Over the past decade, destination choice models have become more common. This is especially true of travel demand models for larger and wealthier Metropolitan Planning Organizations (MPOs), particularly as these MPOs adopt more complex Activity Based Models (ABM) in which discreet choice models play a more significant role than in older more traditional four-step models. Despite the increase in the spread of destination choice models, many small and mid-sized MPOs are reluctant to migrate from traditional gravity models to destination choice models. Concerns pertaining to the data burden to estimate an accurate and effective destination choice model often appear imposing to such MPOs. Large sample sizes are considered normative with some such models boasting of being able to rely on sample sizes of up to 20,000 trip records. MPOs unable to fund such large data collection efforts feel inclined to abandon aspirations to developing destination choice models for their regions. Given that comfort with destination choice models can be a stepping stone to more advanced modeling techniques, such as ABM, this is unfortunate.

This paper will show that destination choice models can be estimated using relatively small sample sizes, especially with the availability of data collected from cellular phones to act as a validation check of the resulting model’s trip distribution. This paper will describe the development of a destination choice model for Wilmington, NC using 202 home based trip records from the 2009 National Household Travel Survey. The method hinges on developing a single model based on generalized cost for all home based purposes while developing additional interacting variables specific to individual home based purposes. This results in what is fundamentally a single home based purpose with home based work, shop, and other trips being special cases of this home based purpose. The resulting trip lengths and distributions were measured against trip tables developed by a data vendor from cellular phone positional data. The end result demonstrates the viability of developing a low cost destination choice model from a small sample size.