AN APPLICATION OF DISAGGREGATE MODE CHOICE MODELS TO SYSTEMS-LEVEL TRAVEL DEMAND FORECASTING

This paper describes the development and application of disaggregate models to systems-level transit forecasting in the Niagara Frontier Region (Buffalo, New York). In this study, the disaggregate mode choice models are developed from trip information directly obtainable from standard home interview surveys conducted by most State or local transportation planning agencies. The data set used is a 12% sub-sample of the 1962 home interview in the Niagara Frontier. Binary logit models are calibrated for four kinds of trips, classified by trip purpose and auto availability. For each model, the individual's probability of using transit is related to two system variables: the transit-to-auto travel time ratio, and trip length. Aggregated transit usage proportions for the entire system are then obtained by combining the disaggregate probabilities with an approximation term which reflects the within-zone variance of system characteristics. The prediction performance of the disaggregate modeling approach is compared to that of the conventional modal split model. Results suggest that the disaggregate mode choice method can produce at least as accurate modal split predictions as the conventional method, by using the same type of information, but only 12% of the sample. The report also contains a description of the procedures necessary to integrate the disaggregate mode choice method into typical urban transportation planning modeling systems. /Author/