CNSTAT October 19, 2004

Download Report

Transcript CNSTAT October 19, 2004

Emerging Issues for
Transportation Data Users of
ACS
Elaine Murakami, FHWA
[email protected]
TRB Planning Applications Conference, April 26, 2005
First, the good news…
• CB has added several transportation-related
tables to their planned set of ACS
tabulations.
• This includes
– Residence tabulations of households
such as: household size * # of vehicles
– A FEW tables by Place of Work
More positive things about ACS
• The ACS will provide reliable data for large
geographic units on an annual basis.
• The flip side is that the data for small areas
will be compromised, with much larger
standard errors, and the risk of the inability
to tabulate small area (TAZ) home-to-work
flows.
Questions for today’s discussion
• Do you want home-to-work flow (Part 3)
from ACS if the geography is limited to
tract-to-tract, or “tract group”- to- “tract
group”
• Should AASHTO to sponsor a pooled fund
for a CTPP from ACS?
AASHTO pooled fund
• Special tabs for large geography completed
annually, starting perhaps in 2007? 2008?
• Special tab for “small geography”,
conducted every 5 years, starting with the
accumulated records of 2005 thru 2009, OR
accumulated records of 2008-2012.
Transportation-sponsored research
on ACS
• 1995 meeting (1996 BTS publication) on
Continuous Measurement
• NCHRP project for Guidebook on using
ACS data.
• FHWA sponsored research at CB RDCs and
Suitland using ACS microdata.
• TRB Conference for May 2005.
Findings
• Seasonality does not seem to be a problem
for JTW variables (Hampden County only)
• Loss in # of O/D pairs is significant, and
may cause problems for modeling
applications (Westat, NCHRP)
• Tract level data by mode of transport seems
reasonable by residence geography (chart
next slide)
ACS-000200
Tract Number
CTPP-041400
ACS-041400
CTPP-039902
ACS-039902
CTPP-038600
ACS-038600
CTPP-037400
ACS-037400
CTPP-036100
ACS-036100
CTPP-034000
ACS-034000
CTPP-031800
ACS-031800
CTPP-028900
ACS-028900
CTPP-025100
ACS-025100
CTPP-022702
ACS-022702
CTPP-021502
ACS-021502
CTPP-020100
ACS-020100
CTPP-017500
ACS-017500
CTPP-015400
ACS-015400
CTPP-013200
ACS-013200
CTPP-011501
ACS-011501
CTPP-008500
ACS-008500
CTPP-006700
ACS-006700
CTPP-005200
ACS-005200
CTPP-000200
Number of Workers
Bronx County, Streetcar, trolley car, subway, or elevated
3,000
2,500
2,000
1,500
High
Estimate
Low
1,000
500
0
Key issues for Transportation
• Sample size / geographic level of reporting /
data quality / error
• Thresholds for FLOW tabulation
• CB Disclosure Review Board
• How to do Trend Analysis?
Sample Size
• Long Form 16.6% of addresses
• ACS (60 months) 12.5% of addresses
• HOWEVER, that is the sample FRAME,
not the number of completed returns.
Because the Mail-back response rate is
much lower for ACS, even with 1:3 NRFU,
the # of records in ACS is significantly
lower than LF.
Ratio of ACS/LF unweighted person
records
Broward*
Multnomah
Bronx*
ACS sampling
rate (3 yr total)
10%
15%
10%
Unweighted LF
person records
183,840
90,320
149,310
Unweighted
ACS person
records
Ratio ACS/LF
94,288
64,445
57,069
0.51
0.71
0.38
Implications of smaller number
of completed surveys
• Kiss TAZ data GOODBYE.
• Need a geographic unit maybe a “Tract
Group” or a TAZ that is as large as 2 or 3
tracts together, especially for home-to-work
flow data.
Thresholds required for
CTPP2000 for flow tabulation
• Thresholds are based on unweighted
records.
• The smaller number of completed records in
ACS results in a loss of O/D pairs meeting
the threshold.
Impact of Thresholds
Comparing Decennial LF to ACS Part 3
Broward County, FL Tract-to-Tract
Pt 3 WITHOUT
Thresholds
Pt 3 WITH Thresholds
# of Pairs
# of Pairs
Total
Workers
Part 1 total
workers
Total
Workers
LF
36,300
777,900
8,200
424,700
742,600
ACS
21,500
616,500
2,800
237,200
747,400
22% of O/D pairs
30% of workers
8% of O/D Pairs
Does 60 months of accumulation
result in reduced risk of
disclosure? We think “yes”
• People move residence
• People move workplace location
• People change means of transportation to
work.
• People change departure time.
• People change # of vehicles in hhld.
Therefore…
• The Census Bureau should not require as
much rounding or require thresholds for
tabulation!
What rules will the DRB impose?
What thresholds will the DRB impose for
flow tabulation?
– We hope NONE!
– If the threshold is 3 for key tables, as in 2000,
will transportation planners find it useful or
not?
Biggest threat
• The combination of smaller samples and
potential requirements of thresholds
(unweighted records) for tabulation by DRB
could result in a CTPP that is primarily used
for residence tabulations, with limited
workplace tabulations, and flow data
limited to total worker counts.
What do we want from the CB?
• Alternatives to rounding and thresholds for
disclosure avoidance.
• Improved allocation procedures. Want to
continue working with CB on allocation and
imputation of means of transportation, # of
vehicles, and place of work geocoding.
How to use TREND data?
Example: Tulare County, CA
Tulare County, 2000
2000 ACS
CA
Census
2003 ACS
Median hhld
income
$33,983
$31,467
$36,343
%hhlds with 0
vehicles
9.7%
8.1%
7.8%
%workers
carpooling
18.7%
14.0%
12.6%
How to use TREND data?
Example: No 2000 ACS data available
County with
2000
2000 ACS 2007 ACS
population
Census
below 250,000
Median hhld
$34,000
n/a
$38,000
income
%hhlds with 0
vehicles
10.0%
N/a
8.0%
%workers
carpooling
18.0%
N/a
12.0%
How should we augment
ACS data?
• Explore alternatives for Home-to-Work
Flow matrices, including LED
• Consider workplace surveys
• Add group quarters surveys, unless ACS
implements as planned.
• NHTS for 2008, for trip length distribution
curves for various trip purposes/activities
Questions for today’s discussion
• Do you want home-to-work flow (Part 3)
from ACS if the geography is limited to
tract-to-tract, or “tract group”- to- “tract
group”
• Should AASHTO to sponsor a pooled fund
for a CTPP from ACS?
• Please come to the TRB conference in May!
Thank you for your attention.