aggregation of results is justified by assuming that TDM programs generate modest modal changes providing an acceptable trade-off between accuracy and ease of use.
The pivot logit equation approach simplifies the estimation process and drastically reduces data requirements, making the model available to a broader, less technically oriented, audience of planners and employers. Coefficients derived from regional or area specific travel demand models are used as inputs and applied to the pivot logit equation to estimate changes in baseline mode shares spurred from specific TDM strategies. The coefficients are assumed to be derived using sound statistical methods to guarantee statistical robustness.
Among the constraints of such an approach are:
Trips and VMT estimates are strongly dependent on pre-specified parameters;
No guarantee that the pivot logit equation will predict actual mode shift (predicted mode shift will lie on the logit curve);
The logit equation is based on discrete, mutually exclusive choices (auto vs. transit, without admitting concurrent choices of transit and, say, walking);
Coefficients are affected by factors such as the variables included in the model (and the interactions between the variables), calibration procedures, and the quality of the underlying data; and,
There is no distinction between short run vs. long run effects.
The default parameters are obtained from traditional four-step travel demand forecast models. These models are usually estimated and calibrated for specific regions and uses, with little potential for a generalized use, transferability across different regional areas, and predictive power. This is more evident when trying to estimate the impact of TDM strategies in areas where regional transport demand models are not available.1 The trade- off of using pivot-point modeling versus more intensive computational methods, like four-step travel demand forecasting model is justified by assuming that for modest change in mode shares, such as those generated by TDM strategies, the incremental extrapolation is fairly accurate.
The COMMUTER procedure manual suggests that there exist other ways of applying the incremental or pivot-point modeling approach, for example by applying elasticity parameters from empirical work to extrapolate changes in base values. The use of elasticities was not considered as it was argued that they are “limited in being able to take into account the interactive effects that occur when multiple actions are applied or
This assertion, though,
of the pivot-point logit approach.
seems to contradict the logit equation is based on
a multinomial that different,
discrete choice model, which by its own definition estimates the likelihood mutually exclusive choices are simultaneously taken by an individual. The
1 The user manual that accompanies the Commuter model reports that travel and emission impact estimates are “highly sensitive to the values of these coefficients, especially cost coefficients.” The user is warned against creating hybrid equations or altering the default parameters in the absence of detailed local data from travel forecasting models.