The walkable community (Daybreak) children achieved 4.65 more minutes MVPA during the 1/2 hours before and after school than students from the less walkable community. The mixed community students achieved 5.66 more minutes of MVPA and may have benefitted from the fact that part of their route to school traversed the walkable community The longer walk perhaps accounted for their greater MVPA relative to Daybreak students. The walkable Daybreak community boys engaged in about 18 more MVPA minutes during the late afternoon/evening time period than children from all other groups. These results are consistent with claims that more walkable community designs support more physical activity.
Although selection threats are common in observational studies of community design, separate tests suggested that both selection and environmental influences operated in the study. Recall that selection in this study would occur if walkable Daybreak simply attracts more physically active residents. If pre-existing high activity levels and preferences cause these families to choose walkable Daybreak, family members should show greater MVPA across many settings, including at school, but this did not occur. If selection in the form of parental preferences for their children to walk to school causes the results, then controlling for these parental preferences should have reduced to insignificance all community differences, and it did not. It did reduce from significance (p =.007) to marginal significance (p = .061) the tendency for walkable Daybreak students to achieve more MVPA on the walk to school. Thus, like many other studies of selection effects in the walkable design and physical activity literature [32, 48], selection may play some role, but does not account for all physical activity differences across communities. Also, it is possible parental preferences did not predate the move to walkable Daybreak; parents may have developed their preference for their child to walk to school after experiencing the walkable community. Ideally, attitudinal controls for selection would be measured prior to the move to the new environment, to see if the preferences really predated the move. Therefore, measuring parental preferences post-move may have controlled for reactions to environment rather than pre-existing preferences, making this an especially conservative statistical adjustment.
To provide more comprehensive tests of selection, future research is needed on large samples of households that are moving. Ideally, researchers could track pre- and post-move physical environmental features of walkability, physical activity supports such as parks, objective physical activity levels, and extensive attitudinal measures that assess why families move to new neighborhoods. A more extensive list of reasons for moving to a neighborhood could provide a more comprehensive assessment of attitudinal confounds. By tracking the moves of residents to more and less walkable neighborhoods, researchers could provide the strongest quasi-experimental study design, short of random assignment.
Of the models identified in Table 1, the results (row 4) show mixed support for all three, but strongest support for the environmental model. Consistent with the environmental effects model, walkable Daybreak and mixed community children are more active during the 1/2 hours before and after school, when likely exposed to the walkable features of Daybreak (effects are marginally significant in favor of Daybreak students in the 1/2 hour before school, but significant for the 1/2 hour after school). Walkable Daybreak boys--but not girls--are more active during school-day late afternoons/evenings. No weekend differences emerged across communities; perhaps the mixed and less walkable community students go to more PA-friendly areas on weekends; GPS tracking could address this possibility.
Boys were often more active than girls, a common finding [49, 50]. However, the gender gap was closed for the half hour after school. Given pervasive gender differences in PA, any settings that overcome those differences are of particular value. The greater MVPA on school evenings for walkable Daybreak boys but not girls suggests a gender by environment interaction, consistent with some past research. For example, in past research, PA was higher for boys living near park and recreational land , but higher for girls living near private recreational facilities . New urban designers and health advocates need more research to understand gender-specific appeals of PA design supports in their communities.
Notice that this study contrasted MVPA differences across communities, rather than the more typical test of MVPA differences between walkers and non-walkers to school. It should be more difficult to detect MVPA differences across communities compared to detecting differences across groups defined by whether they engage in the physical activity of walking to school or not. Yet the extra 4.65 minutes during the 1/2 hour before/after school for walkable Daybreak students, computed for the whole group, is within the range of differences other studies have found when they compare walkers to non-walkers. Walkers' extra minutes in past research include: 3.5 in the to-school hour in Denmark ; 8-14 in the 2 hours before/after school in England ; 5.98 and 9.77 (for 0.5-1 mile and 1-5 mile distances from home to school, respectively) in England ; 3.2 for to-school, 4.7 for from-school, and 9.5 for to/from school walkers in six U.S. states ; and about 10 minutes for walking to/from school in South Carolina among regular walkers .
Longitudinal studies are needed to address the inherent weakness of our cross-sectional design, although this study provided more comprehensive tests of the selection threat than do many cross-sectional studies. Sample sizes are also small, especially at walkable Daybreak, which limits power. The school district also restricted survey length, which limited controls for other potential covariates, such as income. GPS data were also not available to confirm where PA occurred. Finally, we consider the possibility that parental preference controls yield conservative results, given that those preferences may have arisen after the move to the community, instead of indexing pre-existing preferences.