
helps the Bureau obtain more accurate data and reduce what has
been an errorprone and costly nonresponse follow- up workload. To
help boost public participation in the census, the Bureau has
instituted both a local and national outreach and promotion
campaign. The local effort is largely based on partnerships with
various governmental and nongovernmental organizations across the
country, while the national initiative centers on a paid
advertising campaign developed by a consortium of private sector
advertising agencies, and includes an initiative to promote the
census through the nation's schools. Both programs face the twin
challenges of raising the public's awareness of the census and
motivating people to respond. In a noteworthy development over
past censuses, for 2000, the Bureau has augmented its partnership
efforts to include almost 90,000 organizations across the country
ranging from state and local governments, community groups, and
businesses. The Bureau's partnership program stems from its
recognition that without the assistance and support from members
of local communities, the message that participating in the census
is important and confidential will not reach everyone--
particularly those in hard- to- count areas. The Bureau hopes that
local people who are trusted by the communities they represent can
promote the census and persuade everyone to respond. Bureau
partnership specialists are to work with Local Outreach and
Promotion Efforts Are Proceeding With Mixed Results Local Partners
are Promoting the Census to Varying Degrees Statement 2000 Census:
Update on Essential Operations Page 3 GAO/ T- GGD/ AIMD- 00- 119
local groups to help them initiate and sustain local outreach and
promotion initiatives. A key component of the Bureau's partnership
program are Complete Count Committees, which consist of local
government, religious, media, and other community leaders. The
committees are to promote the census by sponsoring community
events and placing articles in local newspapers, among other
grassroots initiatives. However, the committees differ regarding
the resources they have available to promote the census. For
example, one Complete Count Committee, the Philadelphia Census
2000 Project, has a $426,000 budget and full- time staff. Others,
such as the Laredo, Texas, Complete Count Committee, have no
budget and are staffed part time by government officials with
other areas of responsibility. Some committees are very active.
The Chicago Complete Count Committee, for example, has arranged
for 1.5 million census posters to be distributed through the
Chicago Sun Times. The back of the posters lists 35 different
local telephone numbers that people can call to obtain census
information in such languages as Arabic, Korean, Polish, and
Russian. Complete Count Committees are also helping the Bureau
count hard- toenumerate populations. For example, the Cuyahoga
County, Ohio, Complete Count Committee printed maps showing
municipalities in which hard- to- count areas were located so they
could target their outreach efforts accordingly. The Navajo
Nation's committee allocated $20,000 for its own billboard
advertising campaign and other promotional materials. They also
translated the census long- and short- form questionnaires into
their native tongue for use by census enumerators. However, this
level of effort is not uniform. For example, a representative of
the Erie County, New York, Complete Count Committee said the
committee has no activities planned at this time. Regarding the
support the Bureau is providing Complete Count Committees, the
situation appears to be mixed. On the one hand, committee
representatives we contacted were generally pleased with the
assistance and guidance they were receiving from the Bureau's
partnership staff. The local committee representatives were also
satisfied with the quality and quantity of English- language
materials they have received from the Bureau. On the other hand,
several committee representatives we spoke with said that the
amount of foreign language materials were insufficient. For
example, a Texas Complete Count Committee with a large Statement
2000 Census: Update on Essential Operations Page 4 GAO/ T- GGD/
AIMD- 00- 119 Spanish- speaking population said that it has not
been able to obtain sample Spanish questionnaires to train people
assigned as Questionnaire Assistance Center staff. Other committee
representatives were concerned about the lack of Asian language
materials. To address this problem, local census staff and
partners often arrange for the translation of Bureau materials
into languages needed by their communities. In addition to the
partnership program, in 1997, the Bureau hired a consortium of
private- sector advertising agencies led by Young & Rubicam to
develop an extensive paid advertising program to promote the
census. As part of this initiative, Scholastic, Inc., was hired to
develop a package of educational material and a marketing plan to
promote the census in the nation's schools. Known as the Census in
Schools program, the Bureau hopes to increase response rates by
encouraging students to remind their parents or guardians to
complete their census forms. During the spring of 1999, the Bureau
sent information about the Census in Schools Program to all
elementary school teachers and secondary social studies and math
teachers located in hard- to- count areas of the country, or about
40 percent of all schools. Teachers could return an enclosed form
and request teacher kits containing maps and other informational
material. In September 1999, principals, department chairs, and
district curriculum coordinators in the remaining schools were to
receive a fall recruitment package including a sample teacher kit
to distribute along with information for ordering additional kits.
