A stochastic user-operator assignment game for microtransit service evaluation: A case study of Kussbus in Luxembourg

Tai-Yu Ma, Joseph Y. J. Chow, Sylvain Klein, Ziyi MaarXiv 2020 · arXiv

Introduction

Microtransit services—flexible, on-demand shuttles designed to bridge the gap between private vehicles and traditional fixed-route public transit—have seen a surge in popularity. However, the economic viability of these services remains a significant hurdle. High-profile failures in the industry underscore a critical challenge: operators often struggle to balance ticket pricing, ridership volume, and high operating costs, frequently resulting in a heavy reliance on government subsidies.

This paper addresses the need for robust evaluation tools by introducing a stochastic stable matching model. By treating user-route assignments as a cooperative game and accounting for the inherent randomness in human decision-making, the framework provides a method for operators to optimize pricing and forecast ridership more accurately.

Methodology: Stochastic Stable Matching

The authors model the microtransit ecosystem as a two-stage assignment game between "buyers" (users) and "sellers" (operators):

  1. User-Route Assignment: This stage matches users to specific shuttle routes to maximize the total generalized payoff, which is the sum of user utility and operator profit, while respecting vehicle capacity constraints.
  2. User-Operator Stable Cost Sharing: The core innovation is the extension of deterministic cost-sharing models into a stochastic realm. By treating a user's travel utility as a random variable (comprising a deterministic base value and a normally distributed error term), the model captures the "noise" of real-world behavior.

To ensure computational efficiency, the authors utilize the concept of α-stability via chance constraints. This mathematical approach converts probabilistic stability requirements into deterministic linear inequalities, allowing the complex game to be solved rapidly using standard linear programming algorithms.

Case Study: The Kussbus Service

The framework was applied to the Kussbus microtransit service, which operated between Luxembourg and its border regions in Belgium and France.

Geographic distribution of Kussbus routes and stops between Arlon and Luxembourg.

Datasets

  • Kussbus Ride Data: The study analyzed 3,010 real-world trips from April to October 2018, including booking times, pickup/drop-off locations, and fares.
  • Mobility Survey: Data from 370 commuters in Luxembourg was used to estimate the local Value of In-Vehicle Time (VOT) at €24.21/hour, ensuring the utility functions were grounded in local economic reality.

Key Results and Insights

The Failure of Flat Fares

The analysis revealed that Kussbus's actual tariff (which included several free rides followed by a flat €5 fee) was financially unsustainable. Under this baseline, the operator incurred a massive loss of -€4,135 across the test dataset of 465 matched users.

Optimizing for Sustainability

By applying game-theoretic pricing, the model identified a "buyer-optimal" stable ticket price (median €10.98). While this adjustment caused a slight drop in ridership (to 426 users), it successfully swung the net profit to a positive +€187. Conversely, "seller-optimal" pricing, which attempted to maximize profit per ticket (median €49.91), caused ridership to collapse to just 6 users, demonstrating that aggressive pricing destroys the stable market match.

Comparison of ticket price distributions under different pricing strategies, including the current Kussbus tariff and the proposed optimal models.

Fleet Composition and Predictive Accuracy

  • Mixed Fleets: The study found that a mixed fleet (combining 7-, 16-, and 19-seaters) is more economically efficient than a homogeneous fleet of large vehicles. Larger vehicles carry higher operating costs that force ticket prices up, ultimately suppressing ridership.
  • Model Validation: By embracing randomness, the stochastic model achieved up to 79% accuracy in predicting user-route matches on training data, proving its utility for real-world transit planning.

Conclusion

The research demonstrates that microtransit can transition from a loss-making, subsidized model to a financially sustainable one through strategic pricing and optimized fleet management. While the current study focused on pre-scheduled routes, future work aims to extend this stochastic matching framework to fully dynamic, real-time shared mobility platforms and broader Mobility-as-a-Service (MaaS) ecosystems.

Summary updated

BibTeX
@misc{ma2020stochastic,
  author = {Tai-Yu Ma and Joseph Y. J. Chow and Sylvain Klein and Ziyi Ma},
  title = {{A stochastic user-operator assignment game for microtransit service evaluation: A case study of Kussbus in Luxembourg}},
  year = {2020},
  eprint = {2005.03465},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2005.03465},
}
Figure