Summary of the June 25, 2004 COPAFS Meeting

In the absence of COPAFS Chair Don Muff, Ed Spar started the meeting by expressing his appreciation to all who rearranged schedules after the meeting was switched from June 11 due to the government closing for the Reagan funeral. Ed also noted that it has been a tough few weeks, as we mourn the loss of three colleagues - Pat Doyle, Chuck Waite and Charlie Jones.

Shifting to Executive Director report mode, Ed described a reorganization at CDC which puts an additional reporting layer between NCHS and the CDC director. The new structure combines NCHS with the National Center for Public Health Informatics and the National Center for Health Marketing, as part of a "Coordinating Center for health Information and Services." There is concern that the new arrangement could be detrimental to NCHS. Ed then drew our attention to a Bill Frey article on micropolitan statistical areas, and described the Federal Register notice soliciting comments on proposed ACS data products.

Turning to the federal budgets, Ed noted that despite understandable skepticism, there is talk of budgets passing on time this year, and some mark ups have been made. The House mark up of the Census budget hits the Bureau with a $55 million cut - with $19 million to come from the ACS. There is talk that part of the cut may be made up for by eliminating group quarters from the ACS-a move that seems to have some support both in and out of the Census Bureau. The other $36 million in Census cuts are likely to be distributed among economic statistics, MAF/TIGER, and 2010 decennial. The Economic Statistics Administration (which includes BEA) has also been assigned a budget cut, and Ed described it as almost a double cut, as ESA has been asked to spend $2 million of its budget on studies of outsourcing.

Ed noted that Connie Citro is the new director at the Committee on National Statistics. He then wrapped up with a pitch for the December 15-16 conference on data quality, and reminded us that the remaining COPAFS meeting dates are September 10 and December 10.

The 2002 Census of Agriculture - New Data and New Information

Ronald Bosecker. National Agricultural Statistics Service

Ron started by reminding us that a farm is defined as a place with sales of $1,000 or more of agricultural products, or which would normally have such sales during the census year. The Census of Agriculture provides data on the type and size of farms, the demographic characteristics of farm operators, farm income and expenditures, and commodity inventory and production. The census provides approximately 15 million cells of information.

The 2002 census counted 2.13 million farms in the U.S. - down about four percent from the 2.22 million counted in the 1997 census. Total land in farms dropped less rapidly (1.7 percent) to 938.3 million acres. The pattern differs by size of farm, as farms with 10-49 acres or 2000+ acres actually increased in number since 1997. The 2002 census also indicates that the value of farm sales is concentrating at the high end, with 7 percent of farms now accounting for 75 percent of sales. Direct sales of farm products to consumers (e.g., farmers markets) is an area of growing interest, with sales increasing from $592 million in 1997 to $812 million in 2002.

Farm operators are an important unit of analysis for the Census of Agriculture, and the 2002 census counted 3.1 million operators-0.8 million of them women. The 2002 census is the first to allow for more than one operator per farm (so long as they have an active role in operating the farm). Total persons in farm households numbered 6.6 million, and there are about 2.2 million farm households. Ron noted that a single farm can have multiple households, while some farms (such as major corporate farms) have no households.

Based on the 2002 census, the average farm had 441 acres (up from 431 in 1997), sales of $97,320 (up from $93,270 in 1997), and an operator who was 55.3 years of age (up from 54.0 in 1997).

Response to the Census of Agriculture is mandatory, but nonresponse is an issue - even for large farms. Nonresponse follow up is conducted by telephone, and in some cases with personal visits. Coverage is measured against a farm list, which Ron said they maintain using "every source we can find" including data from associations, government payments, and even subscriptions to farm magazines.

Ron described the Census of Agriculture as having a wide range of users, including schools, university researchers, and organizations that serve farms. Data products are now mostly on the website and CDs, and these are primarily in PDF rather than database form. Future products include an online query friendly database, state and county profiles, congressional district rankings, Native American statistics, and data from a farm and ranch irrigation survey (of interest in western drought areas). Planning for the 2007 Census of Agriculture is underway, and Ron noted that they are taking suggestions for new items. He asked, however, that users please also suggest items that can be taken off the census.

A Review of the Local Employment Dynamics Program

Jeremy Wu. U.S. Census Bureau

Jeremy explained that the Local Employment Dynamics Program (LED) is a component of the Longitudinal Employer-Household Dynamics Program (LEHD). The objective of the LEHD is to merge varied data sources to provide more complete information about the US Society and economy - or as Jeremy described it, turning multiple data snapshots into movies. The program is still being developed, and Jeremy's presentation focused on the LED, which seeks to combine data from demographic surveys, economic censuses and surveys, linking them to business registers and person level records. Confidentiality protection is definitely a major issue. In response to a question, Jeremy explained that "noise introduction" is the primary method used to protect confidentiality.

