Citation

McCreery-Phillips S, Heydari S. Transp. Policy 2023; 142: 152-161.

Copyright

(Copyright © 2023, Elsevier Publishing)

DOI

10.1016/j.tranpol.2023.08.007

PMID

unavailable

Abstract

As the need to encourage modal shift from motorised vehicle use to active modes becomes greater, it is important to understand the key factors influencing the decision of how to travel. This paper explores the association between bicycle commuting and a range of sociodemographic and built and natural environment characteristics across wards and boroughs in Greater London, UK, with an aim to identify the key factors which influence participation. We employed a Bayesian multilevel heteroskedastic model with heterogeneity in variance, which can address dependencies in the data and unobserved heterogeneity more fully. This allowed us to account for unobserved/unmeasured covariates such as collective attitudes and the existence of cycling cultures that may differ between Greater London boroughs. We found that the propensity for bicycle commuting increases with an increase in the employment rate, the populations of white British and mixed white and black Caribbean, the proportion of terraced houses, and cycle network density. Conversely, we found that the propensity for bicycle commuting decreases with an increase in the absence of academic qualifications, the area of non-domestic buildings, the population of Indians and Pakistanis, and the number of cars per household. Our analysis also revealed important between-borough variations in the effect of key explanatory variables. Notably, the effects of the populations of Indians, Pakistanis, and mixed white and black Caribbean, and the number of cars per household all vary across Greater London boroughs. Finally, by allowing for heterogeneity in variance, we found that rates of bicycle commuting are more dispersed in Inner London and as the number of cars per household increases. Our analysis highlights the importance of cycling infrastructure in promoting bicycle commuting.

Language: en

Keywords

Active travel; Bayesian multilevel heteroscedastic model; Bicycle commuting; Cycling network; Neighbourhood characteristics