Transferability and updating of disaggregate travel demand models
Three of the most highly regarded disaggregate mode split models incorporate very different estimates of the responsiveness, or elasticity, of mode choice to changes in auto travel times and costs.These differences appear to be due in part to the varying specifications used by the model, and particularly whether certain variables (such as a dummy variable for CBD destinations or automobile ownership) are included in addition to the more traditional variables (such as travel time, cost, and household income).In technical terms, this assumption is referred to as the "temporal transferability" of the models. Patruni (2009) Improving the Treatment of Cost in Large Scale Models, European Transport Conference, Noordwijkerhout, The Netherlands. Hess (2010) Review of Evidence for Temporal Transferability of Mode-Destination Models. Gunn, H., (2001) Spatial and Temporal Transferability of Relationships between Travel Demand, Trip Cost and Travel Time. This paper summarizes the findings from a literature review that demonstrates there is little evidence about the transferability of mode-destination models over typical forecasting horizons. Disaggregate travel choice models have been extensively developed in recent years (CRA, 1972; PMM, 1973; Ben-Akiva, 1973; Richards and Ben- Akiva, 1974; Lerman and Ben-Akiva, 1975; and others).
To provide further insights and evidence, models of commuter mode-destination choice been developed from household interview data collected across the Greater Toronto and Hamilton Area in 1986, 1996, 2001, and 2006. Pursula (1997) Transferability Analysis of Disaggregate Choice Models. Wilmot (1982) Transferability Analysis of Disaggregate Choice Models. (1981) A Comment on Interspatial, Intraspatial, and Temporal Transferability. The flow diagram in Figure 1 calls for the use of a prediction procedure which obtains the desired aggregate predictions based on predicted input data and the estimated disaggregate model.The objective of this paper is to identify prediction procedures for use with disaggregate models which (1) use commonly available data or additional data which can be obtained easily, (2) are computationally inexpensive, and (3) provide relatively accurate travel predictions.