To date, the Bureau has filled orders for about 1.5 million
teacher kits for elementary, middle and high schools. However,
problems occurred in distributing the materials. The fall
recruitment package was not mailed until December, rather than
September as originally planned. As a result, schools in this
group that wanted to participate in the program did not start
receiving materials until about January 2000. Further, Bureau
officials reported that orders for some materials, especially for
kindergarten through 4 th grade teachers, were not filled between
October 1999 and mid- December 1999 due in part to delays in
printing materials for the fall recruitment package and the need
to reprint teaching kits. To address the backlog of orders
resulting from the printing delays and to expedite the fulfillment
of new orders, the Bureau changed its distribution method in
January. Since then, new orders for Census in Schools materials
are taking between 2 and 4 weeks to be filled, according to Bureau
officials responsible for the program. Census in Schools Program
Seen as Effective But Hampered by Distribution Delays Statement
2000 Census: Update on Essential Operations Page 5 GAO/ T- GGD/
AIMD- 00- 119 The Complete Count Committee and local census office
representatives we spoke to thought the idea of promoting the
census through schools was a good one and commended the Bureau and
Scholastic on the high quality of the Census in Schools materials.
However, several representatives were affected by the distribution
backlog. One committee representative we spoke to in early- March
reported that the Census in Schools materials arrived too late to
be effectively integrated into the curriculum. Another Complete
Count Committee was so concerned that Census in Schools materials
would not come in time that, according to a committee
representative, it spent over $400 printing the material from the
Bureau's Internet site. The Census Bureau's Philadelphia office
also reported that they printed and distributed Census in Schools
materials on their own due to the distribution backlog. To help
ensure a complete and accurate count of the nation's population,
the Bureau conducts a nationwide field follow- up operation in
which enumerators visit and attempt to collect census information
from each nonresponding household. In addition, the Bureau has
developed certain coverage improvement initiatives aimed at
improving the count of hard- toenumerate population groups. The
Bureau's staffing needs are enormous. To conduct its peak field
operations, and to cover for the possibility of high turnover
rates, the Bureau estimates it will need to recruit nearly 2.4
million qualified applicants for census employment. To qualify for
census employment, candidates must pass a basic skills test and a
personal background check. The Bureau's update/ leave operation,
which began March 3, was the first test of the Bureau's ability to
staff its peak field activities. Under the update/ leave
operation, enumerators are to deliver questionnaires to some 24
million housing units located mainly in small towns and rural
areas where address systems have less geographic structure. Over
70,000 enumerator and other staff are now in the field conducting
update/ leave, and the Bureau does not anticipate that it will
experience any significant problems staffing this operation.
Nevertheless, while the Bureau met its update/ leave staffing
needs, and has a nationwide pool of 1.7 million qualified
applicants for subsequent operations, it understands that it still
faces a formidable recruiting challenge. Over the remaining weeks,
the Bureau needs to recruit an additional 700,000 qualified
applicants to meet its April 19th goal of 2.4 million qualified
applicants to conduct nonresponse follow- up operations.
Experience from the 1990 Census suggests that if the Bureau
encounters staffing shortages, it could Field Follow- up Efforts
Face Challenges The Bureau Is Taking Action to Address Staffing
Challenges Statement 2000 Census: Update on Essential Operations
Page 6 GAO/ T- GGD/ AIMD- 00- 119 delay the completion of
nonresponse follow- up in some locations and compromise data
quality. According to the Census Bureau, as of March 2, the latest
date for which data were available, the Bureau set a benchmark to
recruit about 71 percent of the 2.4 million qualified applicants
needed. Nationally, the Bureau was ahead of this objective, having
achieved 74 percent of its recruiting goal. However, as I
previously noted, national data can mask local challenges; in
fact, 7 of the Bureau's 12 regional offices fell short of the 71-
percent benchmark. The current shortfalls ranged from between 4and
14- percentage points. As shown in figure 1, 6 of the regions were
also below the Bureau's 52- percent benchmark as of February 9,
when the Bureau initially provided us with data. Statement 2000
Census: Update on Essential Operations Page 7 GAO/ T- GGD/ AIMD-
00- 119 Source: GAO Analysis of U. S. Census Bureau data. At the
local level, 270 of the Bureau's 511 local census offices fell
below the Bureau's March 2 benchmark of 71- percent. 2 Of these
270, 22 had recruited fewer than half of the qualified applicants
that the Bureau estimated it needed as of March 2nd. 2 Our
analysis did not include nine local census offices in Puerto Rico.
