Mr. Chairman, Mrs. Maloney, and Members of the Subcommittee:
I am pleased to be here today to provide an update on the status of Census 2000 activities. Last week, I had the honor to report good news about the state of civic responsibility in our country as reflected in the final tally for our '90 Plus Five program. Some 17 percent of local jurisdictions met the demanding goal of exceeding their 1990 mail response rate by 5 percentage points. I reported that questionnaires had been returned for 65 percent of the housing units in America. This matched the mail response rate for the 1990 census and exceeded the target in our budget by 4 percentage points. More importantly, it means that the country has stopped a 30-year decline in meeting our civic responsibilities. This is a serious achievement.
In reaching 65 percent, the public out-performed expectations. More than 100,000 census partners deserve credit. Congratulations are owed to thousands of mayors, county commissioners, teachers, community advocates, houses of worship, and other tribal, local government, civic, and business leaders; to Young and Rubicam and their partner agencies for the excellent advertising campaign; and to you, Mr. Chairman, Mrs. Maloney, and other Members of Congress who have encouraged response to the census. Our partners and the public have treated the census as the serious civic event intended by the founders when they wrote the census into the U.S. Constitution.
But, unfortunately, we cannot rest on this achievement. We now move from one phase of the census to another--the very demanding task of convincing people who did not return their forms to cooperate when census-takers come to their door. While we received questionnaires for 65 percent of the housing units, that leaves more than a third of housing units that we must still visit. If we stopped now, the results could be disastrous for many communities. So, the Census Bureau will make a concerted effort to get responses from those remaining addresses.
Census enumerators were trained last week and they have begun making their rounds. These are our neighbors doing America's task. It is a difficult task. I know that from personal experience. I was an enumerator 40 years ago during the 1960 census when I was working my way through school. The job is more difficult today, in part, just because it is harder to find people at home. Households are smaller and more people are in the labor force. People are more concerned about privacy. There have been some indications that people may not cooperate, and I'll talk more about that in just a moment. So we need the help of our partners to sustain the civic momentum and commitment we saw in the first phase of the census. Those who did not meet their civic responsibilities previously now have a chance to do so by cooperating with census enumerators. I will discuss the nonresponse followup operation in more detail later.
The good news about the mail response rate is tempered somewhat by our concerns about noncooperation and the potential loss of data. As I explained at the last hearing, every question we ask in the census serves an important purpose and all have a specific federal legislative or judicial mandate or requirement. Very early this year, an advocacy group issued a press release that said: "Real Americans don't answer nosy Census questions. You can strike a blow for privacy, equality, and liberty by refusing to answer every question on the Census form except the one required by the Constitution: How many people live in your home?"
This misreading of the Constitution--which states that the census is to be conducted "in such manner as [Congress] shall by law direct"--ignores the fact that the Nation's founders directed that the census be a tabulation of the population by such characteristics as age, gender, race and household composition. Every census has been more than a simple headcount. Moreover, the advice given is a prescription not only for poor data quality but also for an increased undercount. If people don't cooperate with the census at all or just give us a number, whether when they returned the form by mail or when the enumerator visits, that will not be sufficient. Beyond the number of people at an address, we require some minimal characteristics to complete an enumeration. Otherwise, we have no way to know if we are dealing with real people. In cases where no cooperation is forthcoming, we will have to compensate by attempting to get the data through interviews with other knowledgeable individuals.
We are also concerned about potential loss of data due to opposition to the long form. There was approximately a 12 percentage point difference between the mail response rates for the long form and the short form, double the 1990 rate. We do not have data at this point about item nonresponse rates--that is, for example, how many people who mailed back the long form did not answer specific questions, such as income, disability, education, and so on.
Comments we have received give us reason to be concerned about the long form problem. Let me just cite two of these comments:
I have this day read my long form and promptly ripped it in two and burned same. Don't bother sending another as I won't fill it out nor will I pay the $100 fine.
I am refusing to complete the long form. You can arrest me if you want. But I am not going to complete it!
