Main findings
In this cross-sectional analysis, we found no overall association between exposure to urban motorway infrastructure and the likelihood of local walking or cycling, or of the number of local destinations accessed on foot or by bicycle.
However, in the analyses stratified by study area, we found that in the area around the new motorway (M74), a unit increase in proximity to the motorway was associated with a lower likelihood of local recreational walking and cycling (OR 0.63, 95% CI: 0.39 to 1.00, p = 0.050) a pattern not found in the area with the longstanding motorway (M8). Conversely, we found that in the area near the longstanding motorway a unit increase in proximity to motorway infrastructure was statistically significantly (p = 0.014) associated with about a 12% decrease in count of types of destination accessed via active travel, a pattern not found in the new motorway (M74) study area. Sensitivity analysis results, with an altered exposure definition, also differed by area.
This provides important quantitative evidence that an association between exposure to urban motorway infrastructure and local walking and cycling behaviour exists and suggests it may vary by the characteristics of the motorway to which a participant is proximal, such as the age of the motorway. The conflicting nature of the findings mirror the conflicting prior claims [21] that living near a motorway could have both positive and negative impacts on local walking and cycling. The lack of associations seen in the study area with no motorway suggests a threshold effect whereby beyond a certain distance from a motorway, additional distance makes no difference.
Strengths and limitations
This study adds to a growing body of evidence examining the association between the built environment and health behaviour and is one of few to address the associations of walking and cycling with the presence of urban major road infrastructure. Using a GIS, we objectively defined exposures, using models controlled for a series of potential confounders.
We also acknowledge the study limitations. In this cross-sectional study, temporality of association between exposure and outcome could not be demonstrated. Because the exposure could not be assigned at random there may also be some residual confounding in unmeasured factors or characteristics [33]. The potential confounding influence of socioeconomic status, which may influence both location of residence and options for travel beyond walking and cycling, is particularly important. We used car and home ownership as proxies of socioeconomic status in the fully adjusted models, however these are blunt measure for this nuanced characteristic. We also derived a measure of whether participants were employed, in education or volunteering (or not) which may have captured both a measure of socioeconomic status, but also in part provided a purpose for travel or reasons to leave the house.
There was a comparatively low response to the survey, which may limit the external validity of the findings, although our response rate was not unusual for this type of natural experimental study [34, 35]. Our examination of likelihood of active travel reports relative impacts from a controlled comparison across three similar study areas, which improves the methodological robustness of the study.
Further biases may be introduced due to the self-report nature of our outcome. However, error introduced during self-report is unlikely to differ by level of exposure to motorway infrastructure. Our self-report tool was a relatively blunt indicator of local walking and cycling and did not assess duration of activity. Similarly, the count of types of destinations accessed should not be misinterpreted as number of trips made to particular destinations, or reflective of the variety of destinations of that type in the area. For example, a person making one journey in the 7 days to one shop would be recorded as the same as a person making daily journeys to seven different shops during the study period.
In an effect termed ‘resident/migrant bias’ or ‘residential self-selection’ [4], individuals may choose to live in areas which reflect their pre-existing lifestyle choices. In this example, someone who does not intend to walk or cycle locally might move to an area where it is not easy to walk and cycle because it does not matter to them, or a person intending to travel predominantly by car move near a motorway. In this case, the behaviour ‘causes’ the area choice, resulting in reverse causality. Several studies investigating residential self-selection, however, indicate that regardless of personal preference and intentions on relocation, the built environment still has measurable effects on walking and cycling [36,37,38].
Finally, it is usually challenging to assess the degree of external validity or generalisability of natural experimental studies. While considerable effort was put into the delineation of the study areas and recruitment, we make no claim that our sample was representative of the local population. Self-report postal surveys may lead to over-representation of more educated and/or less deprived individuals due to the cognitive burden of completion, although compared to other survey methods they may have less social desirability bias [39]. However, it is reasonable to assume that some association between urban roads and local walking and cycling would also be seen in other contexts.
