Axis of travel: Modeling non-work destination choice with GPS data

Arthur Huang, David Levinson

Research output: Contribution to journalArticlepeer-review

33 Scopus citations


The advent of GIS and GPS has revolutionized how we monitor and model destination choice. New methodologies in building choice sets can be developed and new insights on travel behavior can be gained given the real-time GPS travel data. To this end, this research investigates how land use, road network structure, and axis of travel influence home-based, non-work destination choice based on in-vehicle GPS travel data in the Minneapolis-St. Paul Metropolitan Area in 2008. We propose a novel choice set formation approach combining survival analysis and random selection and a new approach to deciding choice set size. Mixed-effects logit models are used to model our data with repeated observations for each participant. Our findings identify the following factors that influence non-work destination choice: (1) Walkable opportunities and diversity of services at destinations, (2) Route-specific factors such as turn index and speed discontinuity, and (3) Axis of travel measured by relative travel time to work, home, and downtown. A destination closer to the axis of home and work, all else equal, is more likely to be selected. A destination far away from downtown is more attractive to auto users. This research contributes to methodologies in building choice sets for modeling non-work destination choice. The results enhance our understanding of non-work destination choice and have implications for transportation and land use planning.

Original languageEnglish (US)
Pages (from-to)208-223
Number of pages16
JournalTransportation Research Part C: Emerging Technologies
StatePublished - Sep 1 2015

Bibliographical note

Publisher Copyright:
© 2015 .


  • Axis of travel
  • Destination choice
  • GPS data
  • Land use
  • Non-work trips


Dive into the research topics of 'Axis of travel: Modeling non-work destination choice with GPS data'. Together they form a unique fingerprint.

Cite this