We are very concerned that refusal to respond fully to the census can pose a serious risk to Census 2000 data. As I previously testified, the Census Bureau would have to determine whether the data are sufficiently reliable to perform the functions expected of them.
I will now summarize some of the activities we have completed and results achieved since the last hearing.
At the last hearing, I reported on the successful completion of the update/leave operation and delivery of questionnaires in mailout/mailback areas. I will discuss these operations in more detail later.
Since the last hearing, we have completed or are progressing on schedule with several other major operations. These include:
--The enumeration of remote areas of Alaska.
--List/enumerate, which is conducted in remote, sparsely populated areas where it is not efficient to compile a precensus address list.
--Update/enumerate, which we conducted in communities with special enumeration needs and where most housing units may not have house number and street name addresses. These include selected American Indian Reservations, unincorporated Spanish-speaking communities along the border of Texas and Mexico (colonias), and resort areas with high concentrations of seasonably vacant housing units.
--Service based enumeration, in which we interviewed people at shelters, soup kitchens, mobile food van stops, and targeted outdoor locations.
--Group quarters, where we count people living in such places as college and university dormitories, hospital and prison wards, migrant farm camps, and nursing homes.
--The enumeration of land-based and shipboard military personnel and people aboard U.S. flag-bearing merchant vessels.
In data capture, all operations are on schedule and have very high quality. We have scanned about 70 million questionnaires. Optical character recognition is
working well; it has "read" over 80 percent of the write-ins with over 99 percent accuracy.
Be Counted sites closed on April 11 and we stopped accepting questionnaires on April 30. In all, we received about 573,000 Be Counted forms out of some 16 million that we distributed. About 465,000 of these were in English, 95,000 Spanish, and the balance in other languages. Questionnaire Assistance Centers closed on April 14; these centers received about 350,000 walk-in visits.
We have received over 5.8 million calls at our Telephone Questionnaire Assistance centers. TQA will operate until June 8; the number of calls has decreased dramatically and we have reduced staff accordingly. TQA staff took approximately 120,000 census enumerations over the telephone. Except for the first two days of the operation, TQA was able to service nearly all callers.
We received some 2.3 million requests for foreign language questionnaires as a result of the advance letter and we have sent questionnaires to the requesters. This is an indication that the advance letter worked.
In all, we received about 66,000 responses through the Internet. We consider this an operational success. It was the first time we had allowed this option and we wanted to see if we could do it. It worked.
At previous hearings, I discussed potential problems that could put the census at risk. They included failure to complete the update/leave operation; problems with our payroll system that would prevent us from paying our employees on a timely basis; widespread problems filling enumerator positions, despite our extensive pool of qualified applicants; problems with the Census 2000 address file that would prevent our employees from being able to fulfill their responsibilities; a major breakdown in the telephone questionnaire assistance operation; breakdowns in data capture systems or in questionnaire delivery; unexpectedly low mail response rates; any event that could undermine faith in the confidentiality of the data, such as a hacker on our Internet site; or a failure to meet our promise to provide the mailback response rate to the public. None of those potential problems has occurred. In fact, census operations have been quite successful.
Now, we enter the nonresponse followup operation, which is the largest, most complex, and most costly operation in Census 2000. Nonresponse followup raises its own set of potential risks. These would include high turnover rates for enumerators, more outright resistance from respondents that would adversely affect productivity or data quality, a breakdown in our payroll system, or random events such as attacks on enumerators or natural disasters. Turnover has been very low in early census operations, such as update/leave, but nonresponse followup is a more difficult and frustrating operation. The controversy over the long form, as I have said, gives us some reason to be concerned about resistance and data quality. Our payroll system has worked very well so far, but nonresponse followup is such a big operation that it will be the major test for the system. So we face potential risks during nonresponse followup that could affect accuracy, data quality, and budget. However, as I wrote in my letter to you of April 20, the Census Bureau has completed contingency plans to the extent feasible.
