Associations between neighbourhood walkability and cycling in Denmark

Background: Previous research has established four environmental attributes that contribute to neighbourhood ‘walkability’: street connectivity, land use mix, residential density, and retail floor area ratio. There is emerging evidence that these attributes are related to not only walking behaviour but also cycle use. Given the significant health benefits associated with regular commuter cycling, an understanding of the environmental correlates of cycling is essential. The aim of this study was to examine the link between walkability and transportation choices across three Danish cities where cycling culture differs from most other countries and bicycle share is much higher (17% of all trips in Denmark (2011)).
Methods: Geospatial and transportation data representing 123 geographic zones were extracted from the Danish National Transportation Survey (DNTS). A geographic information system was used to calculate a walkability index for each zone by combining z-scores for street connectivity, land use mix, residential density, and retail floor area ratio. Multiple linear regression and Pearson correlations were used to quantify the associations between walkability and walking, cycling, and passive transportation practices for each zone. Furthermore, the DNTS zones were divided into deciles on walkability index scores, higher education and personal income and a 4-level variable was created according to each group’s expected mode of transport: 1) high walkability – high % higher education; 2) high walkability – low % higher education; 3) low walkability – high % higher education; 4) low walkability – low % higher education. The 4-level variable made it possible to compare groups with similar walkability index scores but different educational levels.
Results: Walkability index scores and active transportation were positively correlated with coefficients ranging from 0.40 to 0.65. Significant differences were found on all transportation variables (active and passive) from neighbourhoods ranked as high walkability-high % higher education and neighbourhoods ranked as low walkability-low % higher education. Simple linear regression analysis on cycling showed that the walkability index scores had significant Pearson coefficients in relation to cycling (cycle kilometres: 0.15, p<0.001 and cycle trips: 0.06, p<0.01) and that inclusion of other variables in a multiple linear regression model (educational level, age and city) weakened the association, but walkability index scores remained positively associated with cycle- and walking trips and negatively on other trips (p<0.05).
Conclusion: Built environment factors related to walking behaviour are to a minor extent also applicable to cycling in Denmark but the associations are confounded by age, city and either % higher education or % lower education. The association between walkability and cycle- and walking trips cannot be explained by variation in the other variables (education level, age and city) even though the association is weak. This suggests that there is an association between the walkability index and the number of trips which implies that the walkability index functions best as a predictor for the number of trips within the DNTS zone, but not kilometres travelled by bike or by foot. This indicates that the use of a more bike-friendly index is needed to encompass e.g. the larger radius of action bicyclist have, compared to pedestrians. This information is potentially useful for future studies that link the built environment with cycling even though the study is in large descriptive and hypothesis generating and more detailed studies on an individual level is needed to state any causal relationship.