The geography of ridesharing: A case study on New York City

Abstract

Despite the popularity of ridesharing, there is limited empirical evidence on how ridesharing activities differ across regions with different levels of accessibility and the implication for consumers. In this paper, we study the market for rides across New York City neighborhoods. We construct a novel data set that contains massive API queries on route-specific estimates of pricing, wait time, and travel time of Uber, Lyft, and the public transit. After linking this data with actual trip records of taxis, Uber, and Lyft, we document a strong pattern that ridesharing has a larger market share relative to taxis in neighborhoods with lower accessibility, defined either in terms of geographic distance to Midtown Manhattan or “economic distance” to job opportunities. Next, we estimate a discrete-choice model of demand for rides and interpret the geography of ridesharing through the lens of the model. We find that consumer surplus from ridesharing varies drastically across geography: passengers that are 5 to 15 miles (resp. more than 15 miles) from Midtown experience a 60% (resp. 19%) larger consumer surplus relative to passengers that are within 5 miles from Midtown. Over half of these gains comes from reduced wait time.