Modelling shared e-scooters: A spatial regression approach

Abstract

Shared e-scooters have appeared quickly and in large quantities, yet little is known about their use. In this study, we explore spatial drivers of demand for shared e-scooter trips in Louisville (KY). We estimate a generalized linear mixed model with conditionally autoregressive random effects using 15 months of booking data, points of interests from Open Street Maps and US census data. We find that population density, the presence of bikeways and university campuses have the strongest positive effect on shared e-scooter trip destination counts. We find a significant, yet less substantial positive effect of bus stops suggesting some first/last mile use and hypothesize tourists to be an overlooked, yet important segment in shared e-scooter demand.