The LED initiative started in 1998 with funding from the Census Bureau and several foundations. The driving focus behind it was the need for local data among employers and those involved in workforce information boards, economic development agencies and transportation. The focus is on data for small geographic areas, demographic sub-groups, and for data that are dynamic-allowing comparisons over time.

The program involves a voluntary federal-state partnership with state labor market information agencies. States provide data that the Census Bureau merges with federal and other sources to provide Quarterly Workforce Indicators (QWIs) that are then returned to the states. QWIs are local-providing data for state, county and sub-county areas-and the focus is on employment by characteristics including age/sex, industry and earnings. And again, the data are dynamic-tracking job gains, losses and flows quarterly as far back as 1990.

LED currently has 31 participating states-with 24 shipping data, and 19 having produced QWIs. An additional 10 states are at various stages of joining the program. The LEHD website-introduced in February 2004-is described as a modest, but effective site that is a good source of documentation, and which will become the primary means to deliver QWIs. Jeremy showed examples of the site's pages, and described features for accessing data. The future LEHD website will be demand driven by customers and stakeholders, and will provide better applications, better data dissemination, and better online analytical capabilities.

Additional LEHD products and plans include the building of relations and applications, special topic reports, mapping applications, and "virtual" research data centers, where analysts can go beyond the standard tabulations to work with the basic data.

Jeremy wrapped up by showing example tables, and described the emphasis on drilling down. For example, one might start with a table of the top 10 industries in New Mexico, drill down to the top 10 industries in New Mexico for females age 35-44, and then drill down to a table of earnings in the 10 industries in New Mexico for females 35-44.

For further information, Jeremy directed us to the LEHD website at http//

Proposed Data Tables for the American Community Survey

Douglas Hillmer. U.S. Census Bureau.

Doug described the Federal Register notice that currently solicits comments on proposed ACS data products, and directed us to the Census Bureau website for more information. He described three dimensions of the ACS-content, time and geography-and their relevance to data products. Because the ACS questionnaire is similar to the census long form, one would expect data products to be similar to those from the long form-as has been the case with ACS test data. Geographic levels also would be similar, with both the long form and ACS providing data down to the block group level. But the time dimension is something new, and is not independent of the other two dimensions. For example, the more ACS sample is accumulated over time, the more geographies qualify for estimates (as the minimum population is reduced), and the more content (category detail and crosstabulation) can be provided for a geographic area.

Doug then described the ACS product release schedule, noting that the first data for three-year averages (based on 2005-2007) for areas of 20,000+ population are scheduled for release in 2008, and that the first five-year average data for block groups (based on 2005-2009) are due for release in 2010. In 2011, we can expect three-year data based on 2008-2010, and five-year data based on 2006-2010.

Given the mixed signals that we have had from the Census Bureau in the past, it was noteworthy that Doug's slides made explicit reference to the provision of ACS data for block groups.

A slide describing the geographic levels for which ACS data will be released in 2006 lists "Decennial 5% PUMAs." PUMAs are added for the benefit of areas that would otherwise have ACS data only for very large areas, such as state remainders or very large cities. This option prompted a question about a hypothetical PUMA consisting of two counties-one with a population of 80,000, and the other with 40,000. The question concerned whether we would get data for such a PUMA, since data from the county of 80,000 could be subtracted to determine data for the county of 40,000 (which would not qualify for such data). Doug did not have a definitive answer, but it was noted that two issues are involved with respect to the county of 40,000-the first being disclosure, and the second being data quality.

Doug described three criteria the Census Bureau will use in determining what to include in ACS data products. First is whether the data are used to support a current federal program mandated by law, second is whether the data are widely used as a key measure (such as median earnings), and third is whether the data can be processed and released in a timely manner.

Next, we looked at specific changes proposed for ACS data products. Doug stressed that changes need to be determined before full implementation, and that the Census Bureau is gathering input from workgroups and data users. The proposed products would include three levels: 1) Profiles and Ranking Tables, 2) Subject Tables, and 3) Base Tables. The profiles are similar to the Quick Tables provided for Census 2000, and the Base Tables are the foundation product from which other products are derived. The subject Tables represent a new "middle layer," providing crosstabulations related to about 50 major topics. Change tables also are proposed, and would present data for multiple years, such as 2000, 2001, and 2002. Sub-Population Profiles would provide data for subgroups, such as Native Hawaiians or persons less than age 18-produced directly from the microdata, rather than the base tables.

According to Doug, the Base Tables are likely to be a super-set of SF3-going beyond its already extensive content. They are not yet sure about providing SF4 type tables, but Doug noted that the objective of long form replacement would require SF4-type data.

The ACS will also have a special tabulation program modeled after that for the decennial census. And as with the decennial, the Disclosure Review Board will have different rules for special tabulations than for regular products.