I want to reemphasize that the Census Bureau will fully apply its procedures to account for every address that is on our list to be visited during nonresponse followup. Those procedures are extensive and include making up to 6 attempts--3 by personal visit and 3 by phone (when a phone number is available)--to complete the enumeration of a housing unit. The procedures also include extensive quality assurance procedures and supervisory controls. But they also reflect our experience that the longer we are in the field and the farther we get from Census Day, the more the quality of respondents' answers deteriorates. It is important to keep in mind that we are using a part-time, temporary staff to which we have been able to provide only basic training in survey methods. Extending nonresponse followup beyond the already extensive level of effort we plan would not only increase census costs, it could lead to a reduction in data quality.
To implement nonresponse followup requires massive logistical operations. We began with a universe of about 120 million addresses in mailout/mailback and update/leave areas. As questionnaires were returned to us, we checked them in against the Master Address File to note those that had returned a questionnaire. Nonresponse followup is concerned with those addresses that have not returned a questionnaire, so after April 11, we generated nonresponse followup assignments for each of our 520 local census offices. The local census offices loaded these new files into their databases and printed file assignment directories, address listings, and labels for each questionnaire that will be completed in nonresponse followup. After April 18, we produced yet another set of files for each local census office that contained lists of those addresses for which a questionnaire had been checked in between April 11 and April 18. These are what we call late mail returns. Clerks in each local census office were assigned to cross out by hand the late mail returns on the nonresponse followup address listings.
This massive operation of organizing address listings was just part of the preparatory work the field offices had to do to get ready for nonresponse followup. They had to refine their staffing levels based on the actual mail response rates for small geographic areas. Maps, training kits, and supplies for the enumerators had to be distributed and sometimes redistributed based on the workloads for specific areas, which, as I said, could not be known until we had generated the nonresponse listings.
All of this work was very time-sensitive; it had to be completed in a few days so we could begin training for nonresponse followup on time. While it was going on, we continued to receive mail responses. Some of these made it into the late mail return files, but some did not. Some people who mailed back their census form will be visited in nonresponse followup. We realize that this will irritate some of the public who will wonder why we are bothering them again. All we can do is explain why this is largely unavoidable given the magnitude of preparing for nonresponse followup, a process that I described earlier. We have to have a cutoff date to begin preparing the assignments and to get all the maps and kits to the right training sites. We do the best we can to strike the late forms that come in from the nonresponse followup universe. Forms are still coming in as we speak. We have also received many forms, such as Be Counted forms, that do not have identification codes on them so they can be checked in quickly. We have to undertake a labor-intensive matching and place coding operation to code them to the right geographic area. So, this sometimes correct complaint that "I already sent the form in" is something our enumerators are trained to deal with. Of course, they will try to complete an enumeration at these housing units anyway, because many people will say they have returned a questionnaire, even when they haven't, and the enumerators will have no way of knowing. Their job is to get a completed questionnaire for every housing unit on their list.
To get ready for nonresponse followup, the Census Bureau conducted an aggressive and highly successful recruiting effort. We have sufficient staff to begin nonresponse followup on schedule in every local census office in the country. We front-loaded our training selections, which means that our goal was to train and give assignments to twice as many people as we needed. That way we will have staff to offset attrition. We have also identified over 50,000 individuals for replacement training so that we can keep replenishing the pool of available workers. We attained this two-to-one redundancy at the vast majority of sites.
We are continuing to recruit in targeted areas even as we speak. This may mean, in the end, that some qualified job applicants may not be hired. We realize they may be disappointed, but we believe we must keep the applicant pool active to assure we have sufficient staff to cover attrition. Thus far, we have identified 2.6 million qualified applicants or 108 percent of our goal. We used more than 5,000 print ads, over 160,000 radio spots, 25,000 bus posters, and over 150 million flyers and brochures. We generated 6 million calls to our toll free jobs line and over 10 million hits on our Internet job site. Our aggressive recruitment strategy worked.
As I said earlier, census takers are in the field now visiting housing units. I urge the public to cooperate with them and provide the information that is being requested. These are not professional survey takers. They are not sales or marketing people. They are part-time workers, our neighbors, who accepted the challenge to do this very difficult job for America. There are several ways to identify a census taker. Census 2000 enumerators will carry a red, white, and blue badge with their name on it; they will carry a black and white tote bag with a census logo on it and census forms; they will offer printed statements about the confidentiality of the data; they will not ask to come into anyone's home; they will give a local census telephone number to verify that they are a census enumerator; and they will not ask for credit card numbers or for a Social Security number.