Figure 1: Regional Offices Are Progressing Toward Their Recruiting
Goals Statement 2000 Census: Update on Essential Operations Page 8
GAO/ T- GGD/ AIMD- 00- 119 Factors contributing to the Bureau's
successes and shortfalls vary widely across the country. For
example, managers at a Phoenix local census office told us that
they were competing for employees with the Bureau's nearby Census
Data Capture Center, which has indoor work and higher wages. In
contrast, the Phoenix local census office offers applicants the
lower- paid job of enumerating people in extreme heat. Other local
census offices, such as those in the western and southern parts of
Chicago, are better able to meet their staffing goals compared to
the northern and central Chicago local census offices, which are
located in areas with lower unemployment levels. Additionally,
according to local and regional Bureau officials we contacted,
local census offices that are conducting the update/ leave
operation have been more successful at meeting their recruiting
goals because, among other reasons, they can offer immediate jobs.
The regional and local census offices continue to aggressively
recruit applicants to ensure that their recruiting goals will be
met. For example, the Bureau is working with communities to set up
testing sites at community events, meeting with leaders of local
civic organizations to understand how to best recruit their
members, and placing recruitment advertising in community
newspapers to recruit individuals to work within their own
neighborhoods. Moreover, since November 1999, the Bureau has been
working with state governments to obtain exemptions so that
individuals receiving Temporary Assistance to Needy Families,
Medicaid, and selected other types of assistance would not have
their benefits reduced when earning temporary census income. Such
exemptions could expand the applicant pool by making census
employment more desirable to these individuals. According to the
Bureau, 36 states and the Virgin Islands have granted an exemption
for 1 or more of these programs as of March 6, 2000. Bureau
headquarters has also developed a list of actions that its
regional offices can take to help local census offices that are
not reaching their recruiting goals. These actions, which vary by
the severity of the shortfall, range from saturating neighborhoods
with recruiting advertising and targeted postcard mailings in the
less severe instances, to raising pay rates and sending in special
recruiting teams in the most dire cases. We found that several
census regional and local offices are implementing some of these
actions. For example, the Atlanta and several other regional
offices are mailing postcards to targeted zip codes that local
census offices have flagged as hard- to- recruit areas. In
addition, regional offices have authorized pay rate increases for
some local census offices. In February Statement 2000 Census:
Update on Essential Operations Page 9 GAO/ T- GGD/ AIMD- 00- 119
2000, for example, the Manassas, Virginia, local census office
raised pay rates from $11.25 to $12.75 an hour and the
Placerville, California, office raised enumerator pay rates from
$11.00 to $13.00 an hour. The Bureau has included several coverage
improvement programs in the 2000 Census that are aimed at
increasing the count of various hard- toenumerate groups. The
hard- to- enumerate include minority groups, migrant and
undocumented workers, the homeless, and those with little or no
English- language skills, among others. Such individuals are often
missed during the Bureau's initial enumeration and nonresponse
follow- up operation. Two coverage improvement programs we have
been examining are the Bureau's walk- in Questionnaire Assistance
Centers and Be Counted Program. Questionnaire Assistance Centers
are to help people especially those with little or no English-
speaking ability complete their census questionnaires by providing
assistance in various languages on a walk- in basis. Assistance
guides in 37 languages that are designed to help people complete
their English language forms, as well as large- print English
guides, are to be available at all assistance centers. Assistance
guides in 12 additional languages will be available at selected
centers on the basis of specific community needs. The centers,
which are located in various public locations, such as community
centers and libraries have been open since March 8 when
questionnaires were first distributed, and are to close on April
14, 2000, about 2 weeks before the beginning of nonresponse
follow- up. The Bureau is staffing the centers with a mix of
volunteers and paid employees. The Be Counted Program is designed
to count people who believe they did not receive a census
questionnaire or who were otherwise not included in the census.
