Fall 2001 Meeting Minutes
quarters. The census provides a new base for population estimates, and the Bureau is trying to determine where that base came from and exactly what it means. Important factors that have been identified include count improvement and immigration. In some cases, the changes in the census count may not have been improvements. An additional decision on census adjustment is scheduled for October 14th, 2001. In March 2001, the Bureau decided not to adjust the data provided for redistricting. SF1 was then released without adjustments. The next decision pertains to the population estimates base and federal survey controls, and it will reflect additional data that were not available before [Note: The Bureau decided not to use the adjustments in long form data, the population estimates base, or survey controls. Adjustments were used in survey controls for the 1990’s.] The Bureau has intensively examined immigration data, using CPS data re-weighted to the 2000 Census, ACS results, and preliminary tabulations from the immigration question on the census long form. The Bureau is getting a better handle on foreign immigration than ever before. Data on temporary migrants, work permits, out-
Notes on Fall 2001 FSCPE Meeting Kansas City, MO September 10-11, 2001
Note: An overview of these notes can be found in the document "overviewfscpeKS" on the Bureau’s website. (http://www.census.gov/population/www/coop/s erver/agenda/overviewfscpeKS.html).). in the Additional information can be found meeting handouts and in the slide shows for each presentation, which are linked to the meeting agenda on the Bureau’s website (http://www.census.gov/population/www/coop/s erver/agenda/fscpeKS.html ). The meeting was opened by FSCPE Chair June Nogle. Introduction and Welcome: John Long, Chief of Population Division John thanked FSCPE members for all the work they have done on the census as well as on population estimates. The Bureau is currently digesting information from the census, which counted more people than expected. Problems were also found in count review, particularly regarding group
Fall 2001 FSCPE Meeting Minutes 2migration, and other components of migration flows are being examined. In past analyses, it has been assumed that 1% of births are unregistered. Some evidence suggests that birth registration is now almost complete. The Census Bureau welcomes opportunities to continue to work on its relationship with the coops. That relationship has grown, as reflected in count review. The rest of the Bureau has come to understand what the coops can accomplish, and the Bureau is looking for ways to work together, and for ways to use information from the coops in time to improve census data. The Bureau may work more closely with the coops on local data from the census and the American Community Survey. In the American Community Survey, group quarters will not be handled through the same process as the household population. The ACS will have to find a new method for dealing with group quarters, and that may involve working much more closely with the coops. Updating of the Master Address File and TIGER are also necessary for ACS, and that will also involve working with the coops. It was noted that one advantage of the ACS is that activities can be refined from year to year. Update from Population Estimates and Projections Area / Signe Wetrogan Personnel issues: Lisa Blumerman is acting chief [now chief] of the estimates branch. She has a degree in applied demography from Georgetown and a degree from the population studies program at the University of Massachusetts. She has worked as a special assistant for John Long, and she previously worked with the American Public Welfare Association. Goals for the population estimates program include: maintaining the current product line timeliness [a list of dates for input data submission and data releases for 2001 estimates was included as a handout] completing evaluation of estimates from the 1990’s introducing the Census 2000 base in estimates for 2001 enhancing methodologies improving dissemination re-benchmarking estimates for the 1990’s The annual release schedule will be: state totals end of December county totals April subcounty totals July county age/race/sex/Hispanic June or July state/county households and housing units June or July National projections are due in late 2002 or early 2003, and state household projections are due in July 2003. The Bureau has completed its preliminary analysis of estimates for April 1, 2000. Unlike previous decades, the Bureau completed a full set of estimates for 2000, rather than just extrapolations of estimates from a prior year. Estimates were prepared for state, county, and subcounty totals, county age/race/sex/Hispanic origin, and county housing units and households. These estimates for April 2000 and evaluation results are scheduled to be available by October 2001. There will be an “evaluations” section of the population estimates
