NOVEMBER 9-10 2004 MEETING OF THE DECENNIAL CENSUS
ADVISORY COMMITTEE
As described in an October 4 APDU email message, the first day of the fall meeting was a joint meeting of all census advisory committees devoted to the “data privacy, confidentiality, and dissemination” issues surrounding the Census Bureau’s provision of data on Arab American ancestry to the Department of Homeland Security (DHS).
Flanked by high ranking Census officials at the head table, Census director Louis Kincannon acknowledged to the full conference room (and overflow spectators watching on video in the auditorium) that the DHS episode has intensely disturbed some people, and noted that the Census Bureau is very concerned with this perception. He assured the assembled representatives of the high importance that the Census Bureau gives to protecting the confidentiality of census responses.
Kincannon described the episode as one in which the Census
Bureau provided data on Arab ancestry to DHS – twice. In response to a first request, Census provided data for places
of 10,000 or more population, and in response to the second, Census provided
similar data for ZIP Code Tabulation Areas (ZCTAs). In both cases, the data were already available to the public on
the Census Bureau’s website. And in
contrast to “special tabulations” in which the Census Bureau accesses
confidential detail files to produce aggregations not previously published, the
data provided to DHS were “extracts,” of data that had already been
published–and reviewed to ensure against disclosure. Confidentiality is not the issue, but
rather perceptions and sensitivity.
In response to the controversy that followed these transactions, the Census Bureau has implemented “interim” procedures for responding to such data requests. The procedures require Census employees to seek to identify the identity of the requestor, and if the requestor is affiliated with a law enforcement or intelligence agency and/or the request involves sensitive populations, the request must be referred to senior census staff (associate director level or above). The director stressed that these procedures are not designed to restrict access to statistical information. The Bureau already had procedures in place to monitor response to requests for special tabulations, and now has procedures for responding to requests for “extracts” of published data.
Kincannon asserted that the Bureau does not judge data users, and that data are equally available to all. However, the Bureau realizes the nature of the world we live in, and that is why the new procedures have been adopted. He emphasized that the procedures are interim – that the Bureau is not aware of their full implications, and wants to hear from data users.
Scheduled for one hour, the Q and A session ran long as a great many advisory committee representatives waited their turn to ask questions and offer comments.
The representative from the National Asian Pacific American Legal Consortium (NAPALC) expressed appreciation for the Census Bureau’s prompt response, but noted that the interim procedures do not specify what the Census Bureau would look at, and what data they would provide. The representative also likened the provision of Arab ancestry data to the 1942 tabulations that contributed to the internment of Japanese Americans during World War II. Saying he was not in a position to discuss 1942, director Kincannon explained that, under the new procedures, the Bureau would decline to provide data if it would compromise public confidence in and response to the census. He specified that a request for small area data on Arab ancestry—that might not have been sensitive in 1999—would now be denied, although the Bureau might refer the requestor to the Census website or other sources for the already published data.
The rep from the Population Association of America interjected a comment on the precarious ACS funding situation – noting the stake that those in the room have in the ACS, that final budget decisions are expected soon, and that this was our last chance to communicate support for the ACS.
Returning to the DHS controversy, the rep from the National Association of Latino Elected and Appointed Officials Educational Fund echoed the Asian concerns, asserted that confidence in the confidentiality of the census has been shaken, and expressed concern that circumstances are such that we could repeat what happened in 1942. The rep also asked for a Census Bureau expression of awareness that data on individuals cannot be used to harm them. Kincannon observed that just as aggregate data can be used to help groups, they can also be used to their disadvantage. Asked for a clearer definition of “sensitive” groups, the director said these would include all official minority groups, but that judgment would determine which ancestries are sensitive. Kincannon identified Scotch-Irish as an ancestry that probably would not be considered sensitive.
The American Civil Liberties Union rep charged that the Census Bureau’s assurances that it had not been approached by enforcement agencies were false, and that the Census Bureau has not taken this issue seriously. Your APDU rep’s recollection is that the Bureau’s assurance was that they had not been approached about providing data protected by Title 13—an assurance that still stands. When asked about this distinction, the ACLU rep reported that he had had other conversations with Census on the subject. The rep also commented that the Census Bureau has done great damage to public trust, argued that the public makes no distinction between special tabulations and extracts, and criticized the director’s side-stepping of the 1942 analogy—pointing out that “for Arab Americans, this is 1942.”
Kincannon responded that, with the interim procedures, the standards for special tabulations now apply to extracts, but argued that there is an important distinction between special tabs and extracts. The data provided to DHS were extracts of data already available to all, and the director noted that no one protested when these data were made available to the public.
