Parameters and predicted probabilities of mode choice and ride pass subscription for microtransit in Arlington, TX

We provide parameters and predicted probabilities of mode choice (at trip level) and ride pass subscription (at individual level) for microtransit in Arlington, TX. The parameters are estimated by an agent-based, nested behavioral model that is developed in the C2SMARTER project “Multi-modal Tripchain Planner for Disadvantaged Travelers to Incentivize Transit Usage” (Award #69A3551747124). We separate the choice related to microtransit into two parts: travel mode choice and ride pass subscription choice. Synthetic population data from Replica Inc. and microtransit service data from City of Arlington are used for estimation. 

In the lower-branch travel mode choice, individuals decide on mode to use by considering factors such as travel time, cost, trip purpose, tour type, and mode-specific preferences. The mode choice set includes driving, walking, biking, carpool, and microtransit. Different from traditional mode choice models, we allow time and cost parameters to vary across individuals, and we allow mode specific constants to vary across trip OD pairs. This makes sense when we do not have sufficient data for socioeconomic attributes and built environment variables. The assumption we made here is that the impacts of these unobserved variables are included in the nonparametric distribution of individual and OD pair-level parameters.

In the upper-branch ride pass subscription choice, individuals decide whether to purchase a weekly ride pass, a monthly ride pass, or no ride pass at all. By subscribing to a ride pass, travelers pay an amount of money in advance and enjoy free microtransit trips until the ride pass expires. The utility of purchasing a ride pass consists of four components: (1) the utility related to the prices of ride passes, (2) the utility related to the change in consumer surplus (or compensating variation) brought by free microtransit trips with a ride pass, (3) the utility specific to microtransit users, and (4) the alternative specific constant of the ride pass. Given the data availability, we consider the ride pass model as a simple MNL model with six parameters to calibrate. These ride pass parameters are calibrated using the Nelder-Mead Simplex Method. The cost function to minimize is the squared distance between the predicted ride pass market share and the observed one.

Accordingly, this dataset consists of six .csv files:

• Mode_choice_parameters_weekday.csv
• Mode_choice_parameters_weekend.csv
• RidePass_subscription_parameters.csv
• Mode_choice_probability_weekday.csv
• Mode_choice_probability_weekend.csv
• Ridepass_subscription_probability.csv

Field definition in "Mode_choice_parameters_weekday.csv" and "Mode_choice_parameters_weekend.csv"

Field definition in "RidePass_subscription_parameters.csv"

Field definition in "Mode_choice_probability_weekday.csv" and "Mode_choice_probability_weekend.csv"

Field definition in "Ridepass_subscription_probability.csv"