Taste variation in environmental features of bicycle routes

Abstract

In this paper we look at route choice modeling based on observational GPS traces collected by bicyclists in Amsterdam and surroundings. We consider factors influencing bicycle route choice such as distance and environmental factors such as cycle-way infrastructure, land-use environment, tree cover and the effect of noise emitting roads using data from a noise emission model. We estimate a route choice model, comparing multinomial logit, mixed logit and mixed path size logit specifications. Our results show that cyclists have a highly stochastic behavior that are likely to prefer detours to drive over cycle-way infrastructure, near greener landuse and near water, and on less busy roads. Models such as mixed logit that can estimate the stochasticity of cyclists perform best to capture this behavior.