The program also allows people who had no usual residence on
Census Day, such as transients, migrants, or seasonal farm workers
to be included in the census. Be Counted questionnaires
essentially a modified short- form are to be available in
Questionnaire Assistance Centers as well as in other public
locations for people to complete and return to the Bureau. The
program is to run from March 31 until April 11, 2000. The Bureau
anticipated setting up as many as 66,895 Be Counted locations (an
average of about one Be Counted site for each of the nation's
census tracts). Of these, 34, 725 were to be stand- alone sites,
and the remainder were to be located in each of the 32,170
Questionnaire Assistance Centers the Bureau expected to establish.
Key Coverage Improvement Programs Appear on Track Statement 2000
Census: Update on Essential Operations Page 10 GAO/ T- GGD/ AIMD-
00- 119 Bureau data as of March 1 indicate that the Bureau has
established 17,209 stand- alone Be Counted sites and 29,161
Questionnaire Assistance Centers. Table 1 shows how these numbers
compare to the Bureau's projections for each census region. Census
region Projected number of Questionnaire Assistance Centers Actual
number of Questionnaire Assistance Centers Percentage of projected
number Projected number of Be Counted sites Actual number of Be
Counted sites Percentage of projected number Atlanta 969 1,109
114% 854 643 75% Boston 1,400 1,905 136 925 1,310 142 Charlotte
1,544 871 56 6,905 702 10 Chicago 12,000 8,395 70 3,900 997 26
Dallas 1,500 2,922 195 2,000 836 42 Denver 1,500 1,476 98 2,000
1,425 71 Detroit 1,455 1,136 78 2,000 899 45 Kansas City 2,414
1,346 56 1,521 1,166 77 Los Angeles 3,000 2,231 74 3,000 2,700 90
New York 2,000 1,887 94 3,900 1,716 44 Philadelphia 1,088 1,956
180 2,220 1,575 71 Seattle 3,300 3,927 119 5,500 3,240 59 Total
32,170 29,161 91% 34,725 17,209 50% Source: U. S. Census Bureau
data as of March 1, 2000. Although short of the Bureau's
projections overall, several regions are close to or are exceeding
the Bureau's expectations, and the combined number of sites 46,370
is still substantial. Indeed, by comparison, there are about
12,600 McDonald's restaurants in the United States. As we
discussed in our February report, in addition to the number of
sites, other factors will be critical to the effectiveness of the
two programs. 3 The factors include (1) working with local
partners to identify sites with locations and schedules that best
meet the needs of targeted groups and are adequately publicized;
(2) ensuring that sites have street- level visibility so that
targeted groups are able to find them; (3) monitoring usage so
that people will be able to find forms and obtain assistance when
and where they are supposed to be available; and (4) in the case
of Questionnaire Assistance Centers, making sure staff are
available with appropriate foreign language skills. In
implementing the two programs, the Bureau appears to be taking
steps consistent with these factors. For example, officials from
the Los Angeles 3 2000 Census: Actions Taken to Improve the Be
Counted and Questionnaire Assistance Center Programs (  GAO/GGD-
00-47 , Feb. 25, 2000). Table 1: Status of Questionnaire
Assistance Centers and Be Counted Sites by Census Region Statement
2000 Census: Update on Essential Operations Page 11 GAO/ T- GGD/
AIMD- 00- 119 Regional Census Center have worked with the Los
Angeles Unified School District to open Questionnaire Assistance
sites at public schools in hardto- enumerate areas. In addition,
the Charlotte Regional Census Center has worked with the
consortium of Historically Black Colleges and Universities within
the region to recruit student volunteers to work in the
Questionnaire Assistance Centers. The census work satisfies a
public service graduation requirement. Similarly, in Detroit,
regional officials noted how they were giving priority to hiring
assistance center staff with foreign language skills. In February,
we testified that the Bureau faced formidable challenges in
ensuring the operational readiness of its four data capture
centers (DCC), as much remained to be done to complete and test
DCS 2000 4 within a very short time frame. In particular, we noted
that when we concluded our review in early January, measures of