Fall 2001 FSCPE Meeting Minutes 3website. Evaluations of projections are also underway. Evaluation results will be presented at the Southern Demographic Association meeting in October and at the next FSCPE meeting. The estimates will use census counts as a base even where there are known to be group quarter problems. The methodology will not be affected by the decision about census adjustment. New estimates will reflect the OMB race reporting guidelines. Estimates will be developed for 31 race categories. New modified age-race-sex (MARS) data will also include multiracial detail. Some categories may be combined for dissemination An “alone or in combination” approach may be used. Planned methodological improvements for the county age/race/sex/Hispanic estimates include introducing a cohort-component approach, and using Social Security Administration and IRS migration data. Other planned improvements include integrating American Community Survey data with the estimates, improving handling of Special Census results, and improving estimates of international migration. New estimates will reflect DAPE (demographic analysis population estimates), and national results for non-immigrants, unauthorized immigrants, and emigrants. The delay in reflecting boundary changes in the estimates will be reduced. The 2001 estimates will reflect census geography, and the 2002 estimates will reflect the 2002 Boundary and Annexation Survey results for December 2001 boundaries. Methodologies submitted by several states are being evaluated for incorporation into the Bureau’s estimates for 2003. There will be a 3-tier approach for each product including the estimates themselves, a description of the methodology, and a report discussing the estimate results. Estimates will be disseminated via the Internet through American Factfinder and through the estimates and projections site. In addition to tables, there will be maps, charts, data files, and press releases. There will also be “cross product analytical reports” such as reports on coastal counties, metro fringe counties, etc. A new set of estimates for the 1990’s will be prepared that will benchmark to both the 1990 and 2000 censuses. [Update: These estimates are expected by the end of April.] The new national and state totals will be released in 2002, and they will incorporate new information on international migration and mathematical smoothing. New age and race estimates will be released in 2002 or 2003. They will be consistent with new age and race categories, and they will reflect Census 2000 sample data as well as results of research on changes in race identification. Estimate Evaluation 1: Demographic Analysis and Pop Estimate Project (DAPE) / Lisa Blumerman The DAPE research project started in April and involves 60-70 people. The work of this group influenced the March 15, 2001 decision by the ESCAP committee not to adjust the census for undercount. The population figure which initially emerged from demographic analysis was lower than the census count. Alternative scenarios indicated areas for additional research, particularly international migration.
Fall 2001 FSCPE Meeting Minutes 4[Note: In the Bureau’s “demographic analysis” methodology, the national population for each age/race/sex category is estimated from historical birth statistics, death statistics, and migration estimates. The data reflect adjustments for under-registration of vital events, unauthorized immigration, and other factors. Another demographic approach involves applying recent birth, death, and migration estimates to alternative 1990 base figures.] DA provides an independent measure of census coverage. It allows assessment of the reasonableness of the change observed from one census to another; and it allows assessment of the consistency of implied change for age/race/sex groups. New benchmark data were developed; CPS results were weighted to 2000 population statistics; Census 2000 Supplemental Survey data were used, along with provisional Census 2000 sample data. The 1990 base was reviewed for all population components, and estimates of change from 1990 to 2000 were revised. Three meetings were held to consult with outside experts, including a meeting at the Census Bureau on 3/20/01 and meetings at the Population Association of America meeting on 3/29/01 and 3/30/01. The short-term research agenda which emerged from these meetings included a focus on international migration and studies to validate the estimated components of intercensal population change. The longterm agenda included dedication of Bureau staff to international migration studies, and increased collaboration with Mexico and Canada. The DAPE project included 15 research projects, scheduled for completion in Fall 2001. Topics include birth data, Medicare data, legal immigration, unauthorized immigration, emigration, temporary migration, other migration flows, and consistency and quality of foreign-born data. New DA estimates are due in late Fall 2001 to reflect new benchmarks and revised levels of international migration. Long-term plans include annual estimates of major components of national demographic change, re-calibration of vital events to better match race responses in the census. This will produce more accurate and reliable measures of the size and characteristics of the national population. Lisa can be reached at Lisa.M.Blumerman@census.gov, or at 301-457-2385. Estimate Evaluation 2: National Estimates / Melissa Therrien Cohort-component estimates by age/race/sex/Hispanic origin were examined. The new multiple-race census question prevents direct comparison of estimates to census counts and produces minimum and maximum populations for each race. Evaluation focused on numeric and percent differences and sex ratios. For evaluation purposes, the population was grouped into six broad age categories from 0-17 to 85+. Overall, the census count was 2.4% higher than the estimates. Counts were 2.6% higher for age 0-17, 4.7% higher for age 18-29 and only 1% higher for age 65-84. The census was 0.4% lower for age 85+. The census generally exceeded the estimates by a higher percentage for males than for females. Exceptions were the age groups 65-84 and 85+. For Hispanics, the census was 8.9% higher overall for the census, including a 19.7% difference for age 18-29. However, the es
