Optimal Pricing for Distance-Based Transit Fares

Richard Hoshino, Jeneva BeairstoAAAI 2018 · DOI

Abstract

Transit authorities frequently struggle to balance equity, ridership, and revenue when designing fare structures. Traditional zone-based systems often penalize short-distance travelers crossing arbitrary boundaries while subsidizing long-distance travelers within a single zone. This paper presents a mathematical framework for transitioning to distance-based pricing using quadratic programming. By applying the Cauchy-Schwarz Inequality, we derive a closed-form solution that allows for the optimization of fares to maximize either revenue or ridership. The model is validated using real-world data from Metro Vancouver’s TransLink network, proposing a five-tier distance-based system that maintains revenue while improving fare equity for the majority of users.

Introduction

Most urban transit systems rely on flat fares or zone-based pricing. While simple to administer, these systems are inherently inequitable. A passenger traveling a short distance across a zone boundary may pay significantly more than a passenger traveling a much longer distance within a single zone. Urban planners increasingly favor distance-based fares, where the cost is proportional to the distance traveled. However, implementing such a system requires balancing the competing goals of maximizing ridership and maintaining revenue. Setting distance-based fares too high risks losing riders, while setting them too low threatens the financial sustainability of the transit network.

Problem Formulation and Methods

The challenge of determining optimal distance-based fares is formulated as a constraint optimization problem. Specifically, the authors utilize Quadratic Programming (QP) to model the relationship between fare prices, ridership, and revenue.

Unlike traditional iterative methods such as the Frank-Wolfe algorithm, this research identifies a more efficient approach. By leveraging the Cauchy-Schwarz Inequality, the authors derived an exact, closed-form solution. This allows transit authorities to calculate the "perfect" price for various distance tiers ($X_i$) instantaneously. The framework provides two primary optimization paths:

  1. Revenue Maximization: Determining the fare structure that yields the highest revenue for a given ridership target.
  2. Ridership Maximization: Determining the fare structure that maximizes the number of passengers for a given revenue target.

Central to these calculations is the concept of price elasticity ($k$), which quantifies how ridership responds to price changes. In the North American context, an elasticity of 0.2 is typically assumed, meaning a 10% increase in fare results in a 2% decrease in ridership.

Data and Empirical Analysis

The model was validated using comprehensive data from Metro Vancouver’s TransLink SkyTrain network. The dataset included 30 days of Compass Card data from September 2016, representing approximately 13 million boardings across 46 stations. The analysis utilized matrices for trip counts, geodesic distances, and existing zone-based fare counts for all station pairs.

Distribution of trips by distance and zone count

The empirical analysis quantified the "unfairness" of the existing system. As shown in the data, a significant portion of ridership involves a mismatch between distance and cost. Specifically, 17% of passengers were found to be "overpaying" for short trips across zone boundaries, while 7% were "underpaying" for long trips within a single zone.

Results: A Proposed Distance-Based Structure

To ensure logical consistency—where longer trips do not inadvertently cost less than shorter ones—the authors grouped distances into practical tiers. This "bundling" approach ensures monotonicity in the fare structure.

Proposed 5-tier distance-based fare structure

The resulting proposal for Vancouver is a 5-tier system ranging from 0km to 22km. The recommended fares for these tiers were calculated at $1.50, $2.00, $2.25, $4.25, and $6.00. Implementation of this structure would result in a negligible revenue loss of only 0.25% while providing a fare reduction for over 70% of SkyTrain users.

Discussion and Limitations

While the mathematical model provides a robust starting point for policy, it has limitations. The assumption of constant price elasticity across all rider types may oversimplify real-world behavior; "choice" riders (who can walk or bike) likely have higher elasticity than "captive" riders. Furthermore, the mathematically optimal fare for the longest trips (exceeding $7.00) may be politically unfeasible, necessitating manual price caps.

Despite these constraints, the use of AI and quadratic programming offers a powerful tool for transit authorities. It enables the rapid evaluation of "what-if" scenarios, allowing policymakers to understand the exact trade-offs between revenue, ridership, and social equity.

Summary updated

BibTeX
@inproceedings{hoshino2018optimal,
  author = {Richard Hoshino and Jeneva Beairsto},
  title = {{Optimal Pricing for Distance-Based Transit Fares}},
  booktitle = {AAAI Conference on Artificial Intelligence},
  year = {2018},
  doi = {10.1609/aaai.v32i1.11413},
}
Figure