The Missing Link: Bicycle Infrastructure Networks and Ridership in 74 US Cities.

| David Levinson

Recently Published:

Seattle bike network
Seattle bike network

Abstract: Cities promote strong bicycle networks to support and encourage bicycle commuting. However, the application of network science to bicycle facilities is not very well studied. Previous work has found relationships between the amount of bicycle infrastructure in a city and aggregate bicycle ridership, and between microscopic network structure and individual tripmaking patterns. This study fills the missing link between these two bodies of literature by developing a standard methodology for measuring bicycle facility network quality at the macroscopic level and testing its association with bicycle commuting. Bicycle infrastructure maps were collected for 74 United States cities and systematically analyzed to evaluate their network structure. Linear regression models revealed that connectivity and directness are important factors in predicting bicycle commuting after controlling for demographic variables and the size of the city. These findings provide a framework for transportation planners and policymakers to evaluate their local bicycle facility networks and set regional priorities that support nonmotorized travel behavior, and for continued research on the structure and quality of bicycle infrastructure and behavior.
Keywords Bicycling · Travel Behavior · Networks

One thought on “The Missing Link: Bicycle Infrastructure Networks and Ridership in 74 US Cities.”

  1. Fascinating work. Couple of thoughts:

    1. Would it be possible to make the dataset available in a unified form, either at the ArcGIS level or at the aggregate network structure variables level? (Of course I can’t guarantee that I or others would actually use it… so likely not worth your time.)

    2. Have you considered applying the model to a selection of Dutch cities to see if it continues to hold across a truly wide range of mode shares? They will be way off your bell curve in connectivity, density, and mode share, and it’d be interesting to see how far the model can be pushed. Obviously ACS data won’t be available but I’m sure there are Dutch sources for both infrastructure networks and cycle mode shares that would be good enough.

    Like

Comments are closed.