To place nonresponse followup in context, Appendix 1 graphically depicts each of the major census enumeration operations that precede and follow it.
Mr. Chairman, I will now answer the specific questions in your letter of invitation for this hearing.
The final status of nationwide mail response rates and what those rates mean for the Nonresponse Follow-up (NRFU) workload, other operations, cost savings, and any potential reallocation of funds.
The Census Bureau is redirecting resources and personnel from areas with high response rates to areas that will need more attention in the nonresponse followup operation--hiring additional enumerators in these areas and, in a few areas, raising pay rates to meet hiring goals. In addition, we also are examining local outreach and promotion efforts to determine whether there are additional opportunities for targeting advertising and promotion in communities with high nonresponse populations. The General Accounting Office's estimate of a savings of $34 million for each percentage point above 61 percent is based on our planning assumptions about the productivity and turnover rates of census enumerators. As I have explained earlier, turnover and productivity are two of the serious potential risks that we could face in nonresponse followup. We will assess nonresponse followup as it progresses. We will not know the actual cost of nonresponse followup until we have analyzed turnover and productivity; only then will we know if there will be any savings.
Operational difficulties in Tampa, Florida, Albany, Georgia, New York, New York, and Chicago, Illinois.
I have addressed the issues related to Tampa in a letter I sent you earlier this week. I will answer any specific questions you might have about Albany, Georgia, New York, or Chicago.
The final status and assessment of update-leave and mail delivery.
We began the update/leave operation on March 3 and completed it by the end of March, as planned. Census enumerators left questionnaires at approximately 24 million housing units in areas including Puerto Rico. Local census offices conducting update/leave operations had sufficient staff and turnover was low. From March 13 through March 15, the U.S. Postal Service successfully delivered questionnaires in mailout/mailback areas of the country. There were some instances where households received duplicate questionnaires. This occurred because during all the overlapping processes used to build the Master Address File, we wanted to minimize the chance that we would eliminate an address that should be retained. We have procedures in place to remove these duplicate addresses from our files before the final census data are tabulated. There were also some small towns where residents receive their mail at a box at the Post Office and the U.S. Postal Service does not deliver questionnaires to their home. In some of these cases, local census workers distributed questionnaires and some were returned. We will visit the rest of these housing units during nonresponse followup.
The assessment of advertising and local partnership efforts for the mail out/mail back phase and plans for advertising and local partnership activities for NRFU.
As I stated earlier, the efforts of our 100,000 partners and the success of our advertising campaign certainly contributed to our ability to reach a 65 percent mail response rate. Formal evaluations of both the advertising campaign and the partnership program are planned for completion at a future date. We will report these findings when they are available.
The original plan for Census 2000 advertising called for 4 weeks of nonresponse followup advertising, beginning April 17 and running through mid-May. Ads will be seen on television, radio, and in print across the country. A number of new ads were created for this phase to deliver the message about the benefits of the census as well as to inform the public that Census 2000 is not over. The ads tell residents that census takers will be in neighborhoods to assist those who did not fill out their forms, and they encourage residents to check for official census identification badges and to participate in Census 2000.
To insure that the public continues to be aware that the census is not over, the Census Bureau has decided to use savings from cost efficiencies in the early phases of the advertising campaign to extend the original 4 weeks of nonresponse advertising by an additional 4 weeks, reaching into early June. In weeks 5 and 6, ads will be placed in the same markets and outlets as during the first 4 weeks (with the exception of television, which was sold out). Weeks 7 and 8 will include a targeted radio campaign reaching areas of the country with low response rates. Staff are analyzing initial response rate data to determine the target markets for this additional advertising.
We also continue to involve our partners in census efforts. A second component of the "How America Knows What America Needs" program (the first component was the '90 Plus Five program) is called Because You Count. We have made available through our Internet site various reproducible materials our state, local, tribal, and community partners can use to increase cooperation with census enumerators when they come knocking on doors. We have distributed to each Member of Congress a packet of materials about the Because You Count campaign to help you answer constituents' questions during nonresponse followup. We have also distributed to Members public service announcement scripts and videos you may use to produce public service announcements for nonresponse followup.
