Activity-based modeling is becoming the main travel demand forecasting process in the largest metropolitan areas in the United States, with at least seven models completed and in use and another 10 currently under development. These models have been developed in a variety of ways. The earliest models, as well as many of those currently under development, have been estimated completely from local data sets and include many variations on the model structure. A few models currently under development use parameters transferred from other urban areas. While transferring parameters can provide significant savings in data collection and model estimation costs, the validity of transferring activity-based model parameters has not been rigorously studies, and the transferability of parameters has been asserted.

The Houston and Minneapolis-St. Paul metropolitan areas are currently developing activity-based models, estimating parameters from locally collected travel survey data. These models are very similar in structure, using the TourCast framework initially implemented in Houston. Because of the similarities in model structure, there is an opportunity to compare the same parameters between models in a systematic way. This comparison does not prove or disprove transferability of activity-based model parameters, even between Houston and the Twin Cities, but it can provide valuable insights to practitioners considering transferring model parameters.

The Houston model parameters have been estimated, and the Twin Cities model parameters are being estimated in the fall of 2012. This presentation will show the results of the systematic comparison of model parameters, presenting similarities and differences in the sensitivity of the modeled behavior to input variables such as travel time, cost, person and household characteristics, and land use characteristics. The types of models for which parameters will be compared will include vehicle availability, workplace location, destination choice, mode choice, and time of day choice. While the models for the two areas are not identical, statistical analysis of the similarity of parameters will be presented where possible.