Specification and estimation of mode choice model capturing similarity between mixed auto and transit alternatives

Abstract

To analyse the diversion from auto modes to combined modes such as park and ride, it is common to develop mode choice models based on discrete choice theory. In most cases, park and ride is modelled as an access mode to a main transit mode. This paper proposes an approach to test similarities among modes and the appropriate model structure, providing the flexibility for various model structures. The paper explores the capability of recently developed models by specifying their structure to capture the similarities of the combined modes. The paper presents an example with real data to illustrate the methodology application. Estimation results for different model structures including the Multinomial Logit, Nested Logit, Cross-Nested Logit and the Logit Kernel with all of these previous models as kernel are presented. As expected the best estimation results are obtained for the most flexible model, the Logit Kernel with Cross Nested as Kernel.