Minutes of the December 8, 2006 COPAFS Meeting

COPAFS chair Sarah Zapolsky started the meeting, and introduced past chair Don Muff for the election of the new board and officers. The slate included Ralph Rector (chair), Judie Mopsik (vice chair), Maurine Haver (treasurer), Felice Levine (secretary), Sarah Zapolsky (past chair) and at large members Joseph Garrett, Ken Hodges, Jerry Fletcher and Don Muff. The slate was approved, seconded, and voted in.

Ed Spar. Executive Director’s Report.

Ed Spar started with the budgets, but noted not much had happened since the September meeting. Latest word is that the continuing resolution will carry through February 15, but that OMB has recommended the Census Bureau for an “anomaly” or exception that would permit an increase (or reduced reduction) in funding to allow for the purchase of the hand held computers for use in census field work. A proposed Senate cut of $20 million in the Bureau of Justice Statistics budget would kill the Crime Victimization Survey, but ironically, funding under the continuing resolution is enough to keep it going. Spar described the NCHS budget as flat-lined, and explained that this could cost a month of data collection in the National Health Interview Survey.

Spar noted that Cynthia Glassman has been appointed as the new Under Secretary of Commerce for the Economic and Statistics Administration with oversight of the Census Bureau and the Bureau of Economic Analysis. He then described growing concern with how the census will (or will not) handle same sex marriages. There is interest in data on same sex marriage (for example, among sociologists), but because the census does not determine where or when respondents were married, it cannot determine whether the marriage would be legally recognized. The fact that the census does not plan to tabulate such data may become a source of some controversy.

Spar then put in a plug for next year’s Research Conference (organized by COPAFS), and noted that plans are underway for a conference on the “hot topic” of survey incentives—sometime in 2008. Dates for the 2007 COPAFS meetings are March 9, June 8, September 14, and December 14.

Census director Louis Kincannon and deputy director Hermann Habermann, who both resigned in mid November, were in attendance, and Spar finished with personal remarks of appreciation for their exemplary public service. Spar credited them with bringing rigorous thinking and excellence to the Census Bureau, and expressed that they will be greatly missed.

A U.S. Census Bureau Update.

Louis Kincnnon. U.S. Census Bureau

Having received Spar’s flattering comments, and the hearty applause of COPAFS attendees, Kincannon said he appreciated the remarks, but expressed some unease with the “succession of funereal events” he has attended in recent weeks. He acknowledged that he and Habermann are leaving the Census Bureau, but said he had no thoughts on possible replacements, or how soon replacements would be made. Kincannon joked, though, that Habermann was already sitting in the back by the door.

Kincannon then recalled some of the accomplishments at Census in recent years. The American Community Survey (ACS) was launched, and is becoming established with a solid constituency of users. The short form only census is proceeding – now free from the long form, which was the source of most of its controversy. As Kincannon put it, the ACS gets letters of complaint, but no one is on talk shows complaining about it. Other accomplishments include the proposed new SIPP design, and the efforts underway to improve the demographic estimates.

Ed Spar asked Kincannon to address the ancestry and group quarters issues. The director noted that a three-question option (adding ancestry to the short form census while shortening the race and ethnicity questions) did not work out in testing, so the 2010 census will proceed with only Hispanic and race questions, in a format similar to that used in 2000. Kincannon noted with satisfaction the recent approval of this proposal by the census advisory committees. Turning to group quarters, Kincannon assured that it will be on the decennial, but that due to funding uncertainty, it remains to be seen if group quarters data collection will continue in the ACS. He described ACS group quarters data as “not dead yet.”

