Forecasting paratransit ridership using discrete choice models with explicit consideration of availability

Julian Benjamin*, Shinya Kurauchi, Takayuki Morikawa, Amalia Polydoropoulou, Kuniaki Sasaki, Moshe Ben-Akiva

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

16 Citations (Scopus)


In most developed countries, the population of the elderly and disabled is growing rapidly. These individuals require transportation service suited to their needs. Such service may be provided by applying emerging technologies to dial-a-ride transit. This research develops a methodology to quantitatively evaluate the impact of paratransit services on a traveler's mode choice behavior. The mode choice model explicitly considers availability of alternative modes and includes latent factors to account for taste heterogeneity. Stated preferences are also used to elicit preferences for new paratransit services. The methodology is empirically tested with data collected in Winston-Salem, North Carolina. The model system developed is applied to evaluate the effect of improving service attributes and the impact of the introduction of new cost-effective modes on modal shares. Results of the policy analysis indicate that (a) transit policy changes, such as fare reduction, would have little effect on automobile driver and automobile passenger shares; (b) an improved reservation system for dial-a-ride services would produce shifts in mode share; (c) the proposed new bus deviation service was favored; (d) free bus service reduces dial-a-ride share; and (e) an increase in awareness of a dial-a-ride system would significantly increase its share.

Original languageEnglish
Pages (from-to)60-65
Number of pages6
JournalTransportation Research Record
Issue number1618
Publication statusPublished - 1998
Externally publishedYes

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • Mechanical Engineering


Dive into the research topics of 'Forecasting paratransit ridership using discrete choice models with explicit consideration of availability'. Together they form a unique fingerprint.

Cite this