Formulation and estimation of combined network equilibrium models with applications to Stockholm

Combined network equilibrium models of trip distribution, modal split, and route choice are formulated and estimated for the Stockholm region. These models take into account the feedback effects among three stages of the classical four-step model. A simultaneous structure of the mode and destination choices is studied together with nested combined models reflecting conditional choice probabilities. In the traditional combined model, mode choice is assumed conditional on destination choice. The reverse combined model, in contrast, considers destination choice conditional on mode choice. User-optimal traffic assignment is integrated with these approaches to modeling travel demand in network equilibrium models. The three combined models are estimated using the full-information maximum likelihood technique implying that all preference parameters are estimated simultaneously. The models are estimated on medium-size travel and network data for the Stockholm region in 1986. The results suggest a rejection of the traditional nested model because of incorrect relative values of the estimated cost-sensitivity parameters. The reverse nested model is preferred, even if the overall goodness-of-fit is better for the traditional nested model. An application to a future scenario for the year 2020 is included.