Kincannon concluded by assuring that the 2010 census is in good shape, and on track for a “real dress rehearsal.” An attendee noted the contribution of duplication to the low net undercount in 2000, and asked if enhanced de-duplication capabilities might result in a 2010 census with a higher net undercount. Kincannon described this as a plausible scenario, and something for future directors to be aware of. Asked about the drop dead date for funding for the hand held devices, Kincannon said “we need the anomaly” (exception to continuing resolution funding). Most critical is that funding allow for the use of hand held devices in the 2008 dress rehearsal. In response to a question on whether the Census Bureau can remain politically independent in the present partisan environment, Kincannon noted that the census is inherently political, and that the parties will exert pressure. But he observed that the Census Bureau is very large and difficult to penetrate, and expressed his belief that the Census Bureau’s core integrity is strong, and that data products are not likely to be influenced.

The State of Services Productivity Measurements.

Jack Triplett. The Brookings Institute.

Triplett explained that overall productivity growth in the U.S. was very slow from 1973 to 1995, but there was a great expansion following 1995 – an expansion that data did not detect for several years. Most credited the expansion to advances in information technology, but Triplett explained that that much of the expansion traced to a suddenly dynamic services sector, which was becoming more productive to the surprise of many. Even when information technology “went bust,” the increasingly dynamic services sector drove continuing growth in total productivity.

To illustrate how long the importance of the service sector has been underestimated, Triplett asked for guesses as to when non-goods producing jobs first became more than half of those in the U.S. economy. The surprising answer was 1940 (as reflected in the decennial census), and Triplett commented that it has taken that long to address services in the statistical system. Triplett noted that the effort started in the 1980s, and credited the system with coming a long way since that time.

Having acknowledged the great progress of recent years, Triplett then presented his wish list – literally a list of recommendations in the back of a paper he distributed. To save time, he highlighted just a few items that give the flavor of what he sees as the current needs. Triplett’s first recommendation calls for continuing and accelerating the Producer Price Indexes for services. He also calls for continuing and accelerating Census collection of inputs for services industries and of purchased services for all industries. Other recommendations include combining costs of disease in the National Health Expenditure Accounts, and changing the System of National Accounts concepts of finance and insurance.

Triplett concluded with the observation that sometimes improvements to data reveal problems in the data that previously had not been known. As an example, he cited how recent data improvements in the services area have put a spotlight on inconsistencies in industry classifications.

Statistical Uses of Social Security Administrative Data.

Dawn Haines. Social Security Administration, Office of Research, Evaluation and Statistics.

Haines explained that the Social Security Administration (SSA) is not a statistical agency, but a program agency with a lot of data – including data on eligibility, program participation, and earnings history. Their statistical products include hardcopy publications, public use microdata files, and research and analytic studies.

Research publications include the Social Security Bulletin, a peer- reviewed research journal, a working paper series, and policy briefs. Statistical publications include the Annual Statistical Supplement and chart books providing fast facts and features for audiences such as Congress.

Microdata files include the 2001 OASDI Public Use File (PUF), that provides data on program eligibility, program participation and other Social Security related variables for about 460,000 records. There is also a 2004 Benefits and Earnings PUF that provides the first release of longitudinal earnings data for about 470,000 records. Haines also described the National Survey for Children and Families, that provides data on 8,700 children and young adults with special healthcare needs.

Haines cited high quality as one of the strengths of SSA data. They have information on all beneficiaries, recipients and covered workers. Limitations include the restricted scope of the data, and the lack of research variables of interest (e.g., demographic and socio-economic). Turning to survey data, Haines described strengths including a broad range of topics and links to married couples and race data. Limitations include small samples, lack of representation of some population components, and both sampling and nonsampling error.

An attractive feature of SSA data is the linking of administrative and survey data – providing enhanced records based on matched Social Security Numbers (where available). Uses for such linked data include micro-simulation models to research the impact of proposed program changes. Examples include Modeling of Income in the Near Term (MINT), based on data from several SIPP panels. There is also a Financial Eligibility Model that simulates the effect of proposed changes to SSI eligibility criteria. Another use of linked data is to study differences between what respondents report on surveys and what administrative data show they actually paid and/or received – the results pointing to survey improvements.

