Dear Chuck, Very interesting of your comparisons. In order to have a fair evaluation, we need to have a list of assumptions (methods, procedures, etc) , input data (samples, indicators, etc), reference date (i.e. 1/1/05; 1/1/06; 7/1/2005; 2005 average; etc), and components (total pop, gq pop, household pop, household size; total housing units, occupied housing units, vacant housing units, vacancy rate) for comparisons. The 2004 (12-month) average ACS estimates provided estimates of total housing units and their implied vacancy rates but the 2005 ACS estimates do not have the data. We need to process the 2005 ACS PUMS file to figure out the size of samples for evaluation on the derived household sizes to see if the changes in household size are real. If we can get the numbers of vacant housing units (no response in ACS household survey) from the ACS PUMS samples, then we can calculate to obtain estimates of total housing units for comparison and for evaluation. We have collected some housing indicators (monthly BPUs, COUs, DMUs, and Others) from quite a few municipalities for estimating annual new housing units besides the USCB's (U.S. Census Bureau) annual BPU (Building Permit Units) survey data. It is very important to realize that we need a very good estimate of household population to fill into new housing units and should not apply a traditional housing unit method to use the number of new housing units to claim a certain numbers of household population. For instance you have picked 100 apples (persons) from your apple trees in your yard to fill your fruit baskets (housing units). It does not matter how many new fruit baskets you have brought with you, you still have only 100 apples to fill the baskets. Just please compile the annual (2000-2005) numbers of the U.S. population and numbers of the U.S. housing units for comparison and for analysis and you will get the right answers. Some of the 2005 ACS point population estimates are higher than that of our 7-1-2005 annual Colorado county and municipal population estimates but they are still falling within the lower and the upper bounds of 90% confidence interval. That means the differences are not statistically different at 90% confidence interval even though we are not satisfied with the conclusions. At least we can add the ACS survey estimates to our database for reference. Nice to share. - Richard Lin, Ph.D. Demographer Colorado Division of Local Government (303)866-4989, fax (303)866-2660 richard.lin(a)state.co.us TO: CTPP-News Listserv FR: Chuck Purvis This is an email I prepared for our Bay Area and California Census listservs, regarding newly released ACS data. PLEASE, if there are other metro or state agencies who've analyzed the ACS household population relative to their independent estimates of household population, PLEASE share your research! * * * * * * * TO: Bay Area Census Listserv FR: Chuck Purvis, MTC The "first wave" of the American Community Survey data for 2005 was released yesterday, August 15th. This first year of the full ACS database EXCLUDES group quarters population, so any of the "TOTAL POPULATION" tables as shown by the Census Bureau's American FactFinder are actually "HOUSEHOLD POPULATION" (or population in households, whichever term you prefer.) The data for total population that INCLUDES group quarters will be included in the 2006 ACS data, to be released summer '07. Over the past year we have had concerns that the household population in the ACS (the 2004 ACS, to be particular), was significantly lower than decennial census counts of household population, and households. This is now a real concern with the 2005 ACS, as well. What we are able to compare for California is the household population estimates from the 2005 ACS, to our State Department of Finance's (DOF) estimates of total population, household population, GQ population, housing units, occupied housing units, at the city and county level, from their "E-5" estimate reports, available on the DOF web site (http://www.dof.ca.gov/HTML/DEMOGRAP/Druhpar.asp). There are standard errors in the DOF data, as suggested in their methodology, but DOF doesn't provide explicit confidence intervals around their estimates, at least as far as I could tell. So, I've been able to analyze the data for the 119 California Places (116 cities + 3 census designated places); and the 40 California Counties, that exceed the 65,000 household population threshhold for reporting ACS totals. I've also created tables for the San Francisco Bay Area Cities (25 cities) & Counties (9 counties) in the same EXCEL workbook. This EXCEL workbook is available from the MTC web site, at: ftp://ftp.abag.ca.gov/pub/mtc/planning/ACS/ Of the 119 California Places with household population estimates exceeding 65,000, thirteen of these cities are significantly different, comparing the 1/1/05 DOF (Department of Finance) household population, and the 1/1/06 DOF household population estimates, to the 95 percent confidence interval around the ACS household population estimates. Unfortunately this list includes six of the ten largest California Cities: 1. Los Angeles: ACS = 3,668 to 3,794 thousand; DOF = 3,852 to 3,892 thousand (estimates for 1/1/05 and 1/1/06) 2. San Diego: ACS = 1,176 to 1,240 thousand; DOF = 1,254 to 1,263 thousand 3. San Jose: ACS = 866 to 908 thousand; DOF = 931 to 943 thousand 4. San Francisco: ACS = 699 to 739 thousand (estimated); DOF = 773 to 778 thousand 8. Oakland: ACS = 355 to 393 thousand; DOF = 403 to 404 thousand 10. Santa Ana: ACS = 277 to 328 thousand; DOF = 345 to 346 thousand. The other seven California places with a significant difference, comparing ACS to our DOF, include: Garden Grove (Orange County); Salinas (Monterey County); Pasadena (LA County); Daly City (San Mateo County); Berkeley (Alameda County); Alhambra (LA County); and Merced (Merced County). A critical check in the coming months will be to compare the ACS estimates of housing units and households (occupied housing units) to independent estimates of housing units and households, also available from the DOF Report E-5. Perhaps the decline in population in these large California cities can be attributed to decreasing average household size. On the other hand, it is a lot harder to explain any absolute losses in housing units within a community. It may be useful for the City Planners in Large Cities to have at hand independent estimates of new housing units constructed and old housing units demolished over the 2000 to 2005 time period. These are very serious concerns about the ACS. I believe it has more to do with the "sample frame" and weighting/expansion issues, as opposed to the quality of the characteristics derived from the ACS, but this is an important topic that should be discussed by our state data center network, the Census Bureau and interested stakeholders. Hope this helps, Chuck Purvis ************************************************************** Charles L. Purvis, AICP Principal Transportation Planner/Analyst Metropolitan Transportation Commission 101 Eighth Street Oakland, CA 94607-4700 (510) 817-5755 (office) [new, 8/1/05] (510) 817-7848 (fax) [new, 8/1/05] www: http://www.mtc.ca.gov/ Census WWW: http://www.bayareacensus.ca.gov/ ************************************************************** _______________________________________________ ctpp-news mailing list ctpp-news(a)chrispy.net http://www.chrispy.net/mailman/listinfo/ctpp-news