Generally speaking, the ACS estimates use the FSCPE's (Federal-State
Cooperation for Populatin EStimates) state and county estimates as
controls but not the exact figures. It is more complicate with detailed
adjustments without documented explanations. At least it happens to the
2004 ACS estimates to Colorado counties and municipalities.
- Richard Lin, Ph.D.
Colorado Division of Local Government
(303)866-4989, fax (303)866-2660
>> "Chuck Purvis"
<CPurvis(a)mtc.ca.gov> 9/2/2005 12:20 PM >>>
[I'm cross-posting my message to the national census-transportation
You raise important points:
1. How does the Census Bureau weight & expand the American Community
Survey? Do they weight it to match sub-county "control totals"?
2. How will formula allocation programs (HUD, HHS, DOT) utilize
American Community Survey in allocating funds to local jurisdictions?
How will formula allocation programs handle the higher standard
and confidence intervals, associated with the ACS?
1. The Census Bureau, to the best of my knowledge, uses counties as
finest geographic level for expanding and weighting ACS data. They do
not use sub-county or place-level control totals (e.g., like they did
with the weighting of decennial long form data to match decennial
form control totals). The "control totals" the Census Bureau uses are
the state-federal cooperative population estimates program, which
produces data on TOTAL population by age by sex by race/hispanic
(How the Bureau adjusts these county-level population estimates to
remove the group quarters population, I don't know.) Again, this is
to the best of my knowledge.
For example, our State Department of Finance reports the population
Alameda County, in 2004, as 1,497,000. If we assume the same 1.9% of
Alameda County's population of group quarters, then this means about
1,469,000 persons-in-households in Alameda County in 2004. The ACS
measures 1,427,827 household persons in Alameda County, a difference
4.7%. Why this difference? The pop estimates that the Bureau is using
are perhaps different (?) than the pop estimates that the California
released this past May.
To state that the weighting and expansion of American Community Survey
data is an "issue" is a gross understatement. It's a really, really
2. In terms of the question "how will allocation formulas be affected
by ACS?" This is really a very, very significant issue and honestly, I
For example, the president recently signed the federal transportation
reauthorization bill (SAFETEA-LU). There are two formula programs
on the distribution of DISABLED persons (for the "New Freedom
and the number of PERSONS BELOW THE 150% POVERTY LEVEL (for the Jobs
Access-for-Reverse Commuting - - JARC Program), based on urbanized
population. Right now, the only urbanized area data on disabled
population and poverty population is from the decennial census. So, to
allocate data NOW, the USDOT would need to rely on Census 2000 data.
When data is available from the 2005-2007 ACS (3-year accumulation of
data), down to the SMALL URBANIZED AREAS (50,000 to 65,000
the USDOT MAY CHOOSE TO USE THE ACS 3-YEAR DATA TO ALLOCATE JARC & New
Freedom funds....Hopefully the 3-year ACS data is less jumpy & jittery
than the one-year data we've seen over the past four years. (An
additional point to make is that the disability data from the
census isn't that good. The disability data from the ACS should be a
much better indicator of disability levels.)
The other example, probably more near and dear to city planner's
hearts, is the allocation of community development block grant (CDBG)
funds based on census-tract level estimates of poverty (?) population.
think the funding agency will have to at least look at using the
five-year accumulation of ACS data, but I'm concerned (as are many
others) about the "bright line" rules that regulate whether an area
or doesn't receive CDBG funds. [I am out of my element & comfort zone
when talking about CDBG fund allocations. If there's a HUD-Census user
forum that discusses these issues, that would be great information to
share. Here is a link to a 2002 vintage HUD report on the ACS. I
read it yet: http://www.huduser.org/publications/polleg/acs.html
I've heard comments from the Census Bureau folks that the ACS data is
best for "characteristics, not counts" but the Bureau will need to
realize that we data users want it all: both characteristics (shares,
rates & proportions) and counts (# of persons below poverty level, #
disabled, # of zero-vehicle households, # of transit commuters, etc.)
My strongest recommendation is to "cross-validate" the ACS data with
other, local administrative data. The best cross-validation, at a city
level, will be to compare the citywide total housing unit estimate
the ACS against the City's own records on total units....
This is a long-winded "I don't know" response.
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]
Census WWW: http://www.bayareacensus.ca.gov/
>> JPLevin(a)oaklandnet.com 09/01/05 6:21 PM
This new ACS data raises a lot of questions for us.
It shows our 2004 population as 365,266 with lower/upper bounds of
Our 2000 population was over 399,000. Even subtracting out the 7,000
households in group quarters (who aren't counted in the ACS), this
give us 392,000 in 2000, but only 365,266 in 2004, despite a
increase in our housing stock and by all indications an increase in
Department of Finance estimate is 411,319 for 2004 and 412,318.
Does anyone else have similar problems?
This kind of undercounting can cost cities a LOT of money.
Jeffrey P. Levin (jplevin(a)oaklandnet.com)
Housing Policy & Programs Coordinator
City of Oakland/Community & Economic Development Agency
250 Frank Ogawa Plaza, 5th Floor
Oakland, CA 94612
510/238-3501 FAX: 510/238-3691
ctpp-news mailing list