The ACLU rep also raised an issue with the Census Bureau’s efforts to estimate undocumented immigration. The concern is that such estimates could be perceived as contributing to enforcement activities. The director explained that the shortfall of the 2000 population estimates traced largely to the underestimation of immigration, and that improved estimates of undocumented immigration are critical to improving the estimates of total population. He went on to note that the estimates would be at national level only, and were not thought to be sensitive. Many advisory committee reps did not agree, and the issue was raised repeatedly through the remainder of the meeting.
A rep from the Hispanic Advisory Committee stressed the importance of continued feedback from the advisory committees, and called for the establishment of a privacy and confidentiality group at the Census Bureau. Kincannon noted that the Bureau cannot pledge to not provide published data, in that there are sensitive data already on the website, but assured that they would try to direct users to other means (such as American FactFinder) for acquiring such data. He also noted that a separate group on privacy and confidentiality is planned, and could start meeting as early as March.
The rep from the American Foundation for the Blind asked why DHS was not directed to get the Arab ancestry data from the Census website, and commented that the transaction gives the impression that DHS was given preferential treatment. The director responded that the Census Bureau has worked hard to establish a culture of customer service, and that this is usually good. He also acknowledged that website’s American FactFinder function is not always as user-friendly as it could be. Based on recent experience, your APDU rep concurs that American FactFinder can be a poor tool for acquiring large census files, but also that the Bureau has become more customer focused, and that the DHS extracts were not necessarily an exceptional service for a preferred user.
It was not possible to make this or other points in the meeting, as by this time, there was a very long list of reps waiting their turn to weigh in on the topic of the day.
The Privacy Rights Clearinghouse rep asked if the new procedures had been used yet, and whether they had been discussed with outside groups. She also expressed concern that Census surveys such as SIPP and the CPS collect respondent Social Security numbers. Kincannon said that about a dozen requests had been reviewed under the interim procedures, but none had reached his level. He also explained that Social Security number is asked in some voluntary surveys, and used for matching purposes. He described this as a question for public debate, but noted that it does not relate to the Decennial Census, which does not ask for SSNs. Still, there is concern that any controversy over SSNs in voluntary surveys could impact cooperation with the census.
In contrast to the NAPALC rep, who wondered whether the Census Bureau should continue publishing ancestry data at the ZIP Code level, and an Asian Advisory Committee rep’s focus on the small cell totals in some ancestry tabulations, the rep from California Rural Legal Assistance made the point that response to this issue is not to call for the discontinuation of data collection and publication. Rather, she stressed that the perception of confidentiality is as important as the reality, and with this in mind, asked if the Census Bureau’s response to DHS would have been different. The director responded that now they would explain how the Bureau’s direct provision of these data to DHS could undermine confidence in the census, and direct them to the website for the data. The perception issue resonated with other reps, with some drawing the link to perceptions related to the estimation of undocumented immigration.
The Q & A session ended with the Census 2010 Coalition (Americans overseas) rep’s tongue-in-cheek expression of regret that there are no data on overseas Americans to misuse. He also established that very few reps had seen the DHS extracts, and suggested that those most critical of the transaction should at least have looked at the data.
At this point, the advisory committees held breakout meetings. In our DCAC meeting, the U.S. Chamber of Commerce rep spoke first – raising the issue of ACS funding, and describing the opportunity to express support through a letter drafted by Terri Ann Lowenthal. There was talk of formal reaffirmation of ACS support from the DCAC to the Commerce Secretary, but the thought was dropped on advice that the situation is now in the hands of the appropriators. The ACS discussion was a reminder that some DCAC reps were concerned that the all-day meeting on the DHS episode left little time to address other pressing census issues.
When the discussion turned to confidentiality, the Hispanic Elected Officials rep submitted a resolution calling on the Census Bureau to guard against releases of special tabulations or extracts that could harm, or be perceived as harming, race or ethnic groups—thereby undermining confidence in the census. The resolution also called on the Bureau to eliminate real or perceived uses of census data by enforcement agencies for the purpose of targeting groups by race or ethnicity. Your APDU rep and others questioned how one could hold the Bureau accountable for the applications to which published data are put. The response from some was that the Census Bureau must be held to a higher standard, but the resolution was tabled for revision and discussion the following day.
In considering the “higher standard,” one rep suggested that we reconsider what data are made available, but the prevailing sentiment was against limiting data collection and publication. Your APDU rep then asked: if eliminating transactions between Census and enforcement agencies does not go far enough, but reducing data collection and publication goes too far, what intermediate steps could be taken? There was no immediate response, but the discussion soon turned to the possibility of a Census Bureau Chief Privacy Officer (confidentiality might be a better term). Notably, there had been no mention of the Census Bureau’s Disclosure Review Board (DRB), but that is appropriate since disclosure is not at issue. Perception and sensitivity–issues presumably extending beyond the DRB—are at issue, and the recommendation is for a Census position with widely recognized responsibility for these concerns.
