Robert:

 

Well, the ACS data is what it is – can’t get around that.  The sample size is smaller than that used for the 2000 Decennial Census sample survey (the equivalent of about 12% versus about 16%), so there are indeed many who are dissatisfied with the resulting higher margin of error.  This dissatisfaction is probably increased because Census did not report the margin of error until the ACS came along, in the past allowing Census users (myself included) some degree of blissful ignorance of the actual level of uncertainty.  At the small area level, notable error rates actually have been in the sample data all along.  Whether it will “not be good” at the block group level will depend on how good it has to be for your purposes, and on what data element you are measuring (e.g., smaller percent margin of error for SOV rates than for bicycle commuters).

 

Block groups are of comparable size nationwide, for small communities and large, ranging in the 2000 Census from about 600 to about 3000 and averaging about 1200 (the 2000 Census geography will be that used in the 2005-09 ACS product).  The ACS block group data consequently will be of comparable utility for small communities and large, since it is the size of the block group that counts, not the size of the community.  One exception to this rule: Census uses higher sampling rates in low population areas (e.g., small towns and small rural counties) so that they can get a large enough sample to achieve statistical significance when they have to report data for small political jurisdictions.

 

Bottom line: for better or worse, ACS 2005-09 block group data is all we will have in the next few months with small area socio-economic and journey-to-work characteristics, unless one has the resources to do one’s own surveys.  As Ed Christopher pointed out, the 2010 block data also will be available for our use in delineating new TAZs, though it will be limited to basic demographic information, dwelling counts, and owner vs. renter vs. group quarters.