Looking to "what's next?" Doug noted that the Census Bureau will start working on the Federal Register feedback immediately. They hope to have an internal version of new data products by late 2004, and plan to use the new product format for the summer 2005 data release. This would give them time to make changes to any areas where users have determined that the new products have "missed the mark."

Alternatives for Improving the Measurement of Poverty

Daniel Weinberg. U.S. Census Bureau

Poverty measures involve the definition of income thresholds related to the cost of meeting basic needs. The current measure-adopted in 1965-varies thresholds by family size and number of children. The thresholds are updated each year based on change in the Consumer Price Index, and define poverty as roughly three times the cost of food for a basic diet. The poverty measure is now under OMB Statistical Policy Directive No. 14.

Dan explained that the poverty measure has long been controversial. Common criticisms are that it does not account for the increased standard of living since 1965 or regional differences in the cost of living, and that the CPI does not measure inflation correctly. Critics also complain that the value of food stamps and other income-in-kind should be added to income, and that taxes, medical and other expenses should be subtracted.

Dan reviewed a long list of studies on the poverty measure dating from 1969 through 1992, and described the work of the 1995 National Academy of Science Panel on the subject. Among the NAS panel's recommendations were that poverty thresholds be set relative to expenditures on food, clothing, and other needs; adjustments for geographic differences in the cost of living; and updates based on spending rather than CPI. In response, OMB established a technical working group to assess these recommendations. As Dan put it, "we were told to do some research, and we did."

Key findings from the research were that the deduction of taxes reduces the poverty rate, and that adding in-kind benefits also contributes some reduction. Experimental measures showed a poverty population that is more like total population in terms of socio-economic characteristics, than that from the current measure.

Areas where there is consensus on how to improve the poverty measure include the use of an expanded measure of income, and adjusting the equivalence scales (variation by family size and number of children) to eliminate existing anomalies. Dan chuckled as he noted that the areas of consensus fit on one slide, then proceeded with the five slides summarizing areas of disagreement. These include whether to adjust for geographic differences, how to deal with the cost of medical care, and how to update the thresholds over time. The handling of healthcare expenses is probably the most controversial. The NAS panel proposed to subtract medical out-of-pocket expenses (MOOP) from income, but there is concern that MOOP measurement problems can push even middle income families into poverty.

The Census Bureau is currently producing a total of 21 alternative measures of poverty, and Dan said that they are working to reduce this to a smaller and more manageable number. He also noted that the alternative measures move up and down with each other, and show little change in poverty from 2001 to 2002. Looking to the future, Dan noted that OMB has asked NAS to hold a follow up workshop to assess the state of research on poverty measures, and that seminars are planned on alternatives to income-based measures of well-being. OMB will have to decide on the next steps to be taken. Asked if there was a schedule for the completion for this work, Dan quipped that he has only been working on this for about two and a half decades, so it should be sometime before the end of the century.

Dan concluded by noting that there is a poverty measurement website, and suggested that we check it out at

Concerns of COPAFS Constituencies
  • Howard Leathers, AAEA, U of MD
  • John Knapp, AUBER
  • Ed Goldfield, CNSTAT, Census
  • Ken Hodges, PAA/Claritas
  • Sarah Zapolsky, AARP
  • Margaret Martin, COPAFS
  • Linda Duffy, IASSIST
  • Seth Grimes, Alta Plana
  • Chris Reynolds, American Demographics
  • Paul Bielski, Mitre
  • Denise Miller, Mitre
  • Mel Kollander, IRS, Temple U
  • Tim Tang, NEA
  • John Cromartie, AAG
  • Fred Cavanaugh, Sabre Systems, Inc.
  • Rick Ayers, ESRI
  • Susan Lapham, ESSI
  • Colleen Flannery, Census
  • James Wing, Beyond 20/20
  • Ralph Rector, The Heritage Foundation
  • Gooloo Wunderlich
  • John Wertman, AAG
  • Dhiren Ghosh, Synectics for Management Decisions
  • Karol Krotki, ASA/Mathematica
  • Stephen Tordella, Decision Demographics
  • Carolee Bush, AAPOR
  • Linda Jacobsen, APDU
  • Judie Mopsik, Abt Associates
  • Patricia Becker, APDU/SEMCC
  • Ron Bosecker, NASS
  • Jay Johnson, NASS
  • Jeremy Wu, Census
  • Bob Parker, GAO
  • Jennifer Marks, Census
  • David McMillen, House of Rep.
  • Nancy Torrieri, Census
  • Nancy Gordon, Census
  • Dan w\Weinberg, Census
  • Jennifer Williams, CSR
  • Doug Hilmer, Census

The Council of Professional Associations on Federal Statistics

20 F Street, NW, Suite 700, Washington, DC 20001 |

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