ODOT, Greg Giaimo and Rebekah Anderson

1
The Greater Cincinnati Area Large-Scale (100)
GPS-Based Household Travel Survey

• ODOT, Greg Giaimo and Rebekah Anderson
• OKI, Andrew Rohne
• Laurie Wargelin, Abt SRBI, Prime
• Peter Stopher, PlanTrans, GPS
• Kevin Tierney, Cambridge Systematics, Sampling
• Sharon OConnor, Resource Systems Group, Internet

2
Introduction

• Quoting from the RFP Research Objectives . .
• This is the first large scale GPS-based survey
  conducted in the United States, and therefore,
  beyond the various logistical issues, it is
  uncertain to what extent a GPS-based survey is
  able to capture all the information available in
  a diary-based survey.
• This presentation describes the processes used
  for the Spring 2009 GPS survey pilot, and
  discusses early findings.

3
Goals . . .

• Reduce substantial respondent burden inherent in
  traditional travel diary recordings
• Reduce under-reporting of trip data
• Increase representative response rates
• Provide the detailed geographic information on
  route, speed, and location not captured by
  traditional diary methods that can influence the
  way travel is modeled

4
Sample Design

• Sample Size Record three days travel --
  Complete a minimum of one-day trip records for
  all members of 4,000 households.
• No diary recordings (if over age 12)--GPS
  RECORDING ONLY.
• Address-Based Sampling - so that cell-only
  households are included.
• Exclusion of these households is an increasing
  problem with traditional RDD sampling.
• Estimates of cell-only households are 15-30
  depending on metropolitan area.

5
Phone Matched and Un-Matched Sample with
Address-Based Sampling

• The randomly selected address-based sample is
  from US Postal Service delivery sequencing files.
• Sample addresses are matched with known land
  based phone numbers. (For the pilot, 60 of
  addresses were matched with a land based phone
  number-about the same as reverse matching for
  RDD sample.)
• Addresses without known phone numbers
  (Un-matched) consist of households with
  unlisted numbers, no phone, or increasingly of
  cell phone-only households (total 40).

6
Recruitment With an Address-Based Sample
7
Other Advantages of Address-Based Sampling

• Address-based sample allows oversampling of
  transit access areas and University off-campus
  areas (student households) by census block
  groupnot fully possible with RDD sampling.
• PUMS data (aggregated by block group) can be used
  for monitoring recruits by combinations of
  household size, of workers, of vehicles, and
  HH lifestyle cycles (student, with/without
  children, retired).
• Advance letters can be sent to all sample
  householdsnot possible with an RDD sample.

8
Survey Design- GPS Capabilities

• Personal GPS units so that all travel, not just
  vehicle trips, are recorded. Can be carried in a
  pocket or purse, or clipped on a belt or a wrist
  band.
• Goal of recording three days of travel.
• Every member of the household 13 years
• and older carries a GPS unit for
• three days.
• The GPS devices will be deployed over
• a one year time period beginning in
• July of 2009.

9
Survey Design Pilot Implementation

• Besides GPS units for all over 12, we collect
  limited childrens activities and travel
  information in a diary format (to link with other
  household members travel). Objective is to
  reduce burden and meet child privacy concerns.
• Six-minute phone or Internet recruitment
  interview--three-day travel periods assigned.
• Short-form household and person information forms
  distributed and collected with GPS units.
• Forms collect (1) work and school locations,
  (2) two most frequent household shopping
  locations, and (3) GPS usage status for each
  member, each day.

10
Short-Form Materials Piloted with Deployment GPS
Units
11
Survey Design Pilot Implementation

• Goal is to develop an efficient (low cost) GPS
  data collection process
• GPS unit and forms packages sent by Fed Ex
• GPS units return methods piloted
• (1) Participants provided with pre-paid
  shipping packages that can be deposited in either
  Fed Ex or US Postal Service drop boxes
• (2) Call the project 1-800 number to arrange a
  Fed Ex or personal courier pick-up.
• (3) Follow-up phone calls and Internet reminders
  to arrange courier pick-ups as necessary.

12
GPS Data Imputation and Verification PlanTrans
Processing Methodology

• Imputation of Trip Ends and Mode - Using a set of
  rules that include movement of the GPS for 2
  minutes or more-or lack of movement, or a
  significant change in speed, indicating a
  different mode being used.
• Prompted Recall (PR) Verification - Return of
  respondents travel (in Google Map form) in a
  web-based format for verification. Detailed
  ability to correct travel and purpose
  information, and add travel cost (fare, driving
  and parking costs) and vehicle occupancy for each
  trip.

