Temporal Stability of Modal Choice Models

A number of methods for evaluating the temporal stability of travel demand models are outlined. Several of the methods are applied to disaggregate modal choice models for CBD work trips in Detroit for 1953 and 1965. Graphical analysis is used initially to evaluate the probability of using the bus, auto driver, and auto passenger modes as a function of travel time difference and autos owned. Based on the graphical analysis, binary choice models using linear regression are developed and tested statistically for temporal stability. Chow's tests for the equality of two sets of regression coefficients confirms the graphically observed lack of temporal stability. Improved specification of the auto driver model by including the variable, number of transit captives, does result in temporal stability. The lack of excess time and cost variables limit the generality of conclusions based on the models developed. Compatible data for a third time period are necessary to test fully the temporal stability assumption.