Fall 2001 FSCPE Meeting Minutes 5timates were 9% higher for ages 65-84 and 33% higher for 85+. Slides were presented for each race, including separate bars for people reporting that race alone and people reporting that race in combination with other races. Estimates are higher than census counts for age 85+ for each race category. Census counts are generally higher than the estimates for most other age groups, particularly age 18-29. The ratio of males to females falls more slowly with age for the census than for the estimates. Census 2000 briefs on how to talk about race can be found at: www census gov /prod/www/abs/briefs.html Explanations discussed for these patterns include changes in census coverage, international migration, quality of mortality data, and differences in census age reporting. Estimate Evaluation 3: State/County Estimates and Characteristics / Josie Baker The state and county estimates for April 2000 are interpolations between July 1999 and July 2000 estimates. The estimates are 2.4% lower than the census overall. The MAPE (mean absolute percent error, a measure of accuracy) is 3.3%, down from 3.6% for 1990. The MALPE (mean algebraic percent error, a measure of bias) is –1.6, compared to +1.8% for 1990. The census count exceeded the estimates for roughly two-thirds of all counties. MAPE’s were higher for the west and south than for the midwest and northeast. Among states, MAPE’s were highest for Hawaii, Nevada, Arizona, and Colorado, and smallest for Ohio, Iowa, Maryland, and Pennsylvania. [See talk handout and table of state estimates.] MAPE’s tended to be highest for counties with fewer than 2,500 persons, and lowest for counties with 50100K persons. MAPES tended to be lower in 2000 than in 1990 for most county size categories, particularly categories ranging from 5K to 50K persons. However, MAPE’s were higher than in 1990 for counties with over 100K persons. MAPE’s tended to be lowest for stable counties and higher for counties with significant population declines or with rapid growth. MAPE’s were highest for counties with very low or very high group quarter populations, and for counties with very low or very high percentages of population over age 65. MAPE’s tended to be higher for counties with higher proportions of Hispanic residents, and for counties with higher rates of growth for their Hispanic population. MAPE’s also tended to be higher for states and counties with higher percentages of population growth through international migration. It was noted that counties with very low or very high percentages of group quarter population tend to be very small counties, which may explain why they tended to have high MAPE’s. The Bureau’s estimates were also compared against two simple alternatives: using 1990 census counts in place of population estimates, and assuming the same rate of growth between 1990 and 2000 as was observed between 1980 and 1990. It was noted that the 2000 estimates differ from the Bureau’s normal population estimates because they were not subjected to FSCPE review. However, normal review was not possible because FSCPE input could have been biased by the familiarity with 2000
Fall 2001 FSCPE Meeting Minutes 6census counts that was gained through the count review process. Out of the three alternatives, the population estimates have the lowest MAPE (3.3%). Assuming no change from the 1990 census counts produces the highest MAPE (11.1%). Assuming the same rate of growth as from 1980 to 1990 produces a MAPE of 9.2%. Out of the three alternatives, the estimates perform best for 72% of all counties, while the assumption of no change from 1990 census counts performs best for only 12% of counties. The assumption of no change in the rate of growth performs best for 15 percent of counties. The estimates out-perform the two simple alternatives for all county size categories. Conclusions: The estimates for the 90’s performed slightly better than those from the ‘80’s, with typical patterns with respect to county size and growth rate. The estimates achieve roughly two-thirds of the potential improvement relative to assuming no change in the rate of growth, and they achieve more than two thirds of the potential improvement relative to assuming no change in population. Additional research is needed with respect to the causes and implications of the relationships between county characteristics and error levels. Estimate Evaluation 4: Subcounty Estimates / Jason Devine The subcounty estimates were produced for 40,630 areas using housing unit data to distribute county population totals among subcounty units. Errors are caused by: inaccurate housing unit estimates inaccurate or incomplete boundary change processing inaccurate county population estimates inaccurate assumptions about vacancy and occupancy rates inaccuracies in the 1990 or 2000 censuses.
