Investigating COVID’s Impact on SEPTA with Negative Binomial Distributions

Loading...
Thumbnail Image
Penn collection
The Wharton School::Wharton Undergraduate Research::Joseph Wharton Scholars
Degree type
Discipline
Statistics and Probability
Urban Studies and Planning
Subject
Transit
Negative Binomial Distributions
Ridership Models
SEPTA
Philadelphia
COVID-19 Pandemic
Funder
Grant number
Copyright date
2025
Distributor
Related resources
Author
Eichmann, Oscar
Contributor
Jensen, Shane
Abstract

This thesis investigates how the COVID-19 pandemic has reshaped public transit ridership in Philadelphia and evaluates the extent to which Negative Binomial Distribution models (NBDs) can be used to track and forecast ridership in its aftermath. It tests several approaches for modeling pandemic-related disruption—including binary, count, and tapering multiplier specifications—and assesses their performance across both route- and mode-level datasets. Results suggest that NBDs are well-suited to managing overdispersed ridership data and capturing hierarchical variation, though with limited predictive precision in post-pandemic years. Among the tested specifications, the binary COVID indicator most effectively captured the persistent nature of ridership decline, while the tapering multiplier rarely indicated meaningful recovery. Route-level models offered finer spatial resolution, whereas mode-level models produced more stable estimates of system-wide trends. Together, these findings demonstrate the value and limits of using NBDs to quantify long-term shifts in transit usage, offering a foundation for more nuanced forecasting.

Advisor
Date Range for Data Collection (Start Date)
Date Range for Data Collection (End Date)
Digital Object Identifier
Series name and number
Publication date
2025
Volume number
Issue number
Publisher
Publisher DOI
Journal Issue
Comments
Recommended citation
Collection