DCS 2000's quality and maturity were mixed. Moreover, we added
that as of February, the results of important test events had not
yet been reported, last minute changes to DCS 2000's software and
hardware configurations were still being made to address a
recently discovered operational problem, and critical test events
yet to be conducted would likely identify new problems. As of
today, the Bureau reports that it has implemented the near- term
DCS 2000 system changes and completed all scheduled test events,
including an operational test at each DCC and a test of the four
centers operating concurrently. Further, as of today, the DCCs
have been operating for about 1 week, the first 2 days of which
the DCCs checked in an average of 117,525 questionnaires. This
workload represents about 8 percent of the daily workload expected
later this month when questionnaire processing is expected to
reach the production level of about 1. 5 million questionnaires
processed per day. Thus, actual operations do not yet demonstrate
the DCC's readiness to operate at expected production- level
workloads. Additionally, the information we have seen on actual
operations does not address whether recent changes to DCS 2000 are
functioning correctly. For this and other reasons discussed below,
we remain uncertain about the centers' readiness to meet the full
production- level workload anticipated to begin in the next 2
weeks. 4 Specifically, each DCC will rely on DCS 2000 to check in
questionnaires and determine which households have responded, scan
the questionnaires to produce an electronic image, optically read
handwritten marks and writings from the imaged questionnaires, and
convert the data into files that will be transmitted to Bureau
headquarters for tabulation and analysis. Production- Level
Operational Readiness of Data Capture Centers Remains Uncertain
Statement 2000 Census: Update on Essential Operations Page 12 GAO/
T- GGD/ AIMD- 00- 119 To ensure DCC readiness, the Bureau and its
contractors conducted a series of operational tests at each of the
four DCCs. These tests were conducted in an incremental fashion,
whereby improvements and lessons learned were incorporated in
succeeding tests. These tests were designed to assess a DCC's
ability to perform key operations, such as sorting, checking- in,
and scanning questionnaires at required production levels.
Although the four operational tests have been concluded, the
Bureau's report on the results of its operational test at the
Jeffersonville DCC, performed during December 1999 through
February 2000, is not yet available. However, the Jeffersonville
site manager told us that this operational test, although less
structured than the other DCC operational tests and focused on
qualitative observations rather than quantitative measures, was a
success. Although the DCCs perform the same basic data capture
operations, test results from Jeffersonville are significant
because actual operations, and thus center operational capability
and readiness, are very much people dependent. Therefore, it is
important to know the Jeffersonville operational test results to
determine if the problems identified in the Pomona DCC operational
test 5 have been successfully addressed at the Jeffersonville DCC,
as they were at the Phoenix DCC. Moreover, the Jeffersonville DCC
is unique because unlike the other DCCs, which are operated by
contractor personnel, Jeffersonville is being operated by Bureau
personnel, supplemented with temporary workers. Bureau officials
told us that the operating procedures developed by the DCC
operations contractor were provided to the Jeffersonville DCC, but
that the Bureau's own staff in Jeffersonville are not required to
use these procedures and, in fact, can create their own.
Additionally, the Jeffersonville workload mix is different from
the other DCCs, encompassing other forms, such as Be Counted and
foreign language forms, which the other centers do not process. As
a result of the DCC operational tests, the Bureau realized that
the keying rate for key from image (KFI) was not high enough to
meet its master schedule for completing Census 2000 and delivering
the apportionment counts by December 31, 2000, as required by law.
