RE: [CTPP] Some more points

Patty.. I must correct your misunderstanding in the statement below. The Census Bureau has NEVER used computer algorithms to delineate block groups (BGs). Computer algoritms have only been used to determine the blocks and to number the blocks. Even for the 1990 census where there were BNAs, it was the geographic staff in the Regional Offices who delineated the BNAs and BGs, albeit with an eye towards field operations in the counties with BNAs, not data tabulations. In 1990, the participants who worked on census tracts also were allowed to suggest changes to the BGs for 1990. For Census 2000, the participant statistical areas program asked/allowed/encouraged all participants, nationwide, to review and suggest changes to the the census tracts (including the former BNAs) and BGs. Some participants did review and suggest changes to the BGs, many did not. Where participants did not change the BGs, the geographers in the Regional Offices made any necessary changes in accordance with the guidelines. The guidelines used by participants are on the web at: http://www.census.gov/geo/www/statarea.pdf I would not equate, in any fashion, TAZs to census tracts. The purposes are much different -- for TAZs, the goal is to get zones of approximately equal trip ends so that the trip distribution models work better. For census tracts, the goal is to maintain comparability over time, with a minimum/ideal number of people with generally the same demographics. Thus for TAZs, ideally there would be many, small TAZs in the business district (for the work trip ends) where the CTs would be large because of minimal residential population, while the comparable number of trip ends in a residential area might cover several CTs. Bob LaMacchia Geography Division U. S. Census Bureau -----Original Message----- From: owner-ctpp-news(a)chrispy.net [mailto:owner-ctpp-news@chrispy.net]On Behalf Of Patty Becker Sent: Sunday, January 27, 2002 5:00 PM To: ctpp-news(a)chrispy.net Subject: RE: [CTPP] Some more points Robert, If you actually delineate block group boundaries, you are a rare exception. Yes, I do delineate some of them locally, primarily when tracts have been combined in the City of Detroit (to maintain data for the old tracts), or when there is a unique residential land use (such as a public housing project) which is not large enough to be a tract. I understand that in a smaller metro area, such as Abilene, you probably have more areas which need such special treatment. However, I would venture to say that at least 95% of the BGs in the nation are "accidental," where the BG lines have been drawn by the Census Bureau's computer with no human intervention. However much needed, block group data are basically very, very shaky. At 500 housing units per block group (a rough maximum; most are smaller), you have only 50 cases in a 1 in 6 sample, to say nothing of having fewer because of vacancies, non-response to the long form, and item nonresponse. The sampling error on a distribution of 50 cases is roughly +/- 14 percentage points, if memory serves me right. (I think it's 14 on an N of 50 and 7 on an N of 100.) The non-sampling errors make the data worse. That's why I recommend not using individual block group data. Of course, I come from a large metro area with reasonably high density. I can draw tracts which work so that we don't need block group data. That's also why the TAZs are more like tracts than blocks. I know that it's much harder when you have a smaller population base and still want to work at the small area level. Just keep those error levels in mind, though. Patty Becker At 03:03 PM 01/27/2002 -0600, you wrote: analysis was ---------------------------------------------------------------------------- --------------------- Patricia C. (Patty) Becker 248/354-6520 APB Associates/SEMCC FAX 248/354-6645 28300 Franklin Road Home 248/355-2428 Southfield, MI 48034 pbecker(a)umich.edu