I received this email from Bob Paddock today, and thought that a reply to the listserv might be useful... -----Original Message----- From: Bob Paddock [mailto:bob.paddock@metc.state.mn.us] Sent: Thursday, April 07, 2005 12:14 PM To: Murakami, Elaine Subject: TAZs and the 2010 Census Elaine, If my question should be addressed to someone other than yourself, please let me know. I have been discussing the future of our agency's TAZ system with the person responsible for our travel demand modeling. Collectively we have wondered just how the TAZ system will be utilized for the 2010 Census and ACS survey data. Since we defined our TAZs in the past based on Census block aggregations, will such efforts be needed for the year 2010? If so, how will all of this work with the ACS? Any thoughts on this? BOB ------------From Elaine------ For purposes of this reply, I will assume that there will be a CTPP-like product from a 5-year accumulation of ACS data. 1. Census Bureau Geography Division is already planning to reserve two TAZ fields for TIGER for 2010 Census. Conceptually, these could be considered as a small "urban" TAZ, and a larger TAZ for statewide planning purposes. However.... 2. The sample sizes from the ACS are much smaller than for decennial census due to 2 main reasons: a. The sample size of 2.5 % per year * 5 years = 12.5% is less than the 16.6% (1:6 h.u. sampling rate) for the decennial census Long Form. (Original design of ACS was 3% per year, so that 3% * 5 years = 15% which is much closer to the 16.6% for LF) b. The ACS non-response follow-up is a sample of 1 in 3 (generally speaking), but higher in areas with low mail-back returns. c. Thus, in the ACS test areas, the actual number of unweighted survey records is 50-60% that of the Long Form. This smaller sample size means tabulations for small geographic units is highly variable and much less reliable, and may be subject to data suppression to protect individual confidentialy. This has implications for any TAZ-based tabulation, and has severe implications for any FLOW tabulation. Under the assumption that the CB retains the requirement of 3 unweighted records for FLOW tabulation (same as imposed on the CTPP 2000), we expect that about 40-50% of tract-to-tract flows would be suppressed. 3. What might be some alternatives? Would people want to define TAZs, based on census blocks, that are closer in size to a census tract? Would census tracts be sufficient as is? Would different TAZs be needed: one for Residence only and Workplace only tabulations, and ANOTHER larger TAZ for FLOW tabulation? Assuming that State DOTs and MPO still want to define TAZs, it would be safe to assume that we would use a GIS-based product similar to TAZ-UP for the CTPP 2000. One outcome may be that the TAZs in your MPO model are likely to be SMALLER than any geographic area tabulation you get from the ACS. You will have to figure out a way to distribute ACS results to your model's TAZs. 4. CB has a project called the LEHD http://lehd.dsd.census.gov/led/ They have been doing some exploration on generating small area home-to-work flows from this, but although this has characteristics like age, sex, earnings, it doesn't have mode to work, travel time, departure time. Also, multi-site businesses, government employment, and other issues on the workplace geography are still being researched. My personal opinion is that there are probably issues on the residence geography as well, e.g. particularly areas with "transient" populations such as students, snowbirds, and migrant laborers. 5. And, of course, we haven't even discussed the quality of the workplace geocoding, so there are plenty of things beyond TAZ definition that need to be addressed regarding a CTPP from ACS. If you haven't planned yet to come to the TRB Census conference (May 11-13), now is the time to sign up! http://www.trb.org/conferences/censusdata/ Elaine Murakami FHWA Office of Planning 206-220-4460 in Seattle