Minutes of the September 16, 2005 COPAFS Meeting

In the absence of COPAFS Chair Sarah Zapolsky, past Chair Don Muff started the meeting by announcing that this is Ed Spar’s 50th meeting as COPAFS executive director, and presenting him with a commemorative gavel. Ed then noted that today was Susan Cohen’s last meeting as COPAFS administrative assistant, and presented her with a commemorative gift. Ed also noted that in response to intense interest in the impact of Hurricane Katrina, the Census Bureau agreed to present on the topic. He expressed appreciation for the Census Bureau’s last minute efforts, but also to OMB for agreeing to postpone their presentation on Statistical Policy Directives 1 and 2. Given the large turnout for the Katrina presentation, Ed deferred his executive director’s report until later in the day.

Some Effects of Hurricane Katrina on Federal Statistics

Alan Tupek, Tom Mesenbourg, Christa Jones. U.S. Census Bureau.

Alan Tupek reported that the Census Bureau is already thinking about the impact of Hurricane Katrina on its data products. They have contingency plans to guide them in situations like this, but this event will require major adjustments. For example, the Census Bureau cannot even get to some areas for September CPS data collection, and for October, they are considering the addition of questions to track the movement of hurricane evacuees. The Bureau expects to find about 250 CPS respondents who are Katrina evacuees, and to get some idea of where they have relocated. The ACS question on residence a year ago is expected to provide additional information on the movement of evacuees, and the October ACS questionnaires are being sent out with a letter instructing respondents to be sure to include information on evacuees who are living in their household.

In response to a question on how long it might take for the Census Bureau’s Master Address File to reflect the hurricane’s impact, Jay Waite said it depends on how quickly the Postal Service gets the changes in its Delivery Sequence File, which is the foundation of the Census MAF. Waite expects that the process may take a year or so, and that during that time, the ACS will have to use the MAF as is – and expect high nonresponse rates in the impact areas. Waite stressed just how uncertain the situation is at this time—noting that no one knows how many units in New Orleans will be demolished.

Tom Mesenbourg then addressed some impacts on the Census Bureau’s economic data. Again, initial work involves the description of what was in the impacted areas – in this case based on the 2002 Economic Census. He described trade statistics and construction statistics as two areas facing the greatest impact. The amount of activity at the Port of New Orleans alone ensures a sizable impact on trade statistics, and the Census Bureau is proposing the release of special data to reflect the trade flow impact on ports in the region.

The construction statistics program faces major challenges. For example, the new survey of construction sample does not cover New Orleans, and the survey does not cover demolition and clean up activity. The survey also does not capture federally owned manufactured homes, and is not designed to provide data for local areas. And in situations like this, local officials loosen up the permit issuing process, so administrative may provide an incomplete picture. The bottom line is that the construction program will have to do a number of things differently to capture post-hurricane activity.

Christa Jones spoke on longer term concerns, noting that the Census Bureau is looking at the possibility of expanding existing surveys to track evacuees, and measure reconstruction. They also are looking at possible administrative records contributions, and there is talk of special surveys—such as a a two-year longitudinal survey of the evacuee population. But Jones stressed that none of these possibilities has progressed to the formal proposal stage. Deputy census director Hermann Habermann was among the attendees, and commented that the topic is delicate – with a ferment of activity and ideas, but no funding yet.

In response to a question about tracking data from the Red Cross and FEMA, Jay Waite explained that these organizations moved people to irregular group quarters (such as the Houston Astrodome), but that those facilities are emptying rapidly. Tracking those who have moved out of the shelters, and the many who never went to a shelter, will be more difficult.

Turning to the ACS, Waite noted that it will take some time to get five-year data for the impacted areas, and described some key questions. Some wonder if the Census Bureau should beef up the sample in the impacted areas to get early data, but again, there is the question of funding. And there is concern about the population estimates used as weights, with some wondering if at least for a while, unweighted ACS data might provide a better picture of the population in the impacted areas. Fittingly, the session ended with more questions than answers.

Offshoring: Home-Country Effects on Investments Abroad by U.S. Multinational Companies.

Raymond Mataloni. Bureau of Economic Analysis

Mataloni cited one definition of “off-shoring” as domestic producers ceasing to produce a good or service in the U.S., and substituting this production under a formal contract to a foreign source. One view is that this activity kills jobs in the U.S., but another view is that expansion abroad goes hand in hand with expansion at home. A number of academic studies support the second view, but Mataloni observed that the relevant statistics are incomplete, and data can be found to support each view. Relevant BEA data include those on direct investment positions and capital flows, the operation of U.S. multinational companies (US MNCs), and trade in services.

Despite concern with the production and employment of US MNCs, Mataloni explained that these are still highly concentrated (about three quarters) in the U.S. There is a slight downward trend in this percentage, but growth abroad is largely in high and middle income countries. Mataloni noted further that foreign affiliates of U.S. companies mainly serve local markets, with about two thirds of sales going to local consumers. As Mataloni put it, US MNCs are chasing local consumers more than low wage labor. And the percent of U.S. purchases from foreign sources has been roughly stable over the years. What has changed, according to Mataloni, is that foreign operations are growing faster than domestic operations, and the host economies are growing more rapidly.

Mataloni then described some limitations in BEA data, which can mask firm level heterogeneity, reflect a limited number of observations, and tell only what happened—not why. He cited two academic studies that find little evidence of labor substitution between parent companies and foreign affiliates, but noted that these studies also are limited. For example, the studies are based on firm level data, limited to manufacturing firms, and are at least 10 years out of date—predating the recent concerns over offshoring. And in response to a question, Mataloni confirmed that employment and economic activity in the U.S. territories is not included in the definition of offshoring.