Hard-to-enumerate plans.
Mr. Marvin Raines, Associate Director for Field Operations, and other Census Bureau staff briefed congressional, Monitoring Board, and General Accounting Office staff on April 25 about the hard-to-enumerate plans. I am attaching for the record a paper copy of the slides (Appendix 2, "Developing LCO Tract Action Plans") that were used in the briefing. Just to summarize this information, using a variety of demographic and operational variables from the 1990 census, Census Bureau staff identified 1990 census tracts that were hard to enumerate, difficult to recruit in, or had some other special situation that would require the application of special enumeration tools. Because the data in our planning database were nearly 10 years old, it was left to the experience of field staff to make the final determination of what tracts would pose problems and, more importantly, what tools would be needed to overcome them. This information about each of the more than 60,000 tracts is displayed in detailed Tract Action Plans for each of the 12 census regions and in a national summary. These materials were also provided to the staff at the briefing.
The tools available for use included establishing Be Counted and Questionnaire Assistance Center sites, providing bilingual enumerators and/or local facilitators, using special enumeration procedures, such as update/enumerate, and having enumerators work in pairs or larger teams to conduct their work. In all, about 39 percent of all census tracts were designated as hard-to-enumerate, and a subset of these also had either recruiting problems or other special situations. Not all concerns were concentrated in specific census tracts; some were general problems throughout an entire local census office area, and these were also accounted for in "special considerations plans."
The number of undeliverable forms and status of forms.
Let me begin by thanking the Postmaster General and his team at the U.S. Postal Service for the excellent job they did in the delivery of questionnaires in mailout/mailback areas. As we expected, the U.S. Postal Service could not deliver some forms because of incorrect ZIP codes, vacant housing units, lack of residential delivery in the area, or other reasons. These undeliverable-as-addressed packages are usually returned to the Census Bureau's National Processing Center. For selected pre-identified ZIP codes, the U.S. Postal Service, at our request, held the packages for pick up by the local census offices. The local census offices attempted to redistribute the packages to individual households for the pre-identified ZIP codes. Out of about 9 million undeliverable forms in all, the local census offices completed redistribution of about 1.6 million forms. Targeted ZIP codes were selected based on the projected numbers of undeliverable forms in the ZIP codes. The valid addresses for the undeliverable forms that were not delivered by the local census offices will be included in nonresponse followup; some of the undeliverable forms were for addresses that were determined to be invalid and were deleted from the master file.
The status of LCO hiring (not recruiting) for NRFU as of May 1, 2000.
As of May 2, over 360,000 enumerators had begun work on nonresponse followup.
The reported computer problem which resulted in the omission of surnames information for NRFU operations.
I notified the Congress of this problem on April 18 as soon as I had adequate information to describe it accurately. When nonresponse followup materials were delivered to the local census offices, Census Bureau staff discovered computer programming problems that resulted in the omission of surnames for responding households at those selected addresses for which we had intended to include surnames. These surnames are an important tool that enumerators can use in cases where they encounter delivery problems in multiunit structures, such as apartment buildings, as well as in trailer parks. These circumstances can result in mail delivery in a pattern inconsistent with the Master Address File. If this has occurred and respondents have provided surname information, this information can assist enumerators in resolving discrepancies.
The Census Bureau produced a supplementary file of the surnames and printed supplemental listings so that enumerators have all the information they need to do their job. While this solution mitigated the problem, the listings will be slightly more cumbersome for the census enumerators to use. This may negatively impact their efficiency in some cases. Direct costs incurred as a result of this error are associated with the extra paper used to print the supplementary files and overtime pay for employees in a few cases. The Census Bureau estimates that the total cost will be between $600,000 and $700,000.
