Minutes of the September 15, 2006 COPAFS Meeting

Stepping in for COPAFS Chair Sarah Zapolsky, who was present but ill, COPAFS Executive Director Ed Spar started the meeting, and we went directly to his report.

Ed Spar. Executive Director’s Report.

Spar commented that the budget numbers could be better, and are among the worst he has seen in years. The Census Bureau has been cut by both the House and Senate – in contrast to recent years when cuts were from the Senate only. The dual cuts will make a favorable resolution difficult to achieve. The Census Bureau has testified that these cuts threaten the collection of group quarters data in the ACS, and the use of handheld devices in the 2010 census. In contrast, the BEA and BLS budgets look fine, while NCES is flat. Spar explained that the NCHS budget looks flat, but is not because CDC is pulling out $5 million for administrative purposes – which will have a negative impact on the National Health Interview Survey. Spar described the Bureau of Transportation Statistics budget as “bizarre,” with the House and Senate having zeroed it out, and put it all under the Federal Highway Administration. It is not clear what is going on, or what the implications are.

Next, Spar described the September 6 House subcommittee hearing on the Census Bureau’s population estimates program. The hearing was prompted by the large revision to the Census Bureau’s population estimate for the District of Columbia, and became “a bit bloody,” with the District’s representative, Eleanor Holmes Norton grilling Census Bureau director Louis Kincannon on the under-estimation, and more broadly, on the fact that the census does not count prisoners at pre-incarceration addresses. In the hearing’s calmer moments, there was discussion of the potential for improving the estimates program through the greater use of MAF counts and the housing unit method – a suggestion heard in July at the COPAFS co-sponsored conference on Census Bureau estimates.

Spar reported that plans are in the works for a conference on survey incentives, probably to be held in the spring. Incentives are widespread in private sector survey work, but are still controversial for federal surveys given the potential for bias. Spar will communicate details as they develop.

The final COPAFS meeting of the year will be Friday December 8.

An Update from the Bureau of Justice Statistics.

Jeffrey Sedgwick. Bureau of Justice Statistics

Sedgwick, a political science professor before becoming BJS Director, described his agency’s context as driven by administration interests in issues such as counter-terrorism (things like cyber security and identity theft), human trafficking (poorly defined but including things like migrant labor and commercial sex trade), family violence, and the advancement of justice through technology (such as DNA testing). Priorities coming through the Attorney General’s office include Internet crime against children and prisoner re-entry. Sedgwick said BJS likes recidivism studies, but they are very expensive, as they require following a release cohort for three to five years.

Turning to challenges, and promising that he was not begging for funding, Sedgwick noted that BJS has a lot to do on a limited budget. He described BJS as victimized by the popularity of its reports, which has led to an overloading of the dissemination unit—with about two years of publications currently in the queue. As Sedgwick described it, BJS consumers are different than they used to be. Accustomed to the reporting of sports scores and stock quotes, they want immediate data, and no longer tolerate a 2006 release with data as old as 2002. So there is pressure to release data more quickly, or to at least become better at explaining why 2002 data are still relevant in 2006.

BJS faces a financial challenge in the form of a proposed $20 million cut following years of flat funding, and Sedgwick commented that they have not been good at insisting that they cannot do more work without more funding. Instead, BJS has responded by cutting the size of the National Crime Victimization Survey, which is about 60 percent of their budget, but is “dying a death of a thousand cuts.” The sample size has been cut to the point where a year-to-year change in a crime rates cannot be discerned until it reaches 9.8 percent.

BJS also is challenged by the fact that criminals are innovative. They now have to collect data on crimes such as Internet predation and identity theft – adding further to the workload and the need for funding. Staffing is another challenge as BJS has only 58 people (not all statisticians) to maintain 48 statistical series. Sedgwick remarked that high level Justice Department officials are now sophisticated data users. This means that BJS is no longer the orphan it once was, but it increases demand and expectations – putting further pressure on resources.