BEA hopes to enhance its measures in this area with initiatives including the expansion of detail in its employment surveys, future links with BLS occupation data and input/output accounts, and more academic research.

Ed Spar’s Executive Director’s Report

Ed noted the retirement of Bob Parker as chief statistician at GAO, and that GAO is seeking a replacement. Ed also reported that the Population Reference Bureau has announced a new seminar series (contact Linda Jacobsen for details). PRB also has initiated a bulletin on the American Community Survey.

None of the statistical agency budgets are passed, with the exception of the Bureau of Transportation Statistics, where numbers are down. Ed commented that this is a major problem for BTS, where the director is no longer a presidential appointee, and the agency faces an uncertain future. COPAFS has sent a letter of support to the assistant Secretary of Transportation, and will report on any developments.

Next Ed touched on the controversy at the Bureau of Justice Statistics in which the director was forced out in a dispute over the content of a press release describing a report on police stops. It is unclear what the final situation will be, but COPAFS has commented on the situation and written a letter for the American Statistical Association to sign.

There is once again suspense over the Census/ACS budget, as the House voted full funding for the ACS, but the Senate mark implies a major cut. Similar to last year, the Census Bureau says it will suspend the ACS if final funding is not near the House mark. COPAFS will do its best to “educate” the relevant appropriators. Ed noted that an ACS Users Guide will be released September 30. This will be a “first edition,” and Ken Bryson (attending for the Census Bureau) stressed that the guide is a work in progress.

The Release Rules for the American Community Survey Tables

Douglas Hillmer. U.S. Census Bureau.

Hillmer described last year’s Federal Register notice on ACS data products, and reported that that some user recommendations are already being incorporated. But his presentation was on the data release rules for ACS base tables (aka detailed tables). Hillmer clarified that the data release rules do not apply to the five-year data, where their application would impair the objective of long form replacement. Accordingly, no population thresholds will be applied to the five-year ACS data. Thresholds are applied to the one–year data (minimum population of 65,000) and the three-year data (minimum population of 20,000).

Hillmer described the roughly 500 base tables as the foundation for other ACS data products. They are modeled on SF3, but provide some additional crosstabulations. Among the issues the Census Bureau has yet to resolve are the inclusion of persons in group quarters, the introduction of several question changes in 2008, and the assurance of disclosure avoidance and statistical reliability.

With disclosure avoidance and statistical reliability in mind, the Census Bureau will not release all available estimates to the public (“estimates” is now the preferred term for ACS data). It is here that the ACS faces what Hillmer described as an “identity crisis” on reliability. Again, the five-year data are viewed as a long form replacement, so data release rules are not applied. But the one-year and three-year data are viewed as more like demographic surveys, and the intent is to ensure high levels of statistical reliability through the application of data use rules.

The data use rules begin with a “filtering” process designed to identify “weak” tables. For each table in each geographic area, the coefficient of variation is computed for each cell. If the median coefficient for all cells in a table is greater than 61 percent, the table “fails” for that area (61 percent is the level at which over half the cells are not statistically different from 0). If the table fails, data can be reported for a collapsed version of the table—with fewer cells. If the collapsed version of the table also fails the reliability check, the data for that table are suppressed for the area in question.

Collapsed versions of the base tables have been designed, and testing (and possible re-design) will begin soon. Hillmer presented an example of a collapsed table, and noted that the Census Bureau will publish the structure of the collapsed tables well in advance of the summer 2006 data release.

Improving the Dissemination of Economic Census Tabulations.

Robert Marske. U.S. Census Bureau.

Marske started by contrasting the demographic data the Census Bureau collects from households and the economic data collected from businesses. The economic data derive from a number of annual, quarterly and monthly economic surveys providing limited detail for national level, and the Economic Census, which provides detailed data for smaller areas once every five years. Public sector uses of the data include benchmarking, tracking economic change, attracting new businesses, and assisting business development. Private sector uses include studies of specific industries and markets, and the evaluation of investments.

Marske described the importance of industry classification to economic data, and the challenges posed by the transition to North American Industrial Classificaiton System. Even the NAICS classifications are updated regularly as new industries are added. The revisions can be disruptive to time series, but are considered necessary. As Marske observed, “there were no Internet providers in colonial times.”

Again, the Economic Census provides data for sub-national areas, including states, metropolitan areas, counties and places of 2,500 or more population. Data now are provided for the new micropolitan statistical areas, which has the effect of increasing the number of individual counties for which data are reported. And the data now are available on AmericanFactfinder. Marske also drew our attention to the Subject Reports (separate reports for each sector), Nonemployer Statistics (businesses without paid employees, which account for 70 percent of all businesses and 3.5 percent of all sales), and the Survey of Business Owners (formerly the surveys of Minority- and Women-Owned Business Enterprises).

Summarizing “what’s new” for the 2002 Economic Census, Marske cited updated industrial classifications, expanded service sector data, the expanded Survey of Business Owners, micropolitan statistical areas, and data on CD/DVD and American FactFinder. He noted that printed reports have been almost completely eliminated, and mentioned that they are having second thoughts about CD/DVD – and in particular PDFs, which take about as much effort and resource as printed reports. In response, several attendees described the advantages of PDFs, and made the case for their retention. Marske noted that they have not committed to eliminating PDFs, and in fact have a questionnaire seeking input on this issue. He invited COPAFS attendees to submit one – or more.

Concerns of COPAFS Constituencies

No concerns were raised, and the meeting was adjourned.