The joint meeting was then reconvened, and the individual committees reported on their breakouts.
The rep noted the history of research on the American Indian population without its consent, and reported the committee’s recommendation that tribes and tribal governments be informed when data on them are provided in tabulations and extracts. The rep also described the loss of trust that resulted when a 2000 census LUCA address list for an Indian reservation was sent to a nearby law enforcement agency that had no jurisdiction.
The rep reported that the committee is requesting a copy of the data provided to DHS, and is concerned that some ZCTAs may have very small numbers of Arab Americans. The group was under the impression that small numbers would be suppressed. The committee recognizes that the Census Bureau must balance user friendliness with data protection, but wants to know who is using census data and for what purpose. They called on the Bureau to better assist users in getting data from the website, and indicated the committee’s willingness to provide feedback on specific requests at any time.
The committee expressed awareness that we are in a new era of national security concerns, but stressed that the Census Bureau must aggressively protect data covered by Title 13. They hope to work with the other REACs on a joint statement on this matter.
The rep praised director Kincannon for a masterful job in showing how difficult the Census Bureau’s job is with respect to issues of confidentiality (and one might add, perception and sensitivity). The rep spoke of the Hispanic community’s stake in the census, and the tremendous help the Census Bureau had provided to this community. But, gesturing toward the head table, he was critical of the fact that senior Census staff still are not representative of the US population in terms of race and ethnicity.
The rep’s remarks were largely an expression of the need for a more complete census count of this small and scattered population.
This committee focused on two issues: 1) the generation of tabulations from published data and 2) the Census Bureau’s role in the generation of these tabulations. In their view, it is not possible to control how published data are used, and the only alternative is to limit what data are published. The rep described this option as one where the disadvantages outweigh the advantages. The committee recommends a re-evaluation of the Census Bureau’s mission with respect to special tabs and extracts, and called for greater transparency—for example, the public logging of tabulation requests. They also noted the problem that an acceptable tabulation one day might be “sensitive” the next.
DCAC Vice Chair Sonny Flores reported on the DCAC breakout meeting (described above).
The day concluded with director Kincannon thanking the advisory committee reps for what he called a useful meeting. He commented that the Census Bureau will continue communicating its commitment to confidentiality, and asked that the advisory reps keep sending suggestions, and that we “spread the facts, not the rumors.”
Your APDU rep’s take was that the meeting conveyed how deeply disturbed some groups were by the Census/DHS transaction. Some of the more critical reactions portrayed this as a threat to confidentiality, and the Bureau’s rush to reaffirm its commitment to Title 13 may have reinforced this notion. But disclosure is not the issue, and most reps seemed to get that. The critics were most effective when stressing the importance of perception, and that even aggregate data can be used to harm sensitive populations. The clear message was that the Bureau must go beyond disclosure avoidance in order to maintain public confidence in the census, and it appears that the Bureau hears that message. So a debate that started on familiar turf amid talk of “confidentiality” and “data on individuals,” succeeded in moving toward the more relevant, but also more vaguely defined realm of perception and sensitivity.
Decennial Update
Jay Waite started day two with an update of Census 2010 activities – reporting that things are moving forward despite the uncertainty over funding. Waite described the number one goal as a fully tested 2010 census, and went on to describe numerous tests that are completed or in the works.
Numerous 2004 tests involve evaluations of hand held computers (HHCs), where early results suggest that HHCs are usable for nonresponse follow up, and that field staffs are comfortable using them. It remains to be seen how HHCs will affect field staffing, data quality, and the need for imputation. Other evaluations focus on the reduction of duplication in the Master Address File, person de-duplication, the effect of new race and Hispanic origin questions, the development of the special place/group quarters frame, and procedures for collecting global positioning system coordinates using HHCs.
Next Waite described the 2005 national field test, which will test modified race and Hispanic origin questions, the improved presentation of residence rules, and the identification of households with complex living situations. A 2006 field test will evaluate improved enumeration methods for American Indian reservations, as well as heavily Hispanic areas.
Turning to the ACS budget, Waite recapped that the original request was for $165 million to achieve full implementation. The House mark of $146 million would allow full implementation (with some limitations), but the Senate mark of $64 million would not. As Waite described it, if the final appropriation is not close to the House mark, the Census Bureau will shut the ACS down, and start planning for a 2010 census with a long form. Part of the urgency relates to the need for advance testing of a long form design, if that is to the way things go. According to Waite, the Bureau estimates that a 2010 census with a long form would cost at least $1.4 billion more than a short form census with the ACS. The key, he said, is that it would not be more expensive in 2005.