To resolve this dilemma, the Bureau adopted a two- pass approach
to data capture operations. During the first pass from March 6,
2000, to August 5 The Pomona DCC operational test, conducted in
October and November 1999, showed that the center was not able to
sort or scan the number of questionnaires expected during peak
data capture operations. The test also demonstrated that the DCC
could not achieve the keying rate necessary to handle the expected
key from image workload, which entails manually entering data
because handwritten characters or marks were difficult for the
system to recognize. Jeffersonville DCC Operational Test Recent
DCS 2000 Software Changes Statement 2000 Census: Update on
Essential Operations Page 13 GAO/ T- GGD/ AIMD- 00- 119 31, 2000
the DCCs will capture the data necessary to determine the
apportionment counts, referred to as 100- percent data. 6 During
the second pass from September 11, 2000, to November 15, 2000 the
Bureau will capture the remaining data from the long forms, known
as sample data, which include the detailed social, economic, and
housing information collected for a sample of living quarters in
the United States. Because the keyers will focus first on the 100-
percent data instead of working on both the 100- percent data, and
the sample data simultaneously, the Bureau expects to be able to
complete the 100 percent data capture by August 31, 2000,
providing enough time to produce the apportionment counts by the
end of December. To implement the two- pass data capture solution,
two sets of changes or releases were required to DCS 2000. The
first release, designed to support the first pass, was completed
in early February, according to the Bureau. This work involved
modifying DCS 2000 software to write the long form images to a
mass storage unit and to not present the sample data to keyers.
This release was not in place in time to be included as part of
operational tests at the individual DCCs. However, it was in place
for the final four- site operational test, which I discuss
shortly. Nevertheless, while Bureau officials have stated the
changes were successfully implemented, the Bureau has not yet
provided us with the procedures used for testing these changes and
the results. As a result, we do not have the information needed to
know with any certainty whether these software changes are
performing as intended. To help prepare for the actual data
capture operations during Census 2000, the Bureau and its
contractors conducted a final operational test on February 22 to
25, 2000. This test was important because it involved production-
level workloads for all four DCCs, centralized operations, and
Bureau headquarters simultaneously. However, the test was limited
because it did not include all DCC operations, such as the DCC's
ability to sort, check- in, and prepare questionnaires for
processing. Also, most of the questionnaires used in the test had
machine printed, rather than handwritten responses, with the same
answers on each questionnaire, which would simplify KFI demands.
Because of these limitations and because, as previously stated,
the individual DCC operational tests were incremental, some data
capture operations have not been verified with a production- level
workload at all DCCs. Specifically, neither the Baltimore DCC nor
the Pomona DCC have successfully conducted operational tests 6 The
100- percent data are the population and housing information
collected for all living quarters in the United States and
includes name, sex, and race of each person living in a household.
Four- site Operational Test Statement 2000 Census: Update on
Essential Operations Page 14 GAO/ T- GGD/ AIMD- 00- 119 of their
abilities to support a production load for sorting or document
preparation. The Bureau has characterized the four- site test as
successful, however, it has not yet provided us with the results.
Assessing the results is still important because the tests did
evaluate some important data capture operations, such as the DCC's
ability to scan production- level workloads of questionnaires and
transmit the resulting data to headquarters. As previously noted,
the Bureau still plans on making a second set of modifications to
DCS 2000 to support its new data capture approach. This work will
involve modifying the system to retrieve the images of the
approximately 22 million sample data questionnaires from the mass
storage unit and present those requiring action to keyers. The
Bureau plans to complete development and system- level test
activities for this modification by July 31, 2000. Although these
changes are necessary for the Bureau to implement its approach to
meeting its schedule for apportionment counts, they will require
the DCS 2000 development contractor to divert personnel from the
DCCs and the central system support facility to ongoing DCS 2000
development. The DCS 2000 development contractor identified the
diversion of personnel as a risk that could adversely impact first
pass data capture operations. In addition, the Bureau plans to
suspend operations at one yet- to- be selected DCC in August 2000
to run an operational test of the second pass DCS 2000 software
release. This will require the Bureau to redistribute this DCC's
workload among the other three DCCs. In addition to the previously
mentioned challenges, a new, unforeseen issue surfaced late last
month when it was discovered that an extra digit was erroneously
inserted in front of the house numbers used to mail advance
notification letters. Despite this error, the U. S. Postal Service
found that the postal barcodes were correctly printed, which,
according to a postal official, helped assure the letters were
properly delivered. The advance letter is part of the Bureau's
multiple mailing strategy, which is designed to boost response
rates by increasing the number of mail contacts the Bureau has
with the public. The advance letter notifies people that they will
soon receive a census form, describes why it is important to
respond, and provides an opportunity to request a questionnaire in
Spanish, Chinese, Korean, Vietnamese, or Tagalog. Ongoing DCS 2000
Changes Bureau Confident That Address Error Is Limited to Advance
Notice Letters Statement 2000 Census: Update on Essential
Operations Page 15 GAO/ T- GGD/ AIMD- 00- 119 According to the
Bureau, the misprint occurred during printing and addressing by a
private contractor. Bureau and Government Printing Office
officials believe that the source of the error was a glitch in the
contractor's printing program because, on the basis of Postal
Service and Bureau information, the error occurred on all of the
letters. According to Bureau officials, quality assurance
procedures developed for printing the advance letters and other
census documents, including census questionnaires, were not
designed to check the accuracy of the printed addresses. Regarding
census questionnaires, for example, the quality assurance
procedures were meant to identify printing problems that could
affect the data capture process, such as type size and color
defects. Once they learned of the error, Bureau officials said
that both the Bureau and contractors responsible for printing
census questionnaires checked a sample of each of the six types of
questionnaires with preprinted addresses to ensure they did not
contain the same misprint contained in the advance letters.