The Census Bureau has expanded efforts to examine all aspects of the procedures for evaluating systems and operations. However, this review was not complete prior to the discovery of the surname problem. Steps are being taken to expand and effectively target quality review procedures to ensure that problems are kept to a minimum. However, while the Census Bureau's goal is to be error-free in all operations, there are still going to be problems in an operation this large that get through even the most stringent review procedures.
All Census 2000 operations are integrated and based on computer programing to some degree.
It merits emphasis that the computer program used to produce the nonresponse followup materials is but one of 2,500 computer programs that are being implemented in connection with Census 2000. The successful, timely completion of many Census 2000 operations that I mentioned earlier is evidence that systems are functioning as expected. We are continuing to closely monitor all Census 2000 operations, and we will keep you informed of any significant difficulties we face in the months ahead.
Census Bureau surveys requiring a Social Security number and reports of current Social Security fraud.
There have been a number of news reports about persons illegally pretending to be Census 2000 enumerators so that they can obtain personal information from individuals, such as credit card numbers, other financial information, or Social Security numbers. The Census Bureau never requires a Social Security number, and I will explain this in more detail later. To help avoid any confusion on the part of the public, we are emphasizing the ways that individuals can know if the person who comes to their door is really a census enumerator. I described those ways earlier, but let me reiterate them. Census 2000 enumerators will carry a red, white, and blue badge with their name on it; they will carry a black and white tote bag with a census logo on it and census forms; they will offer printed statements about the confidentiality of the data; they will not ask to come
into anyone's home; they will give a local census telephone number to verify that they are a census enumerator; and they will not ask for credit card numbers or Social Security numbers.
The U.S. Census Bureau is implementing a series of experiments during Census 2000 to measure the effectiveness of new techniques, methodologies, and/or technologies in order to form recommendations for subsequent testing and possible innovations in the design of the next decennial census. One of these experiments is investigating how the public responds to requests for Social Security numbers on census questionnaires. During this experiment, approximately 20,000 households have received a special census form and a letter asking them to voluntarily provide their Social Security number. It is important to stress that this is being done in the mailout/mailback phase of the census, that provision of the Social Security numbers is voluntary, and that no households are being asked to provide their Social Security number as part of the Census 2000 nonresponse follow-up operation.
In a very few instances, current survey enumerators may ask for Social Security numbers, on a voluntary basis, as part of ongoing surveys, not as part of Census 2000 followup. The Census Bureau conducts a number of household surveys, many on a reimbursable basis with other government agencies, under the authority and provisions of Section 182 of Title 13, United States Code. This law requires the Census Bureau to keep all information about the respondent and their household strictly confidential and to use the information only for statistical purposes. In addition to Title 13 requirements, the Privacy Act (Section 552a of Title 5) and the Paperwork Reduction Act (Chapter 35 of Title 44) also require the Census Bureau to state the purposes of the data being collected and the fact that these surveys are voluntary. This is accomplished by sending each sample household an introductory letter prior to our interview by telephone or personal visit. Respondents in the following ongoing surveys are asked to provide the Social Security numbers on a voluntary basis:
Section 6 of Title 13, United States Code, specifically authorizes the Census Bureau to acquire data from other agencies instead of conducting direct inquiries. The Social Security numbers collected in these surveys permit us to combine survey responses with their corresponding administrative data for program evaluation and enhancement. As the Privacy Act requires, respondents are told that the Social Security number is being collected so that information from other agencies may be combined with their survey responses.
The primary use of Social Security numbers in these surveys is in evaluating the accuracy of income and program participation data that we collect in these surveys. The Social Security numbers have enabled us to evaluate how well our surveys measure wage and salary income, pensions, and interest income. Since these measures are used in calculating the official poverty rate, it is important that we understand any errors or shortfalls in survey reporting that we may be able to correct. Another key use of Social Security numbers is to augment information collected in the survey, by adding historical information or data collected in a different context. This additional information enriches the data we collect without adding to respondent burden or data collection costs.
All of these data sharing arrangements are conducted under strict security guidelines, and the individual records are protected from unauthorized use just as the survey responses are protected. Only sworn Census Bureau employees can see data that could identify an individual.
Mr. Chairman, that concludes my testimony. I will answer any questions you may have.