Comparison with other studies
In the small body of existing literature on the effects of busy or large roads on walking and cycling, a conflicting picture emerges. Although in several studies presence of such roads is correlated with lower walking and/or cycling than in their absence (as expressed in a heterogeneous range of outcomes) [40,41,42,43,44,45], in some cases the reverse is true [46,47,48]. In other studies, a mix of both associations was observed or theorised [49,50,51], and in some cases no association could be seen [20, 29, 52, 53].
In a previous analysis from the parent study from which these data originate, utilising 1 day travel diaries rather than the 7 day data reported here, residents living near the new motorway (M74) were more likely to travel by car than those living further away, but there was no evidence of an association with total active travel (either an increase or a decrease) [20]. In a qualitative exploration from the same study, residents living near the new motorway (M74) described mixed effects of the motorway on active travel [26]. Participants more often described active travel changing experientially rather than in volume or frequency. Taken together with these findings, it is possible that the findings of the present study indicate differing relationships between environment and walking and cycling behaviour depending on whether this behaviour is for utility (active travel) or recreation, and that changes in local behaviour following local changes may be distinct from total travel behaviour.
Aesthetic appeal has been positively associated with recreational walking and cycling [24]. A new motorway could diminish aesthetic appeal, reflecting the lower likelihood of recreational walking and cycling we observed in those living in closer proximity to the new motorway (M74).
Our findings suggested people living near a longstanding motorway (M8) accessed a lower number of types of local destinations by active modes. This could reflect processes of community severance, such as fewer community amenities to access nearer a motorway, or a process by which an individual may be unaware of, or not drawn to local places near a motorway as they redefine their neighbourhood into a smaller geographical area [8, 54]. Severance effects have been described as taking time to accrue [55], which might explain why this association was not observed in the new motorway (M74) study area.
Conversely, our findings also demonstrate a positive association between living near the new motorway (M74) and local active travel, and between road network proximity to a motorway access point and local active travel in those living near the longstanding motorway (M8). It is possible that this reflects some of the beneficial effects of motorways proposed by the advocates of the new M74 – namely that motorways divert what would otherwise be local traffic, making the local area more appealing for walking and cycling [21]. However, as proposed in the previous section, this may also reflect the confounding effect of socioeconomic status, whereby those economically obliged to live nearer the motorway are also less likely to own a car.
Implications for research
This cross-sectional study provides a justification for further investigation of the association between major urban road infrastructure and local walking and cycling. Establishing the nature, direction and temporality of the associations found in our study will be important to guiding future policy decisions on road construction and planning. We have highlighted a number of potential mechanisms of effect, warranting further study. Our study adds evidence to the theory that recreational walking and cycling, as opposed to active travel, should be considered separately in research and their differing determinants elucidated. We suggest that further research should explore these themes using longitudinal or evaluative methods and detailed assessments of travel behaviour to strengthen the evidence base. This would ideally include more granular assessment of count of trips to different destinations, access to destinations and the diversity of specific destinations accessed.
Implications for policy
Current national and regional transport policy in Scotland prioritises investment in sustainable transport modes and improving the accessibility of the transport network to disadvantaged communities [56, 57]. In this context, this study offers important quantitative evidence that the association between new roads and walking and cycling is not null, but the direction of effect and explanations for it may be complex. Health impact assessments for new roads should consider potential impacts on total and local walking and cycling.
It is often argued that new roads help regenerate disadvantaged areas [21], suggesting socioeconomic factors should also be considered in any impact assessment. That increased proximity to a new motorway was associated with increased likelihood of local active travel may suggest a health benefit, but may also reflect socioeconomic factors such as the aggregation of those without car access. A recent analysis of Scottish Household Survey 1 day travel diaries found likelihood of an active journey stage was higher for those living in the most deprived areas than for those in the least deprived [58]. The construction of a motorway, intended to facilitate motorised transport, in an area where nearly half the local population do not own a car (44% in our study participants), and therefore cannot make use of the infrastructure, may be contrary to the principles of social justice. However, if socioeconomic circumstances improve as result of a new road, this might have an unintended consequence of increasing car ownership and thus diminishing active travel. How to preserve active travel through such changes is an important public health consideration.