Minutes of the March 6, 2009 COPAFS Meeting

Ed Spar. Executive Director’s Report

Spar reported the morning’s news that the omnibus appropriations bill did not pass. The government is still under a continuing resolution, but another vote was expected the following week. (Subsequently passed on March 10).

Referring to the budget sheet, Spar said we still don’t know what the final numbers would be (apart from additional funding already approved for 2010 census). Nothing has changed on NCHS. The NHIS sample is still cut in half, and they will collect only the bare minimum vital statistics data. We would only get “enhanced” vital statistics data (for example, additional information from birth certificates) if NCHS gets additional funding. And there appears to be nothing for NCHS in the stimulus package.

Spar described a Federal Register notice on NAICS – with comments due April 30. He also mentioned that final decisions have been made on standard occupation classifications (details available online). The Census Bureau is again considering ways to combine the race and ethnicity questions. The focus is not on 2010 (that’s set), or to establish final questions, but to develop guidelines for future deliberations on this topic. Karen Humes at Census is the recommended contact for those interested in following up.

Looking to the June COPAFS meeting, Spar mentioned that we will have a Census Bureau presentations on HUBERT follow up research, and the designation of urbanized areas. The remaining 2009 meetings are June 5, September 11, and December 4.

Spar then introduced Gary Chappell, from the Census Bureau, who gave us a quick update on the 2010 census partnership program. The objective is to sign national organizations as 2010 census partners. The partnerships are largely symbolic, but Chappell noted that concrete contributions can be made (such as providing space for training, promoting the census, providing speaking opportunities, etc.). The Census Bureau has been conducting email outreach to potential partners, and will host a March 30 partnership kickoff meeting in Washington. COPAFS has already signed on as a 2010 census partner.

Estimating and Mapping International Populations at Risk from Disasters
Joshua Comenetz. U.S. Census Bureau
James Fitzsimmons. U.S. Census Bureau.

Jim Fitzsimmons described the Census Bureau’s objective of identifying populations at risk from disaster around the world. And to make the data more valuable in relief efforts, they need to estimate sub-national population distributions. Such estimates can be a challenge because, many countries do not have recent census data. Fitzsimmons noted that the Census Bureau has been maintaining an international database, and has now conducted a study on the topic of international populations at risk of disasters. The project’s use of satellite imagery has received the most attention, but a key to the approach has been a willingness to “take anything we can get.”

Josh Comenetz described a National Research Council study a few years ago that called for improved sub-national population estimates, and a wider availability of these data. The Census Bureau has responded with the goal of maintaining a frequently updated database of such estimates, with an emphasis on countries most likely to require outside assistance with humanitarian disasters. The ideal of a worldwide population register is unattainable, so the minimum objective is a set of estimates of population by age/sex, with data on housing, literacy and language. It is hoped that such data could save lives if they are readily available before, during and after incidents.

An objective is to maintain data for three levels of administrative areas (1. state/province, 2. district/county, and 3. municipality/township). Many countries have only levels 1 and 2, and a few have only level 1. Data for the elusive 3rd level are those that would be most useful for disaster relief. Annual estimates for levels 1, 2 and 3 are based on census data and occasionally official estimates, and are controlled to the national population estimates maintained in the Census Bureau’s International Dababase.

Maps, which are also important for disaster relief, pose a separate challenge. Comenetz explained that, unlike the US, many countries produce maps and demographic data in different agencies, so data and maps can be out of synch. One also finds inconsistencies between maps, and maps are available only on paper must be digitized, rescaled, and checked for accuracy. Permission to use is an issue in some countries where maps are not approved for use outside the government, and changes in the boundaries of administrative units are a challenge.

Comenetz then showed some example data, and explained that countries provide data in different formats and media. Then there are the varying definitions that make cross-national comparisons difficult. Among the data sources used are national censuses and official estimates, Census Bureau products such as the President’s Emergency Plan for AIDS relief, the International Database, United Nations data, and surveys by governments and NGOs.

Satellite images also have been used as a basis to distribute population totals to sub-national areas – based on the distribution of hard or impervious areas produced by humans (such as pavement). Topography also is helpful, as people rarely live on steep slopes. Comenetz noted that the image work is only partially automated, and often requires manual work.

Asked how well population distribution corresponds to impervious surfaces, Comenetz showed the strong correspondence in Haiti (confirmed because Haiti has a good census), but he acknowledged that the correspondence likely varies by country. In response to a question about the tsunami in the Indian Ocean, Comenetz said that is exactly the kind of event they seek to address, but he noted that only some of the impacted countries have data at that 3rd level.

The program’s goals for the future include building a shape file archive and demographic database, the use of remote sensing methods to update population data, and making information available online, so they can be used in disaster relief.

Income Data for 2002: A Comparison of Eight Surveys
John Czajka. Mathematica Policy Research.

John Czajka presented on Mathematica’s study of the consistency of income data across eight major surveys – including:

CPS Current Population Survey
SIPP Survey of Income and Program Participation
ACS American Community Survey
NHIS National Health Interview Survey
MEPS Medical Expenditure Panel Survey
MCBS Medicare Current Beneficiary Survey
HRS Health and Retirement Study
PSID Panel Study of Income Dynamics.

Data were examined for 2002 – the most recent year for which all surveys had income data. And in order to make the data more comparable, they standardized income tabulations based on standardized income measures, family definitions and population groups. There is no gold standard for income data, but comparisons were made against the CPS because it is an official source of income data. All estimates were at the national level.