Initiatives in response to these challenges include a push to adopt new technology for both data collection and dissemination. BJS is looking to move away from document providing mode to electronic information dissemination, and for that purpose, they have hired an information architect. BJS also has contracted with the Committee on National Statistics for a comprehensive evaluation – a fresh look from the outside to help them determine if they are doing the right things with the resources they have.

During the question/answer period, Spar suggested a BJS user seminar, and Sedgwick said he would love to have the feedback it would provide.

An Update from Capitol Hill.

John Cuaderes. House Subcommittee on Federalism and the Census
Mark Stephenson. House Subcomittee on Federalism and the Census
TerriAnn Lowenthal. Consultant

Cuaderes, from the majority, said he had little to say about the budget, then talked only briefly about the hearing on Census estimates. He did say that some interesting things are in store in the next few weeks, but could not say now what they are. Stephenson, from the minority, summarized the Census budget situation, noting the Census Bureau’s claim that the elimination of handheld devices from the 2010 census would cost an additional $1 billion over a period of years. Asked if they could substantiate this claim, GAO said they could not, as the Census Bureau had not provided sufficient information. The subcommittee has asked GAO and the Census Bureau to get together for a definite answer.

Lowenthal commented that the hearing on Census estimates was timely, if not overdue, as many localities now challenge their Census Bureau estimates. She then summarized recent stakeholder activities, including the letter writing and other activities of the Census Project (the stakeholder group she works with), and the letters of support sent by organizations including the International Council of Shopping Centers, and the Population Association of America.

Lowenthal reported that a Committee on National Statistics panel on census residence rules had just released its report, and has encouraged the Census Bureau to collect data on alternative (pre-incarceration) homes of prisoners. She wondered if this recommendation might find its way into the eventual appropriations bill.

Lowenthal described the proposed Census budget cuts as troubling in that they target the census itself. Director Kincannon has expressed that if the cuts hold, the Bureau would prefer the discretion to distribute the them to activities considered more expendable – such as the inclusion of group quarters in the ACS, the ACS methods panel, and the Census Address Updating System. Lowenthal observed that Congress is actually supportive of the census, and has criticized the Census Bureau (and the Department of Commerce) for not communicating earlier why its funding is so critical.

Cuaderes added that GAO is now agreeing that the cost of eliminating handheld devices from the 2010 census would be about $800 million (rounds up to $1 billion?). He then followed through on Lowenthal’s remarks, commenting that the annual struggle with the Census budget traces to the Census Bureau’s lack of presence on the Hill – with advocacy repeatedly falling to stakeholders. When an attendee cited the Census Bureau’s claim that they cannot lobby Congress, a frustrated Cuaderes snapped that the Census Bureau is the only agency that uses that excuse. Cuaderes argued that other agencies can, and do, “educate” Congress on the importance of their activities, while the Census Bureau gets its pocket picked every year because they don’t go to the Hill to say “don’t pick my pocket.” Cuaderes said the Bureau needs to visit those members who voted against the Census budget last year (as other agencies do), and contended that they have not visited a single one. He agreed that Commerce is not without blame, but stressed that if the Census Bureau does not adapt to this reality, we will be going through the same budget fight again next year.

Linking Survey and Administrative Record Data: Developing the Medicaid Undercount Study.

Sally Obenski. U.S. Census Bureau
Ron Prevost. U.S. Census Bureau

Obenski traced the Census Bureau’s administrative records program to Title 13, which mandates the use of administrative data to minimize respondent burden, and also mandates the safeguarding of administrative records. The program began in response to the 1990 census undercount, with the initial objective of seeing to what extent administrative data could supplant the census. It became clear that a nationwide compilation of data was the only feasible way to go, but that this would be enormously complicated.

Compilations of this type have been tested under the StARS (statistical uses of administrative records) program, which compiles short form information on individual records, and where coverage issues are similar to those in the 2000 census. Another activity, the NUMIDENT files, involves hundreds of millions of Social Security records, which can be matched to census records to provide data on race and other characteristics. Obenski also described the Person Validation System (PVS), which enables the Census Bureau to identify individual records based on unique personal identifiers based on name, address, and date of birth – and which replaces the use of Social Security Numbers.

