The aim of this study was to investigate the associations between three walkability parameters (residential density, street connectivity and land use mix) and physical activity and to analyze the mediating and moderating effects of vehicle ownership on these associations. The results showed that residential density and land use mix, objectively assessed within 1,000 meter network buffers around participants’ residences, are positively associated with time spent in MVPA and walking for transportation. This is in line with previous research investigating objectively assessed residential density and land use mix as separate measures
 or when incorporating these measures in indexes of overall walkability
Street connectivity was weakly associated with the amount of walking for transportation, but it was not associated with any of the other physical activity outcomes in this study. The lack of associations between street connectivity and physical activity outcomes is in contrast to some earlier findings from other studies. For example, Frank et al. found street connectivity to be significantly associated with moderate physical activity
. The non-significant association between street connectivity and physical activity in this study could be explained by a relatively high level of connectivity. The median number of intersections per square kilometer was 87 in this Swedish study, compared to a mean of 37 intersections per square kilometer found by Frank and colleagues in North America
. The lack of association between street connectivity and physical activity found in this study is, however, in line with the conclusions of a review by Saelens and Handy on environmental correlates of walking
. They found that while residential density and land use mix were consistently associated with walking for transportation, the findings for street connectivity were more equivocal.
We did not find any significant associations between walkability parameters and cycling for transportation. Even though we included the cycling infrastructure in our data, walkability was developed as a measure of supportive environments for walking and not cycling. Furthermore, a 1000-meter buffer may be too small to capture the area of exposure for cyclists. A study conducted in Stockholm investigating route distances in 110 street-recruited bicycle commuters found a mean commuting distance of 6.7 and 8.0 kilometers for women and men, respectively
. However, even smaller buffer zones (450m) have been used in previous research on environmental correlates of cycling for transportation
. Furthermore, it may be more common for cyclists to commute from residences in low walkable neighborhoods to workplaces in dense inner city areas than the opposite scenario, in order to avoid traffic congestions and parking problems. This would attenuate an association between neighborhood walkability and cycling for transportation. Future studies could explore this hypothesis using measures of walkability parameters around participants’ workplaces as well as their homes.
Previous studies have found positive associations between neighborhood walkability and active transport (walking + cycling)
[10, 26]. Other studies have examined the association between wakability and cycling for transportation alone. For example, participants in the Belgian Environmental Physical Activity Study living in highly walkable neighborhoods (walkability assessed within administrative areas) reported 40 minutes more cycling for transportation per week compared to participants living in less walkable neighborhoods
. Results from a study by Winters and colleagues showed positive associations between objectively assessed population density, street connectivity and land use mix and cycling for transportation
. Furthermore, Titze et al. found a positive association between perceived street connectivity and cycling for transportation
Vehicle ownership mediated a statistically significant proportion of all the significant associations between walkability parameters and physical activity outcomes. For example, 34% of the association between land use mix and time spent in MVPA were mediated by vehicle ownership. To our knowledge, no previous studies have investigated vehicle ownership as a mediator between objectively assessed walkability parameters and physical activity outcomes. Therefore, our results are hard to compare with the currently available knowledge base. However, our results are in line with the findings of a study by Sehatzadeh et al. in which fewer vehicles were owned by households in walkable environments and where the number of vehicles in the household was negatively associated with frequency of walking
. This is also supported by results from a longitudinal study by Mumford and colleagues, where participants reported more walking and less automobile use after moving to a community with a high land use mix
We did not find any significant effect modification by vehicle ownership on the associations between walkability parameters and physical activity outcomes. Participants living in dense areas with a mixed land use spent more time in MVPA and reported more walking for transportation compared to participants living in areas with lower residential density and land use mix, regardless of vehicle ownership. This is in contrast to some previous findings where vehicle ownership, or similar vehicle-related measures, moderated the relationship between the environment and physical activity. For example, driving status modified the association between convenience of bus services and physical activity in a Japanese study
 and preference for passive transport modified the association between walkability and numbers of steps per day in a Belgian setting
. However, the present study and the studies by Kamada et al. and Van Dyck et al. used different explanatory as well as outcome measures. For example, preference for passive transport may have a different influence on the association between walkability parameters and physical activity compared to vehicle ownership.
This study has some limitations that should be considered. It is a cross-sectional study and therefore causality cannot be determined. Self-report measures of walking and cycling for transportation may include bias due to social desirability and difficulties to recall activities during the past seven days. Accelerometers, on the other hand, do not suffer from these biases and provide an objective measure of physical activity on a moderate to vigorous intensity level. Strengths of this study also include the large number of participants (n=2,178) and the objective measures of walkability parameters using network buffers. The network buffers were based on detailed network data, including the road network as well as bicycle paths and footpaths. This provides a more relevant area of exposure for cyclists and pedestrians compared to network buffers based solely on the road network. Finally, participants were recruited from neighborhoods with a wide range of walkability and neighborhood-level SES, which is an additional strength.