Model Users Meeting
August 27, 2010
In Association with
&
DW & ASSOCIATES
Purpose:
Collect data from a statistically valid sample
of transit riders in the 20-county
metropolitan Atlanta area
Identify transit markets and patterns
The results will be used to update the
region’s travel demand model
Goal:
Obtain completed surveys from 10% of the
transit boardings
Transit Systems Included in the Survey
MARTA
CCT
GRTA
GCT
CAT
HAT
CTran
Emory
Note: Emory routes were not officially part of the survey, but data
from Emory routes that was collected by MARTA was included in the
final database)
Survey was conducted as a collaborative effort
between ARC, MARTA, GRTA and GDOT
Conducted face to face interviews
Used tablet PCs to record interviews
On-going coordination with FTA
Sampling methodology
Data expansion
Bus
Collected daily ridership by stop
Stops along each route organized into segments
Goal: Obtain complete surveys with 10% of
riders in each segment
Prevented high volume stops from being
underrepresented
Rail / Bus
Stations and bus stops with >100 daily riders
not grouped
One bus per route was staffed with 2-3 people
between 6:30am-6:30pm
Responsibilities
1 Person completed Boarding/Alighting Counts
1-2 people administered surveys
Bus Routes: Participant randomly selected by
computer generated number between 1-4 based
on the number of people who boarded at each
location
High Volume Stop Locations: Every 3rd person
who arrived at the stop
8-12 staff assigned to each train station
between 6:30am – 6:30pm
Staff assigned to work different areas of the
platform and to cover different cars
Every 3rd person who approached a specific
point was asked to participate in the survey
Interviewers did not have to decide who to
interview. The procedures determined who
would be interviewed
Mail-back surveys distributed to riders who did
not have time to complete the survey
Surveys were precoded so the surveys could be
linked to the time and exact location where the
survey was given to a rider
COLLECT DATA
(From riders at 235 bus
routes or rail stations located throughout region)
PRE-PROCESS DATA
(Build db of transit trip
records. Identify those
records that comply with
completeness and stage-1
logic checks)
QA/QC REVIEW
(Add travel forecast model
variables to survey db.
Conduct visual inspection
and outliar tests. Tag
survey trips that are
perplexing in logistical
sense)
REFINEMENTS/
EXPANSION
(Fix data that was flagged
during QA/QC Review.
Update sample counts by
stratification parameters)
Performed at each step
High quality interviewers were used
Extensive training was conducted
More than 100 checks were included on
the tablet PCs to ensure data collection in
the field was complete and accurate
Phone follow-ups were completed with
more than 18,000 participants
All required fields were filled with valid data
Origin address and type of place
Transfers to get to the current route from origin
Mode of access to the transit system
Boarding address
Alighting address
Transfers to get to/from current route to destination
Mode of egress from the transit system
Destination address and type of place
Home address
Number of autos available in household
Household size
Number of adults in household
Number of workers in household
Respondent’s employment status
Respondent’s student status
Driver’s License status
Age of respondent
Annual Household Income
Performed a series of logic checks
Number of household occupants >= number of employed members of the
household and the number of adults
Number of household occupants >= adults in the household
Number of workers >= than the household size
Number of household vehicles was consistent with the household income and
number of workers
Place names and street addresses properly
and consistently spelled
Reviewed distribution of the results by
individual interviewers to identify any of
their potential biases
Compared Trip Patterns to Transit Networks
Two Step Process
Used Database and GIS Mapping
Techniques
Profile Reports and Maps
Reviewed Trips in Relation to Transit Route
Trip Origin and Destination
Access and Egress Mode
Trip Leg Detail
Frequencies
Access type
Walk access distance
Trip purpose
Trip distance (straight-line)
Number of transfers
Household size
Number of workers in the household
Household income and autos available
Computed a Trip Ratio Test
Based on the ratio b/t survey trip’s path
distance to the straight-line origindestination distance
Distance from origin point to boarding location
Distance b/t the boarding and alighting locations
Distance from the alighting point to the destination
A ratio above 2.5 indicated there could
be a trip logistic issue
Inspect spatial sensibility and key variables
Trips with short distances (short for local bus and rail
was < 1.0 mile; for express bus it was 3.0 to 5.0 miles
depending on transit provider)
Walk and Drive access trips with zero transfers
Walk and Drive access trips with three or more
transfers
Investigate any other suspicious O-D travel
based on trip distance and spatial
orientation
Loaded trips onto transit networks
Converted trips into P & A Format
Review assignment
Market segment
Mode of access
Transfer
Transit mode (local bus versus rail)
Tagged all survey trips with a quality
designation
‘1’ complete
‘2’ inconsistent but easily fixed in our opinion
‘3’ inconsistent and probably difficult to fix
Tagged all survey trips with ‘2’ or ‘3’ with a
diagnostic note
Records were revised and reviewed again
using the same process
Separate Methodology for Bus
and Rail
Performed by Time of Day
Iterative Process
Coordinated with FTA
Step 1: Review the actual distribution of
completed surveys
Step 2: Review the actual ridership
between each rail station by time of day
Step 3: Adjust the distribution of actual
ridership upward to account for
"unknown" trips that were not clearly
coded to both a station of entry and exit
by time of day
Step 4: Calculate the actual gap b/t the number
of surveys completed and the numerical goal
set for each cell in the sampling matrix
Step 5: Calculate unlinked trip expansion
factors (Only 15 cells did not meet the original
sampling goals that were set for the project)
Step 6: Dummy records were created to
simulate trips that were not captured in the
survey involving very small numbers of riders
traveling between certain stations
Step 1: Review the actual distribution of
completed surveys for various stops/segments
along each route by time of day
Step 2: Review the distribution of completed
surveys as a percentage of total boardings for
each of the segments/major stops along the
route by time of day
Step 3: Review the actual boardings (ONs) and
actual alightings (OFFs) for each
segment/major stop along the route
Step 4: Calculate the actual boardings
(ONs) and actual alightings (OFF)
by segment/major stop
Steps 5 and 6: Apply an iterative process
to estimated number of ONs and OFFs
from between major stops/route
segment along each route
Step 6: Calculate the weighting factors
for unlinked trips
Basic Formula = 1/(1+number of transfers)
We may further refine process over time
Finalize Data Expansion and
Database
Finalize Linked Trips
Assign trips to networks and
evaluate transit skimming, pathing
and assignment procedures
HomeBased
Home-Shopping
Based 7.1%
School
Home- 7.7%
Based
College
8.4%
HomeBased
Medical
4.1%
HomeBased
Airport
1.4%
Non-Home
Based
Airport
0.9%
Non-Home
Based
10.3%
HomeBased
Other
15.3%
HomeBased Work
44.7%
3 vehicle
7.9%
2 vehicle
19.4%
1 vehicle
32.0%
4 or more
vehicles
0.0%
0 vehicle
40.7%
18.0%
16.0%
14.0%
12.0%
10.0%
8.0%
6.0%
4.0%
2.0%
0.0%
Male
Female
Home
Work
Airport
College/University
Other
School/Daycare
Store/Retail
Restaurant
Another Home
Hotel
Recreation
Bank/Other Office
Medical
Place of Worship
0.0%
10.0%
>$75,000
20.0%
30.0%
< $20,000
40.0%
50.0%
Mode of Access to Transit
Walked
Dropped Off
Drove Alone
Rode in a vehicle for part of
the trip and walked/biked rest
of the way
Carpooled/Vanpooled
Bicycle
Percent
72.4%
14.0%
10.6%
1.8%
0.9%
0.3%