Modeling safety as a perceptual latent variable to assess cycling infrastructure

Perceived safety is a relevant construct that affects cyclists' behavior. Nevertheless, this qualitative attribute has not been adequately considered in infrastructure design preference models. Even though the use of perceptual latent variable may be suitable for this, they have not been used, likely because of their high data collection and estimation costs. We propose a change to the classic Integrated Choice and Latent Variables (ICLV) framework that reduces these costs. This methodology is used with data from a stated preferences survey carried out in Santiago (Chile), which collected choice and psychometric indicators. With these, we estimate models with a latent variable that describes perceived safety and latent classes. Results show perceived safety has a positive and relevant effect on infrastructure design preferences. Moreover, heterogeneous preferences are shown to exist, with some users putting a significantly higher weight on perceived safety.