A nested nonparametric logit model for microtransit revenue management supplemented with citywide synthetic data

As an IT-enabled multi-passenger mobility service, microtransit can improve accessibility, reduce congestion, and enhance flexibility. However, its heterogeneous impacts across travelers necessitate better tools for microtransit forecasting and revenue management, especially when actual usage data are limited. We propose a nested nonparametric model for joint travel mode and ride pass subscription choice, estimated using marginal subscription data and synthetic populations. The model improves microtransit choice modeling by (1) leveraging citywide synthetic data for greater spatiotemporal granularity, (2) employing an agent-based estimation approach to capture heterogeneous user preferences, and (3) integrating mode choice parameters into subscription choice modeling. We apply our methodology to a case study in Arlington, TX, using synthetic data from Replica Inc. and microtransit data from Via. Our model accurately predicts the number of subscribers in the upper branch and achieves a high McFadden R2 in the lower branch (0.603 for weekday trips and 0.576 for weekend trips), while also retrieving interpretable elasticities and consumer surplus. We further integrate the model into a simulation-based framework for microtransit revenue management. For the ride pass pricing policy, our simulation results show that reducing the price of the weekly pass ($25 -> $18.9) and monthly pass ($80 -> $71.5) would surprisingly increase total revenue by $127 per day. For the subsidy policy, our simulation results show that a 100% fare discount would reduce 61 car trips to AT&T Stadium for a game event, and increase 82 microtransit trips to Medical City Arlington, but require subsidies of $533 per event and $483 per day, respectively.