The MAPE for 2000 was 12.4%, which represents an improvement of 2.8 percentage points relative to 1990. Half of the errors were positive, and the MALPE was 3.4%. Very small units elevate the overall MAPE. The MAPE for units with fewer than 100 persons was 35.1%, while the MAPE for places with more than 2,500 persons was only 6.1%. The MAPE for places with between 50K and 100K persons was 4.0%. MAPE’s improve as size increases, except that the MAPE is slightly higher for places with more than 100K persons than for places with between 50K and 100K persons. There is fairly little variation in MAPE among size categories above 10K persons. MAPE is very high for places growing by 50 percent or more (25.8%) or declining by 15 percent or more (45.6%). The MAPE for places growing less than 25 percent or declining less than 10 percent was below 10%. Areas that lost population tended to lose more population than estimated, and areas that grew rapidly tended to grow more than estimated. Positive errors represent nearly 100% of the total for areas with the largest rates of decline, but they progressively drop to about 10% of the total for areas with the highest rates of growth. The lowest MAPE’s were in Indiana, California, Michigan, New York, Pennsylvania, and New England. The estimates are more accurate than an assumption of no change in population since 1990 for 73% of places with 50K to
Fall 2001 FSCPE Meeting Minutes 7100K persons, but for only 51.6% of places with 100 to 499 persons. For places with fewer than 500 persons, MAPE’s are lower for an assumption of no change since the prior census than for the estimates. Nevertheless, this represents better performance for small areas than the subcounty estimates for 1990. The accuracy of estimated population within counties was measured with the Index of Dissimilarity. This eliminates error due to inaccurate group quarter information or county population estimates. By this measure, the estimates tended to be less accurate for the west, the mountain states, and the south than for the northeast and midwest. Estimate Evaluation 5: Housing Estimates for States / Jason Devine The housing unit estimates have only been released at the national and state level, but county estimates have been prepared for 2000 and for each year in the ‘90’s except 1997 and 1999, and they have been used in analysis related to the Master Address File. The estimates are based on building permits, mobile home shipments, and estimates of housing loss. Errors can be caused by faulty building permit data, inaccurate mobile home allocation, inaccurate housing loss models, and faulty housing counts in the 1990 or 2000 censuses. The national estimate of housing units was 0.3% lower than the number counted in the census, which is very close to the 0.26% figure for 1990. The MAPE at the state level was 1.5%, which is down slightly from 1.75% in 1990. Half of the states had absolute errors of less than 1%. The absolute error was highest in Montana (5.9%) and lowest in Minnesota (0.008%). At the state level, error rates tended to be higher for small states and lower for large states. There was no clear relationship between growth rates and error rates at the state level. The MAPE at the county level was 4.6%. Error rates tended to be higher for small counties and lower for large counties. Error rates also tended to be higher in counties with high rates of increase or decrease in their number of housing units. In general, the estimates based on building permits were more accurate than assuming no change in housing unit counts since the prior census or assuming that housing unit counts follow population estimates. Questions raised in discussion included: Would a population-based estimate be useful for areas with poor housing unit data? Has there been an evaluation ofwhether GUSSIE made the estimates better or worse? Can the estimates be improved through demolition data and other housing data that is available for some states, even though it is no longer available through federal sources? Discussion of Avenues for FSCPE Participation / John Long & June Nogle The interaction of states with the Bureau was seen to be helpful, by both the states and by the Bureau. Three core areas for increased interaction are evaluation of group quarter data, improvement of housing data, and the American Community Survey. State participation can improve the ACS, and data from the ACS can improve population estimates.
Fall 2001 FSCPE Meeting Minutes 8Data from states can improve the Bureau’s information on group quarters. Problems were cited with missed dormitories and deteriorating data on defense facilities. A post-2000 LUCA program for local review of census addresses is under consideration. Confidentiality is a problem with group quarter data. For example, shelters for abused women are often very secretive about the location of facilities. Even if information originally comes from a public source, it becomes confidential once an interviewer is sent to confirm the name of the facility. One possible solution is to share a separate data file with states that does not reflect information that is protected by Title 13. Count Question Resolution / Mike Batutis A total of 4,608 issues were raised in count review, including 1,303 that were documented by FSCPE representatives. The Count Review staff delivered 1,994 issues to CQR (including 731 issues from FSCPE review). However, CQR has accepted only 624 issues (including 329 issues from FSCPE review). Another 90 external cases had been received as of early September. The Count Review staff weeded out issues that did not point toward a solution. The criteria used by CQR staff to determine which cases would be accepted is not known. FSCPE and SDC agencies can submit additional cases to “Extended Count Review.” As of early September, 45 issues had been submitted by 5 states, including 18 issues already submitted and 11 that were already in CQR. Issues can be submitted by sending a letter to the CQR staff and to Mike Batutis. The letter should describe the nature of the issue, provide complete geographic information, and indicate possible resolutions. The review is scheduled to continue through the end of FY 2003. However, that is the date on which all activities must be completed, so the latest feasible time to submit issues would be earlier. Discussion: It was suggested that the Bureau should consider recertifying data where the population of a jurisdiction is affected. For example, one flaw in current census products is that a town in Alaska was placed in the wrong county. The possibility of re-running summary files after CQR is complete is still open. CQR corrections will be incorporated into population estimates and ACS. The sources of census GQ errors are still being evaluated. Some of the errors may have occurred in the field at the enumerator level. Possible errors include not having addresses that reflect physical location, submitting information that could not be processed, coding errors, and transposition of GQ type codes. Type codes are not always consistent. For example, some Hutterite communities are treated as GQ’s and some are not. Geocoding errors and missing facilities are the top priorities; faulty type codes are a lower priority. Facilities with lower counts than administrative records are unlikely to change. John Long will look into havingstates notified of specific CQR challenges that have been filed.