Waite said he was optimistic that we would get the ACS, and noted that planning and data collection are going ahead—with new counties and Puerto Rico being added to the ACS sample, and MAF/TIGER enhancements on schedule for completion by 2008.
Late word (11/20/04) is that the final ACS appropriation was $146 million. Funding for future years is seen as a challenge, but for now, the ACS will proceed with the full 3 million annual sample, although without data collection in group quarters facilities. Congressional staff credit the letter writing efforts of data users as contributing to this outcome.
When asked about a recent GAO report that is critical of the ACS at a critical time, Waite commented that the report’s title is more negative than its content, and that the author has said the report is not intended as an argument against proceeding with the ACS. Still, the Census Bureau disagrees with most of GAO’s criticisms. For example, the Bureau disputes that it has made insufficient comparisons with long form census data, and that it has not determined the source of ACS controls. The Bureau agrees with GAO that it needs to better prepare users for the transition to five-year average data.
With respect to the prospects for a count of Americans overseas, Waite observed that the 2004 test yielded disappointing results, and there is not likely to be money for further testing.
Cognitive Test Results for Race and Ethnicity
At the spring DCAC meeting, the Census Bureau introduced a proposal to ask race and ethnicity with a set of three questions—an abbreviated Hispanic ethnicity question, an abbreviated race question (with the “Some other race” category eliminated), and a write-in ancestry question. Eleanor Gerber described the cognitive tests that have been completed on this option.
The test consisted of 100 cognitive interviews with diverse respondents. Gerber described the results as encouraging, finding that the new questions generally work well for all groups, but with some differences. For example, Hispanics tended to answer the Hispanic and ancestry questions but not race, while African Americans tended not to answer ancestry. The wording of the modified ancestry question may need additional work, as the instruction “Provide more detail” has respondents providing very long answers. The term “ancestry” works well, but respondents tend to associate “ethnic origin” with exotic places to the exclusion of mainstream groups. A note preceding the questions was found patronizing by some, and the instructions in the race question were frequently not read or understood.
Asian groups continue to express objections to the three-question approach—the concern being that the number of persons reporting detailed Asian race categories would be reduced. However, Gerber pointed out that cognitive tests cannot confirm this impact, so we will await the fields tests.
Census director Kincannon said again how productive he thought the first day’s meeting was, and asked for feedback on what the advisory committee reps need from the Census Bureau to defend the census to their constituents. He also announced that the DCAC, and other advisory committees would meet as scheduled in the spring, but would then meet only once a year. The reasons, he said, relate to the timing of the census test results, the fact that the budget numbers are often no longer available for the fall meeting, and the slowed pace of activity at mid decade. Asked if we could expect a return to twice a year meetings when the 2010 pace picks, Kincannon said yes, and in response to a question about the DCAC working groups, Jay Waite said he expects no impact – that they will probably tap the working groups more than once a year.
Ursula Wojciechowski, majority staff with the House Subcommittee, observed that the test of the overseas count did not come out in favor of such a count—with few responses achieved at a cost of $1,500 per response. She noted that Congress is open to alternatives such as the use of administrative records, but at this point cannot guarantee that such a count will be part of the 2010 census. Wojciechowski acknowledged the suspense over ACS funding, and thanked users for their letters of support. She stressed that the ACS still has strong supporters in Congress and the White House, and said they are hoping for a favorable resolution. She also remarked that the subcommittee has been pleased with the Census Bureau’s response to the DHS/confidentiality issue.
David McMillen, minority staff with the House Committee, commented that the only good news on the budget is that we would probably have an answer in the week following the meeting. He congratulated the Census Bureau for holding the confidentiality meeting—noting that it is very important that such issues be addressed before they hit the newspapers. McMillen noted that the policy office must continue to balance the need for statistics with that of confidence, and asked rhetorically if we really need to have state estimates of undocummented immigration by age. Looking to 2010, he urged advisory committee reps to pay close attention to the plans for LUCA (Local Update of Census Addresses), and the address list development effort in general. As McMillen put it, “if you can’t find the houses, you can’t find the people.” He finished by describing the move to one DCAC meeting a year as unfortunate, and quipped that he is glad things have slowed down enough that we can take a break.
In our last order of business, the Hispanic Elected Officials rep submitted the revised resolution from the previous day. The revised resolution still calls on the Bureau to work to prevent data disseminations that could undermine confidence in the census, but no longer calls on the Bureau to eliminate applications by enforcement agencies involving race/ethnic targeting. A recommendation that the Census Bureau consider establishing a Chief Privacy Officer was added. The resolution was passed, as was a separate resolution encouraging the Bureau to implement partnership and outreach campaigns as soon as possible to ensure the success of the 2010 census.
With no speakers wishing to present in the period for public comment, the meeting was adjourned.