According to those officials, the samples came from the Bureau's
earlier quality assurance samples and from the contractors' plant
floors, and included the update/ leave short- form questionnaire
printed by the same contractor that printed the advance letters.
Specifically, Bureau and contractor staff:  visually inspected
printed addresses for the same error identified in the advance
letters,  matched printed addresses against addresses contained in
the census address file,  compared printed addresses against
addresses contained in the Postal Service address files located on
the Postal Service Internet site, and/ or  compared address
information read from the Postal Service barcodes against the
addresses printed on the questionnaires. Bureau officials also
said they checked the addresses on a limited number of reminder
cards, including a visual inspection and comparison of the
addresses against the Postal Service file. Although documentation
of the results of these checks is not available, they gave Bureau
officials confidence that the address error identified in the
advance letter is not present in census questionnaires and
reminder cards. In addition, according to Bureau officials, in the
course of these checks, no other types of errors were detected. We
independently reviewed a very small sample of questionnaires at
four post offices in Statement 2000 Census: Update on Essential
Operations Page 16 GAO/ T- GGD/ AIMD- 00- 119 Northern Virginia
and found they did not contain the same misprint present in the
advance letters. Mr. Chairman, with Census Day just over 2 weeks
away, I would like to end by briefly returning to one of the
themes I have highlighted today public cooperation. The mail
response rate the ultimate measure of cooperation-- will be
central to determining the overall accuracy and cost of the
census. The national response rate will therefore provide a first
indication of the success of the 2000 Census. A high mail response
rate will reduce the Bureau's follow- up workload and,
accordingly, relieve some of the staffing and schedule pressures
the Bureau confronts. On behalf of the Subcommittee, we will track
these and other operational data and continue to monitor the
progress of the census. Mr. Chairman, this concludes my prepared
statement. I would be pleased to respond to any questions you or
other Members of the Subcommittee may have. Contact and
Acknowledgments For further information regarding this testimony,
please contact J. Christopher Mihm at (202) 512- 8676 or Randolph
C. Hite on (202) 5126240. Individuals making key contributions to
this testimony included Thomas Beall, Mark Bird, Cristina
Chaplain, Garry Durfey, Deborah Eichhorn, Robert Goldenkoff,
Richard Hung, Lily Kim, Victoria Miller, Vicky L. Miller, Ty
Mitchell, and Lynn Wasielewski. Page 17 GAO/ T- GGD/ AIMD- 00- 119
Page 18 GAO/ T- GGD/ AIMD- 00- 119 Page 19 GAO/ T- GGD/ AIMD- 00-
119 Viewing GAO Reports on the Internet For information on how to
access GAO reports on the INTERNET, send e- mail message with info
in the body to: info@ www. gao. gov or visit GAO's World Wide Web
Home Page at: http:// www. gao. gov Reporting Fraud, Waste, and
Abuse in Federal Programs To contact GAO's Fraud Hotline use: Web
site: http:// www. gao. gov/ fraudnet/ fraudnet. htm E- Mail:
fraudnet@ gao. gov Telephone: 1- 800- 424- 5454 (automated
answering system) (410555)

*** End of document. ***