Bob,
Well,
The first block groups, in 1980, were definitely done by computer. They
had no human input. After they were done, the bureau decided include long
form data by block group on STF3, and the private sector loved it. (This
was the replacement for Fifth Count File C, which for 1970 had provided
data by enumeration district.)
For 1990, most block groups stayed the same. Yes, it's true that the RO
staff did the BNA splits (usually down main roads, which meant areas that
were meaningless statistically), and we in the tract committees did have
the opportunity to recommend block group lines. I just don't think that
was done very often; it's hard enough to get people to do tracts. So, for
the most part, the 1980 accidental geography remained in place. STF3 was
expanded considerably so more data were available by block group.
I have no sense that there was anything different this time than in 1990,
except that you added that useless fourth digit and renumbered blocks
within BG and tract. So, I stand by my original statement for areas which
did not have good local intervention: block groups are accidental geography
and not statistical geography.
Re TAZs: yes, the purposes are very different. However, in the Detroit
area and I'm sure in lots of other places, TAZs tend to follow tract lines
in residential areas (where there is no high employment density). Of
course the tracts are split up in downtowns and other high attractor
areas. TAZs don't change very much either unless there are major land use
changes that change their status as residential or attractor areas.
As you know, I disagree with the fundamental point that the main objective
in tracts is comparability over time. The main objective is tracts that
work well now and into the future. Most of the time in most places, the
changes are small enough that most tracts can and should be maintained from
census to census, but comparability is not the main reason. The main
reason is that the characteristics that make a good tract haven't
changed. I know that you are quoting the official Census Bureau "line"
about these matters. I am trying to help people understand what's really
happening. Long form block group data are lousy.
Someday, we're going to have another geographic conference where we can
discuss these things. Meanwhile, you might want to look at a paper I wrote
for PAA a couple of years ago. If you don't have a copy, I'll be happy to
send it to you.
-----------------------
In response to the writers from Spokane, Missouri, and RI:
Those of you who have specifically delineated block groups are in that 5%
(OK, maybe 10%) in the nation. I agree that people working in small metros
and who are already concerned about their tracts are most likely to do the
work. The more, the better! I'm all for it. However, it's still true
that the vast majority of block groups are not specifically drawn as
statistical geography. (As Bob LaMacchia mentioned, the bgs in the 1990
BNAs were drawn in the regional offices with an eye toward meeting field
work needs; these needs are very different from those of data users!)
The important thing for everyone to remember, though, is the essential
shakiness of bg data. Your error rates are very high.
Yes, the Detroit metro area is more than 10 times the size of Spokane, so
it makes it easier to use tracts here. Congratulations to RI for
merging/nesting census and TAZ geography, I think that's the best way to do
it.
Patty Becker
-------------------------------------------------------
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