Petrella, M: Travel Behavior Panel Surveys

Travel Behavior Panel Surveys:
Measuring the Impacts of Road
Pricing in Seattle and Atlanta
Travel Survey Methods Committee Meeting
January 25, 2012
1
Project Background
• USDOT Urban Partnership Agreement
(UPA)/Congestion Reduction
Demonstration (CRD) programs fund
selected cities/ regions to implement a
comprehensive, integrated approach
to reducing congestion
– The four T’s: Tolling; Transit;
Technology; Telecommuting
– Recipients: Atlanta, Seattle, Miami,
Minneapolis, Los Angeles, San
Francisco
• National Evaluation being conducted
by Battelle
• FHWA funds Volpe Center to perform
household panel survey
2
2
Source: Wikipedia
Seattle and Atlanta
3
Evaluation Questions
How did travel times, vehicle miles traveled, and
daily travel budgets change at the
individual/household level?
Were there shifts in departure times or modes?
In origin-destination patterns?
For those who used the priced facility less, where did
the reduced trips go? Telework? Combined
errands? Route diversion?
Atlanta: How do 2-person carpools adapt to the new
occupancy requirements?
What are the equity impacts of the road pricing
policies?
4
Overall Study Approach
• Household Panel Study: same households before and after
road pricing
Seattle
wave 1
survey
Atlanta
Express
Lanes open
April/May
2012
Oct. 2011
Nov. 2010
5
Atlanta
Wave 2
survey
April/May
2011
Dec.
2011
April/May
2012
Atlanta
Wave 1
survey
Seattle
tolling
starts
Seattle
Wave 2
survey
Overview of Study Population
• Peak hour corridor drivers: sample through license plate
capture with match to registered address
• Seattle: SR-520 and I-90
• Atlanta: I-85 and Buford Highway
• Peak hour corridor transit users: In-person intercept
• Seattle: bus stops, transit centers; park & rides; on-board
buses
• Atlanta: park & rides in corridor: MARTA stations
• Corridor vanpool members: send email to vanpool
participants
6
Atlanta Sample Development (Drivers)
52, 901 plates captured:
Feb. 15 -16, 2010/ 6-10 AM & 3-7 PM
39,527 on I-85;13,374 on Buford Highway
Clean out unusable records, duplicates, commercial, out-of state
46,245 plates sent to State for matching
91.6% match rate: 42,379 addresses returned
Final quality check (e.g., remove businesses)
35,455 Survey invitations mailed out
765 undeliverables
34,690 addresses in final sample
7
Sample Development for Transit/Vanpools
• Seattle: in-person intercept October 18-21, 2010
– on-board intercept (ipad) on buses crossing Lake Washington
– Postcard handout: Redmond and Bellevue Transit Centers; South
Kirkland and Eastgate Park & Rides; on-board buses; downtown
bus stops
• Atlanta: in-person intercept March 21-25, 2011
– Postcard handout only: Discover Mills, Indian Trail, Mall of
Georgia, and I-985 Park and Rides, Doraville and Lindbergh
MARTA Stations
• Vanpool recruitment: Georgia Regional Transportation
Authority/King County sent an e-mail to registered
vanpoolers who use the corridor inviting them to
participate (~500)
8
Survey Materials
• Advance notification postcard
• Introductory letter
• FAQs
• Memory Jogger
• Reminder postcards and
emails
9
Online survey tool
• Household survey on demographics (completed by one person)
• 2-day travel diary completed by all adult (18+) household members
• Additional survey questions: typical commute behavior; typical use
of the facility; trip satisfaction; attitudes and values
• Phone option available; Spanish version of materials in Atlanta
10
Pilot Study
• Purpose: Test all steps of survey administration
– Recruitment method (license plate capture and transit
intercept; effectiveness of materials; incentive structure)
– Online tool: questionnaire design and functionality, survey
duration
– Obtain estimate of response rate
Modal
Segment
Seattle
Response Rate
Atlanta
Response Rate
Auto
9.6% (N=175)
8.9% (N=176)
Transit
18.3% (N=119)
14.7% (N=49)
294 households
225 households
NOTE: “Completion” defined as all adult members of the household
complete their travel diary
11
11
Pilot Study, continued
• Findings: Overall, no major issues or problems
– Cut survey questions due to comments on length
– Clarify several error messages/instructions
– Add response categories for some questions
• Trip purpose: “exercise/gym”
– Increase automation
• Pre-populate starting point for day 2 trip roster with ending point
from day 1
– Utilize $15 gift card incentive (resulted in 9.4% response)
• $10 gift card: 7.0% response rate
• $10 gift card with enclosed $1 bill: 9.8% response rate
• $15 gift card with enclosed $1 bill: 11.8% response rate
12
Incentives and Panel Maintenance
• Each household receives $15 gift card
after wave 1 completion; $30 after
wave 2 completion
• Approximately 3 contacts per household
– graphic display of findings (~3 months after wave 1)
– Seattle only: letter about wave 2 survey delay
– After pricing: mini-survey to engage respondents and obtain initial
feedback on tolling
– Household update survey (several weeks prior to wave 2 survey)
13
Wave 1 Response
Seattle (November 2010)
Household Completion Rates by Recruitment Mode
Modal
Segment
Wave 1: Total
invitations distributed
Wave 1
Completions
Auto
31,282
2908
9.3%
2,513
396
15.8%
520
52
37,983
3356
Transit
Vanpool
Total
Percent
completed
10%
Atlanta (April/May 2011)
Household Completion Rates by Recruitment Mode
Modal
Segment
Wave 1: Total
invitations distributed
Wave 1
Completions
Percent
completed
Auto
35,455
2090
5.8%
Transit
2,721
303
11.1%
Vanpool
477
19
3.9%
Total
38,653
2412
NOTES: “Completion” = All adult members of the household completed all surveys
14
14
Modal
Segment
Goal for
Wave 1
Goal for
Wave 2
Auto
2600
1300
Transit +
vanpool
400
200
Total
3000
1500
Wave 1 Methodology Notes
• Respondent Burden
– Rough estimates come from RSG data on respondents’ web-based
survey times
– This does not include time for filling in Memory Joggers or other work
– Averaged 4 minutes per household for initial screener
– Approx. 10 minutes per person per day for diary and related questions
– 44 minutes total for a typical 2-adult household
• Item Non-Response:
– Essentially none due to design of online survey
– ~10% selected “prefer not to answer” on income question
• Survey medium:
– At least 95% of respondents used online tool, but telephone option was
available and used
15
Methodology Notes, continued
• Non-response Bias:
– Analyzed via comparison of completed surveys against
partial completions and external benchmarks (including
other corridor studies, Census, and Acxiom data)
– Household size appears to be the only key variable with a
bias – the achieved sample has fewer large households
16
Lessons Learned
• Pilot your survey
• Spend the extra time needed to develop a high quality
online survey; provide clear, concise directions
– Pop-up windows, map of corridor
• Advance planning is critical
– License plate capture process differs by state
• Be flexible
– Added travel days in Atlanta to boost response rate
– Enhanced panel maintenance with “mini-survey” to engage respondents
17
Next Steps
• Focus groups in Seattle (February 2012)
• Panel maintenance “mini-survey” in Seattle and Atlanta
(January/February, 2012)
• Final version of wave 2 surveys
• Household update and wave 2 survey administration
18
Comments? Questions?
Margaret Petrella, Social Scientist
The Volpe Center
[email protected]
19