The Time Stability of a Modal-Split Model for Tel-Aviv

Abstract

In the application of travel-demand models for forecasting, it is assumed that the model form and coefficients are stable in time. This study analyzes the time stability of a three-mode disaggregate modal-split model for trips to work over two points in time. Data were used from the 1972 Travel Habits Survey and from the 1976 Survey of Patterns of Travel to Work. All the trips used were made within the Tel-Aviv metropolitan area. The main method used in judging the transferability of the model was to apply the 1972 model to the 1976 trips thus giving for each trip the predicted probability of choosing the modes. The trips were split up into classes by various criteria, and for each criterion the probabilities of choosing a mode were summed over all trips in each class. This sum is the predicted number of trips in the class that uses the mode, and this was compared with the number actually observed. Another method for examining transferability is to recalibrate the model on the 1976 data, and to compare the coefficients obtained with the 1972 coefficients. It appears that the model calibrated on the 1972 data predicts the 1976 modal split satisfactorily in spite of the major structural changes which took place between 1972 and 1976 which resulted in a considerable change in the modal split itself. There is a problem with the prediction of use of the car-passenger mode, but this arises because of differences in survey techniques that were used for collecting the 1972 and the 1976 data. The model coefficients are stable in time except the coefficient for the number of vehicles. These results add confidence to the use of these models for forecasting.