Haines concluded by identifying some of the challenges related to SSA data. These include the interpretation of results based on different universes, the use of data collected for different purposes, and understanding differences between survey respondents and non-respondents (as well as program participants vs. non-participants and linked vs. non-linked cases). Other challenges include data definitions, protecting confidentiality, public perception, and he fact that Social Security numbers are no longer collected – at least by Census surveys.

The Participant Statistical Areas Program for the 2010 Census.

Vince Osier. U.S. Census Bureau, Geography Division.

Mike Ratcliffe, who was originally on the program, was in attendance, and contributed to much of the discussion during Osier’s presentation. Osier reviewed the Census Bureau’s plans for the geographic areas to be used in the 2010 census and ACS, including the participant areas established in collaboration with local governments. For such areas, the Census Bureau establishes criteria which the local areas use in defining geographic areas. The Census Bureau has been discussing these plans with local governments, transportation planners, GIS user groups, and other stakeholders, and will publish proposals in the Federal Register.

In establishing the criteria for defining participant statistical areas, the Census Bureau considers factors such as consistency with previous censuses, data reliability, and disclosure avoidance. The mention of disclosure avoidance and statistical reliability hit a nerve with some attendees, concerned with announced plans for the ACS to collapse, and sometimes suppress, data for small areas that do not pass statistical reliability checks. However, the discussion confirmed that this practice still applies only to the one-year and three-year ACS data. Ken Bryson, with the Census Bureau’s ACS staff, confirmed that the ACS would still provide five-year data for all block groups. In other words, the five-year ACS data are still exempt from the reliability checks.

Few changes are proposed for census tracts. One minor change would lower the minimum population threshold from 1,500 to 1,200. Tracts could also meet the minimum threshold by having at least 480 housing units, and participants are encouraged to maintain continuity by splitting or combining previous census tracts to meet thresholds. Special land use tracts will be allowed (for airports, large city parks, and other large unpopulated areas), and wherever possible, the criteria call for tracts to conform to American Indian reservations.

For block groups, no comparability between censuses would be required, but there is a proposal to raise the minimum population from 600 to 1,200. With 54 percent of Census 2000 block groups failing to meet this criterion, this change would dramatically reduce the number of block groups defined nationwide. This proposed change was recommended by ACS statisticians who have long opposed the release of ACS data for block groups.

A flood of discussion followed. It was noted that data users have battled to preserve ACS block group data, and the Census Bureau publicly agrees that block group data are required for the ACS objective of long form replacement. That’s why five-year ACS data are exempt from statistical reliability checks. As one attendee put it, the Census Bureau could technically meet the promise of ACS block group data by simply renaming census tracts “block groups,” but that would not constitute long form replacement. It is understood that statistical reliability is low for individual block groups, but that they are valuable building blocks to user defined aggregations for which the data are reliable. Reducing the granularity of these building blocks impairs rather than improves their application.

Osier and Ratcliffe were very patient and constructively responsive during the flurry of discussion, and stressed that the thresholds are only a proposal at this point.

Osier was finally allowed (by Spar’s decree) to finish his presentation. He noted that not many people use Census County Divisions (subcounty areas in states with no legally defined Minor Civil Divisions), so they are considering their elimination in 2010. The Census Bureau also is rethinking Census Designated Places (CDPs), which are statistical equivalents for places defined for well known settlements that are not incorporated. Attendees expressed concern that CDPs are still needed in some states, but the Bureau proposes only to refine the criteria – specifying that CDPs cannot have zero population and housing, and cannot be coextensive with Minor Civil Divisions.

Osier concluded with a reminder that the proposed criteria will be published in the Federal Register in early 2007, and that there will be a 90 day review period. Final criteria will be published in late 2007. Participants will be identified in late 2007 through early 2008, materials will be distributed during summer 2008, and the participants will have 120 days to review and submit boundaries.

Concerns of COPAFS Constituents

No concerns were raised, and the meeting was adjourned.