13
Prompted Recall Web Format
14
Survey Design - GPS Data Processing and
Imputation

• PlanTrans imputes purpose using the frequency and
  duration of visits, the match to one of the
  collected addresses (home, workplace, school,
  frequent shops), and to available GIS land use
  data.
• PlanTrans is also developing an additional
  rule-based procedure for occupancy by family
  members by matching trips from different family
  members by time, location, and mode.
• With the aid of the prompted recall, Artificial
  Intelligence software is being trained, and these
  results will be applied to rule-based software.
  In this process, PlanTrans will attempt to add
  the capabilities to impute occupancy,
  driver/passenger status, and possibly parking
  costs and bus fares.

15
Pilot Sample Plan Designed to Test Response Rates
and Incentives

• Equal sample for Higher Transit Access and Lower
  Transit Access geographic areas
• Equal sample for Phone Matched and Non-Matched
  Sample
• Matched Sample offered 0 or 10 incentive to
  complete
• Non-Matched Sample offered 10 or 25 incentive
  to complete

16
Pilot Sample Results


17
Pilot Sample Results- Incentives
Matched Sample

Un-Matched Sample

18
Pilot Sample Results- by Income
Overall Completes to Recruits
Completes to Recruits by Incentive


19
Processing and Verification of GPS Data Files for
Pilot

• PlanTrans Processes the GPS Data Files Twice
• First for the Prompted Recall Survey
• After the Prompted Recall Survey
• Deletions or additions are made to fix trips
• Mode of travel is rechecked and identified for
  each trip
• Purpose of trip is rechecked and identified
• Trip File is Created
• Monthly or Bi-Monthly Completed Data is Delivered
  to the Client for Rechecking

20
Pilot Results - Address-Based Phone Match vs.
Non-Match

• Internet was the most viable means of obtaining
  recruits from households without land-based
  phones.
• Additionally, 19 of recruits from the phone
  matched sample responded to the advance letter by
  completing the recruitment on-line.
• Only one phone number was obtained from the
  unmatched sample via a return postcard/reply to a
  hot button issue survey.
• Regardless of recruitment method, completion
  rates for matched and unmatched sample were
  equivalent once recruited.

21
Demographics of Pilot Internet Responders

• The Advance Letter to the Un-Matched
  address-based sample (households without known
  land phones) was successful at recruiting a
  substantial percent of households to the
  GPS-Based Survey via the Internet.
• This was particularly true for younger age group
  households (18 to 34 years old) with only cell
  phones.
• These households are typically under-represented
  in traditional Household Travel Surveys.
• As would be expected, there were also a higher
  number of student households in this group.

22
Demographics of Pilot Internet Responders - Age

• These households are typically under-represented
  in
• Diary Household Travel Surveys and subsample GPS
  surveys
• also show their trips and tours to be
  under-reported.

23
Pilot Representativeness of Completed Households

• A very representative sample was recruited and
  completed by HH Size.
• The requirement that all household members age
  13 or older carry GPS units did not prove to be a
  respondent burden barrier.
• A representative sample was completed by number
  of vehicles.
• However, while an appropriate percent of
  zero vehicle households were recruited, extra
  effort (incentives?) will be needed to complete
  zero vehicle households.

24
Pilot Representativeness of Completed Households
25
Pilot Representativeness of Completed Households
Cont

• The completed pilot sample was fully
  representative by lifestyle/family type.
• The higher percent of households with zero
  workers was not due to oversampling of retirees.
  May be attributable to current economic
  conditions.
• The pilot was successful at recruiting low income
  households, but incentives/extra effort will be
  required to complete these households.

26
Pilot Representativeness of Completed Households
Cont
27
Pilot Representativeness of Completed Households
Cont
28
Primary Logistical Problems Return of GPS
Units and Some GPS Battery Outages

• Retrieving GPS units in a timely manner for
• redeployment with minimum loss - is a
• logistical and cost problem
• Loss rate for pilot was 2.7 percent--mostly
  among
• low income/urban households.
• More units needed, higher incentives, longer
  field
• time?
• Battery outages over three days need to
  supply
• chargers

29
ODOT GPS-Based Pilot Summary to Date

• Address-based sampling can be successful in
  recruiting cell-phone only households to a
  GPS-Based HTS via an Internet recruit.
• GPS household completion rates are adequate and
  representative.
• Requiring every household member (over 12) to
  carry a GPS unit for three days was not
  considered an undue burden paperwork was
  greatly reduced.

30
ODOT GPS-Based Pilot Summary to Date

• The child diary needs to be kept simple - perhaps
  only one day travel is needed.
• Significant incentives and additional efforts are
  needed to complete unmatched households, and
  households with low-incomes and/or zero vehicles.
• Added trip accuracy reporting and the value of
  route and location with speed data (as collected
  via GPS) needs to be demonstrated upon completion
  of the pilot PR and trip files in early June
  2009.