Czajka presented a number of tables comparing the surveys on measures such as:

Aggregate income
Income distributions
Average income per capita by quintile of family income
Persons with earnings and per capita amount
Earned income as a percent of CPS by quintile of family income
Unearned income as a percent of CPS by quintile of family income

For many of the comparisons, the ACS income estimates were impressively close to the CPS results, and often closer than those of the other surveys. The ACS/CPS consistency was considered notable given the differences in ACS data collection, reference periods, and residence rules. For example aggregate income estimates were as follows.

CPS 6.47 trillion
ACS 6.35 trillion
MEPS 6.28 trillion
SIPP 5.77 trillion
NHIS 6.12 trillion

Czajka noted that the PSID was the highest of all surveys, at 6.72 trillion, despite a weighted population of 21 million fewer than the CPS.

Czajka reviewed additional tables comparing incomes for persons 65 and older and 51 and older, and described how they investigated the impact of family definitions on poverty rates (the inclusion of unmarried partners and their relatives reduces the number in poverty compared to the CPS family definition). The consistency of many survey results was impressive, and an interesting finding was that results based on only a few income questions were not that different from those based on many questions (interesting given expectations that multiple questions were important for accuracy). Also interesting was the finding that allocation rates were higher in months near tax filing deadlines – months when one might expect respondents might find it easier to report their income. ACS allocation rates were notably low, perhaps because of its status as a mandatory survey. The study also noted that a large percent of incomes are divisible by $5,000 – suggesting that many respondents are rounding when they report.

Czajka concluded by noting that the study was not intended to provide recommendations. However, it provides groundbreaking work that they hope will serve as the basis for recommendations for survey improvements. The full study is available on the Mathematica Policy Research website, and Czajka’s presentation will be posted on the COPAFS website.

Reviewing the Standards for Defining Metropolitan and Micropolitan Statistical Areas: A Look at Committee Recommendations to OMB.
James Fitzsimmons. U.S. Census Bureau.
Paul Mackun. U.S. Census Bureau.
Marc Perry. U.S. Census Bureau.

Ed Spar introduced this session by noting that no major meetings are being held on the metropolitan/micropolitan topic this time around, so this COPAFS presentation is an important opportunity for those interested the plans in advance of the 2010 census. With that opportunity in mind, the entire afternoon was devoted to this session.

Jim Fitzsimmons recalled the exhaustive effort that went into revamping the standards in advance of the 2000 census, and noted that this time, the work of the Metropolitan and Micropolitan Area Standards Review Committee has been more focused. Comments are sought on just a few areas, and we are now midway through the comment period for the Federal Register notice on this topic.

Pau Mackun reviewed the current standards and concepts, including metropolitan areas, micropolitan areas, metropolitan divisions, combined statistical areas, principle cities, etc.

Issues for review include the qualifications and titling of combined statistical areas, the scope of metropolitan and micropolitan updates, and replacing the term “definition” with “delineation.”

CSA Qualifications and Titling
The current standards link metropolitan and/or microplitan areas into combined statistical areas if the have an employment interchange of 25 percent or more (the measure is the sum of the percent commuting from the smaller to the larger area plus the percent of employment in the smaller area accounted for by workers residing in the larger area). The combinations are automatic for interchanges of 25 percent or more, but areas with an interchange from 15 percent to 25 percent can be combined if that is the preference of local opinion. The current proposal is to eliminate the use of local opinion, since it disrupts comparability and introduces ambiguity into the process. Instead, areas would be automatically combined if they meet the 15 percent exchange threshold. Also, local opinion would no longer be considered in the titling of areas.

Scope of Updates
For some purposes, yearly (or more frequent) updates to metropolitan and micropolitan areas would be helpful, but for many purposes, stability is appreciated. The question arises now because the American Community Survey will be providing updated commuting data every year – making the annual redefinition, or re-delineation, of areas a possibility that needs to be considered. Balancing the interests in stability and frequent updates, the committee proposes to conduct yearly updates only on the identification of new metropolitan and micropolitan statistical areas. The redefinition, or re-delineation, of county components would be limited to once every five years (one update between each census).

Replacing “Definition” with “Delineation”
The committee is concerned that the term “definition” is not intuitive for some users, so they propose to substitute the term “delineation.” New data and standards would be used to delineate, rather than define, metropolitan and micropolitan statistical areas.

OMB is seeking comments on these proposals. Comments are due in writing on or before April 13, 2009.

In the discussion that followed, some attendees expressed concern that the 15 percent threshold for combination seemed low – significantly lowering the bar for automatic combination from 25 to 15 percent. The Census Bureau presenters were not sure what percent of “15 to 25” percent areas had chosen (by local opinion) to combine under the current standards.

Concern was expressed that the standards still do not provide guidelines on how to define the widely used “suburban” concept. The convention of defining suburban as metro minus central city was noted, but the concept remains elusive, and there are no official guidelines.

Attendees also sought clarification on the plan to update area components every five years. Again, new areas could be established every year based on annual population estimates, but counties would be added or subtracted from areas only once every five years – to provide greater stability. For example, an initial set of areas would be “delineated” in 2013, and these metro and micro components would be stable until the next update in 2018 (based on ACS journey to work data).

No strong opinions were expressed about the switch from definition to delineation, but in private conversations, there was some grumbling about “delineation” and head scratching about the importance of the switch.

Concerns of COPAFS Constituents

No concerns were raised, and the meeting was adjourned.