Time, Comfort, and Preferences for Shared Transit

Abstract

We evaluate both a natural field experiment and quasi-experimental variation in commute times for a shared transportation service provided by a large technology company during morning and afternoon commutes. We find that reduced commute times of roughly 2.5 minutes through expanded high occupancy vehicle (HOV) lanes and brand new coaches both significantly increased ridership by roughly 5.5%. We leverage these insights to describe some insights for public transit designers. We also highlight the importance of correctly specifying standard errors in randomized control trials (RCTs) with a small number of clusters: p-values increase by multiple orders of magnitudes when appropriately bootstrapping standard errors.