Examples of current research include the use of administrative records to assign age, sex, race, and ethnicity to matched records, the identification of households with coverage problems, and the use of commercial and other lists to improve the group quarters frame. It is hoped that administrative data can eventually reduce ACS small area variance by adjusting the survey weights, and providing characteristics for non-respondents. It is also hoped that administrative data, such as USPS change of address data, can improve the Census Bureau’s reaction to disasters (such as Hurricane Katrina) and other near-real time requirements.

Operational constraints include file acquisition (negotiations with states and other sources can be complicated), differences in definitions and quality from file to file, and file lag time – as administrative sources lag from one to four years. Obenski observed that addressing these constraints may require OMB or congressional assistance, and new challenges continue to arise. But she was upbeat about the long term possibilities, as new resources continue to arise, and are expanding the potential use of administrative data.

Ron Prevost then described the Medicaid Undercount Study, which illustrates an application of administrative data. The problem is that survey estimates of Medicaid enrollment are well below administrative figures. The Current Population Survey indicates about 25 million compared to almost 39 million administrative records. The study seeks to understand the causes of the discrepancy. Possibilities include universe differences, measurement error, and data processing differences, as well as survey coverage error and non-response bias. Other issues include persons receiving Medicaid in two or more states, and respondent knowledge – as some enrollees may not know they have coverage, or might confuse Medicaid with Medicare. Bias is another concern, as non-response may be higher among poor populations.

Prevost described the test in some detail, but the basic approach was to match CPS records to Medicaid records to examine the correspondence between reported and actual enrollment. The test also included a national Medicaid to NHIS person match. The matched records enabled the development of tables detailing explanatory factors and characteristics useful in examining universe differences and measurement error.

Preliminary results suggest that measurement error plays the most significant role in the survey undercount, as some Medicaid enrollees answer that they have other types of coverage, and some report that they are uninsured. As Prevost put it, basically people don’t know where their health insurance is coming from.

The broader conclusion is that the integration of administrative and survey data provides a robust and accurate picture that builds on the strengths of both sources while minimizing their weaknesses. Prevost noted that policy challenges persist (related to privacy concerns and interagency coordination), but that work of this type can be a cost-efficient way of addressing increased data demand in a period of tight budgets.

HHS Gateway to Data and Statistics.

Daniel Melnick. Consultant

Melnick noted that healthcare data are found in many places, and described Gateway to Data and Statistics www.hhs-stat.net as a web based tool to help users find, organize and use healthcare data. The site was prepared by Moshman Associates, and Melnick is the key consultant on the project.

The site does not provide unique or original data, but rather makes it easier to find and use existing sources. Currently, there are about 8,500 sources on the site and another 2,000 are to be added shortly. These include data from federal, state, and local sites as well as federal grantees and contractors. At this time, metadata are provided only for agencies in the Department of Health and Human Services.

As Melnick noted, the site is only one of many ways to access healthcare data. One could use Google, but this tool is more targeted, and easier to use for the material covered. For example, Gateway is catalog based, rather than searching every word in related documents. It is a tool for finding sources—not for finding answers, as it takes users to places where relevant data can be found. One can search Gateway by topics, studies, or types of resources, and can specify levels of geography—or even specific areas, such as “data for New York.” One can use Boolean logic (described as the most powerful way to use the tool), or there is a “guided search” function that holds the user’s hand during the search process.

Melnick went through a series of screen shots illustrating how Gateway works, and what it can provide. He noted that the site is constantly being updated and revised, and said that comments and suggestions are invited—either through a form on the website or regular e-mail. Melnick and his partners are very interested in getting the word out about this site, and expressed hope that COPAFS members would visit the site and find it useful.

Concerns of COPAFS Constituents

No concerns were raised, and the meeting was adjourned.

The Council of Professional Associations on Federal Statistics

20 F Street, NW, Suite 700, Washington, DC 20001 | www.copafs.org

©2010 The Council of Professional Associations on Federal Statistics. All rights reserved.