Fall 2001 FSCPE Meeting Minutes 9FSCPE agencies will also receive copies of the letters notifying communities of issues that have been resolved. A report on the count review program is due on November 15, 2001. It will include a history of the program, a discussion of the lessons learned, and documentation of FSCPE participation. Next steps include: Planning for 2010. Weekly meetings are taking place to plan the next census, and Mike Batutis is part of the effort. Staff from the Decennial and Geography divisions seem to be implicitly assuming that there will be count review in 2010. Evaluating count review. There is frustration that problems could not be fixed before data were finalized. The count review program was developed too late to have procedures in place to implement changes. The list review portions of count review can occur much earlier. There is a new attitude toward the importance of group quarter data, and a desire to get to the bottom of the GQ problems and get them fixed. This would not have come about as quickly without count review. Looking for other areas in which to collaborate. Census sample data will be reviewed by the Bureau’s subject matter analysts from November through February. The geographic division is interested in the GIS tools that were developed for count review, since they need to compare address counts against benchmark data. They may fund tool development that will be useful for future count review. Special Census Program / Lisa Blumerman An interdivisional team has recommended continuation of the Special Census program, which will begin in 2003. There is an issues team, a methodology team, a design team, and an implementation team. The Special Census program is actually a consolidated program that includes not only the enumeration option, but also the estimates challenge process, population certification, and some activities related to the American Community Survey. Special Census enumerations are reflected in population estimates, except that partial censuses are not considered. Annexations are included in the geographic base with a time lag. The Population Division will review all results. The population certification program involves certifying decennial population counts for updated boundaries, governmental mergers and consolidations, and new geographic entities. A fee is involved. The certified counts are accepted for most purposes that require “hard” population counts. Certification is also available for blocks that are split by special-purpose geographic areas. Population counts are derived from GUSSIE, and they reflect changes in TIGER and in the Master Address File. Information regarding this program will be announced in the Federal Register, and the program will be available in mid to late 2002. It was noted in discussion that Special Census results do not always improve the population estimates. For example, the special census for Maricopa county Arizona was conducted in February when Arizona has many seasonal residents. Based on the results of that special census, population estimates were decreased for other
Fall 2001 FSCPE Meeting Minutes 10counties across the nation without any evidence or validation process to determine that those estimates were too high. Challenge Process / Melissa Therrien Challenges are accepted for county and subcounty population estimates. The Bureau will use the same housing-unit based methods in 2001 as in the past, but the method is under review for future years. The method involves using data on building permits, certificates of occupancy, or utility connections in conjunction with decennial census data on occupancy rates and persons per household. Alternatives under consideration include updating occupancy rates and persons per household, and raking subcounty estimates to county totals. At the present time, a subcounty unit can submit data identical to the data used in the initial estimates, and its population estimate will change because the rake factor is not used. Challenges must be received by October 1, 2001 in order to affect estimates for the following year. That means that counties have about six months to submit a challenge, but subcounty units have only a couple months. The involvement of FSCPE agencies in the challenge process is also being considered. FSCPE agencies are notified when challenge packets are sent, but not when inquiries are received, and they do not get copies of submitted data. Challenges have been processed based on the data submitted and the established methodology. In some cases, estimates were increased because of challenges, and then the population went back down in the census. It was suggested in discussion that some communities are estimated better than others through a housing unit methodology, and that it might be best to use other methods for communities that have not been estimated well by housing unit methodologies in the past. FSCPE Business Meeting / June Nogle Because there was no fall meeting last year, there has been no decision on by-laws. It was decided that the current and proposed by-laws would be posted for feedback. There has not been an FSCPE newsletter for a while, due to time constraints caused by Census 2000 activity and postage costs. John Besl and June Nogle edit an applied demography newsletter which goes to FSCPE members as well as to a wider audience. The Spring FSCPE meeting will be held in Atlanta prior to the Population Association of America meeting. Suggestions for the agenda should be sent to June or to Ann Powell. Support was expressed for having the Fall 2002 meeting in Washington D.C. The potential merger of the estimates and projections coops was discussed. It was noted that participation in the projections coop has decreased, and that the Bureau seems to be viewing estimates and projections more as part of a unified whole. It was also noted that the fall meeting agenda does not go to FSCPP participants who are not also in the FSCPE. It was suggested that the FSCPE and FSCPP have joint sessions when appropriate.
