This paper applies discrete multivariate analysis to the specification and estimation of factors that govern work-trip mode choice. Where large data sets are available, this technique is found to have two important advantages over conditional logit analysis: Better model specification is facilitated and parameters can typically be estimated at sharply lower cost. The study focuses on the mode-choice behavior of 9880 Washington, D.C., area households that made work trips in 1968. Perhaps the most striking result is that in-vehicle travel time seems to have a nonlinear impact on the mode-choice logit (log-odds of drive alone versus bus), which has potentially important consequences for policy. For the range in which bus is faster than automobile, changes in bus (or automobile) in-vehicle travel time have the well-known results reported by other studies. But for the interval within which driving is faster than bus, decreases in bus in-vehicle travel time that fall short of making the bus mode absolutely faster than driving will, if our estimates are correct, fail to increase ridership significantly. (Author)

Media Info

  • Features: Figures; References; Tables;
  • Pagination: p. 30-35
  • Monograph Title: Passenger travel forecasting
  • Serial:

Subject/Index Terms

Filing Info

  • Accession Number: 00310699
  • Record Type: Publication
  • ISBN: 0309029813
  • Files: TRIS, TRB
  • Created Date: May 21 1981 12:00AM