TO: CTPP-News:
FR: Chuck Purvis, MTC
RE: CTPP Part 3 Issues
On the one-year anniversary of receiving the Census 2000 CTPP Part 3
data, I thought I would re-release an old e-mail looking at the Part 3
data suppression and data loss issues (with some current observations).
There are 14 total tables in the CTPP Part 3 area-to-area "flow"data.
Five of these tables (3-3 through 3-7) have a minimum 3-worker sample
threshold applied. If there are less than 3 sample workers, total, on an
area-to-area basis, then values for that flow interchange are
suppressed. This can have a significant impact at finer geographic
summary levels, e.g., tract-to-tract, zone-to-zone, and block
group-to-block group. (This is not 'new' news, but is old, bad news.)
This particular data analysis is for the Part 3 data for intra-San
Francisco Bay Area. Other regions of various sizes will probably show
similar patterns. Attached is my "bottom line" spreadsheet that shows
the data loss due to suppression and rounding error across all
geographic summary levels.
Table 3-1 problems are limited solely to rounding errors. There are
insignificant rounding errors going from the county-to-county to the
place-to-place geographic summary levels, but modest rounding errors at
the tract-to-tract level (a loss of 1.5%) and an incredible rounding
loss of 3.5 percent at the zone-to-zone level. Note that rounding is
done on an interchange cell by interchange cell basis. There is no
"bucket rounding" applied in the CTPP to correct for this rounding
error.
Table 3-2 (household workers by means of transportation) also shows a
similar rounding error problem, topping out at a 4.2 percent loss due to
rounding at the zone-to-zone level.
Table 3-6, all workers by detailed means of transportation, shows minor
data loss due to rounding and suppression at the place-to-place level
(-1.2 percent), but major data losses at the tract-to-tract level (-34.4
percent), and zone-to-zone level (-71.5 percent). Yes, 71.5 percent of
the detailed means of transportation is lost at the zone-to-zone level.
The only usable information from Table 3-6, at the zone-level, is the
number of at home workers (Table 3-6, cell 18.) (NOTE: If you need block
group or tract level workers working AT HOME, USE SUMMARY FILE #3
DATA!!!)
The CTPP Part 3 does NOT have an "all workers universe" table on
workers by DETAILED means of transportation, WITHOUT the sample size
threshold. On the other hand, we can derive "all workers by means of
transportation" by dividing the "aggregate travel time by means" (Table
3-14) by the "average travel time by means" (Table 3-8). This "back
door" approach to estimating workers by means of transportation
(universe: all workers) works exceptionally well. There is basically a
negligible data loss across all geographic summary levels.
For the past year we have been working with this "derived" commuters,
zone-to-zone, based on dividing Table 3-14 by Table 3-8. This is
essential if you're examining commuter flows, average commute distances
(in miles or kilometers) and commute length frequency distributions.
One problem with the Table 3-14 / 3-8 commuters is the "Other" category
which combines five sub-groups: bicycle, walk, motorcycle, taxicab, and
"other means". What this really means is that we (and probably all
others in the U.S.) will be unable to create zone-to-zone bicycle or
walk commuter matrices. (This is unlike the 1990 Census, where we
created zone-to-zone bicycle and walk commuter matrices for purposes of
analyzing commute distances and commute distance frequency distributions
for these non-motorized commute modes....)
(Actually, the Part 3 data came too late (May 2004) to be used in our
travel demand model validation efforts, which we completed April 2004.
We used the county-to-county flow data, derived from the 5% PUMS, by
means of transportation by household income quartile, for validating our
work trip distribution and work trip mode choice models.)
That's all I have to offer for the moment. Think it over, and hope to
see folks at the Census / Transportation Conference in Irvine, CA this
week!
Chuck Purvis, MTC
**************************************************************
Charles L. Purvis, AICP
Principal Transportation Planner/Analyst
Metropolitan Transportation Commission
101 Eighth Street
Oakland, CA 94607-4700
(510) 464-7731 (office)
(510) 464-7848 (fax)
www:
http://www.mtc.ca.gov/
Census WWW:
http://www.bayareacensus.ca.gov/
**************************************************************