Fall 2001 FSCPE Meeting Minutes 11November 11, 2001 Population Estimates Program for 2001: Overview / Lisa Blumerman FSCPE members are urged to contact Lisa Blumerman or Mike Batutis regarding any Census 2000 issues that require a special look by the Bureau, whether in CQR or in other contexts. Estimates are used to allocate funds amounting to over 200 billion dollars, and they serve as controls for the Current Population Survey, the American Community Survey, and other surveys. They also serve as denominators for statistics such as birth and death rates, per-capita income, etc. They also serve as a basis for decisions in the private sector. The population estimation methodologies for different levels of geography were briefly described. A handout provides timelines for input data and estimate products. Goals include maintaining the current range of estimates, ensuring timeliness, evaluating estimates from the 1990’s, implementing 2000 base data, enhancing methodologies, and improving dissemination. Areas for potential enhancement include international migration, boundary changes, implementing a cohortcomponent approach for county age/race/sex estimates, and integrating results of the American Community Survey and special censuses with the estimates. Improvements to the population estimates web page and other improvements in dissemination of estimates were described. Population Estimates Program for 2001: State and County Estimates / Melissa Therrien In addition to the population estimates that are publicly disseminated, outputs of the estimates program include controls for monthly surveys, denominators for various statistics, briefing tables for media and government use, and analytical tables. Input data include vital statistics from the National Center for Health Statistics as well as from FSCPE agencies, international migration data from the Immigration and Naturalization Service and the State Department, internal migration data from the Internal Revenue Service, the Social Security Administration, the Medicare program, and group quarter data from the FSCPE. The estimates for 2001 will reflect a reworking of the distribution of international migration at the sub-national level. The schedule for input data, review, and data release was presented. [See handout.] It was suggested in discussion that information or documentation should be provided for changes that the Bureau makes to IRS data. Some counties can be replicated from the IRS county-to-county file, but others are widely divergent. County age/race/sex/Hispanic estimates are scheduled for release on July 31, 2002. [Update: These estimates may be released at the end of August.] The Bureau is considering having FSCPE agencies review such estimates of population characteristics in the future. Population Estimates Program for 2001: Subcounty and Housing Unit Estimates/ Jason Devine The timelines and methodology for the subcounty and housing unit estimates were reviewed. Changes under consideration include improving the processing of building permit data and developing a different approach to
Fall 2001 FSCPE Meeting Minutes 12estimating housing loss. One idea under consideration is to use the rate of housing unit loss experienced from 1990 to 2000. Census data on mobile homes for 2000 are not yet available because housing type was only on the long form sample. Decisions about improving mobile home data have not yet been made. The Bureau is exploring the possibility of improving assumptions about vacancy and household size based on data from the American Community Survey, the Current Population Survey, and the American Housing Survey. The current methodology produces household estimates based on CPS data on household formation by age. The focus may shift in the future from households by age of householder to county housing unit and household data. County housing unit estimates will be released. The remainder of the meeting was cancelled due to terrorist attacks in New York and Washington and the suspension of air travel. State and County Estimates Produced by FSCPE Agencies / Sam Davis [Presentation cancelled. A slide show for this presentation is linked to the meeting agenda on the Bureau’s website. Incercensal Estimates Beyond 2001 / Signe Wetrogan [Presentation cancelled. A slide show for this presentation is linked to the meeting agenda on the Bureau’s website. Online GIS Demonstration [Presentation cancelled. A slide show for this presentation is linked to the meeting agenda on the Bureau’s website.]