In this study, cross-sectional analyses support an association between less urban sprawl and more walking, as well as lower prevalence of overweight and hypertension. But in longitudinal analyses, men moving from more to less sprawling counties did not increase their walking or lower BMI. These observations suggest that the cross-sectional findings may reflect selection bias (i.e., active individuals choose neighborhoods conducive to physical activity), rather than indicating that the built environment – at least, as measured by urban sprawl – increases physical activity. However, the longitudinal analyses were limited by small numbers of men moving from more to less sprawling counties, and vice versa.
The hypothesis that physical activity can be influenced on a large scale by modifying the environment is attractive and plausible. A parallel may be drawn with tobacco control, where environmental and public health policies have decreased smoking levels more than individual interventions.33
In designing urban areas, land use and transportation planners have long considered how design affects human behavior, primarily in relation to costs and benefits.8
More than 50 studies in this field have examined the built environment and utilitarian travel;8
walking and bicycling for transportation were consistently associated with less sprawling communities.34
In the public health literature, several studies also have reported similar associations.35,36
The primary limitation of existing studies is that they are almost exclusively cross-sectional.9,12
This study design cannot differentiate whether the environment influences physical activity (“environmental determinism”37
), or whether active individuals choose activity-friendly neighborhoods (self-selection). A recent study reported that persons preferring, and living in, walkable neighborhoods walked more (34%) and drove less (26 miles/day) than those preferring, and living in, car-dependent neighborhoods (3% walked; 43 miles/day driven).38
Cross-sectional studies which adjust for residential preferences continue to show an association between walkable neighborhoods and more walking.38,39
Limited data are available from longitudinal studies, which can mitigate self-selection bias. Environmental changes in a San Diego naval air station were associated with improvements in the fitness of active-duty personnel,40
while improved lighting on three urban London streets,41
and improved bicycle paths on six streets in Toronto42
were associated with more persons walking and greater bicycle traffic.
For body weight, cross-sectional studies also consistently show significant associations with the built environment,43
but few longitudinal data exist. The 1997 US National Longitudinal Study of Youth (NLSY) observed findings similar to the present study: significant cross-sectional associations between more sprawl and higher BMI, but no significant association between changes in sprawl and BMI in longitudinal analyses.23
However, another study of 262 men and women who originally participated in the 1979 NLSY reported a significant association between moving to less sprawling counties and decreased BMI.20
This study also indicated some self-selection: persons with lower BMI were more likely to move to less sprawling counties.20
A further limitation is that many studies have relied on subjects’ perception of their physical environment.36
Active persons may be more aware of exercise facilities, or they may be more willing to walk further to their destinations. Studies comparing subjects’ perceptions with objective data derived from geographic information system (GIS) technology indicate low to fair agreement, with the highest agreement among regularly active subjects.44,45
This could bias findings towards an association between activity-friendly environments and physical activity levels.
The major strengths of this study are its large size, longitudinal design, and objective and validated measures of the built environment. It also possessed validated and detailed measures of physical activity, and a homogenous population, which minimizes confounding by socioeconomic status and race.
Limitations include, in particular, few men changing urban sprawl levels over time. However, even disregarding statistical significance, the point estimates do not support the hypothesis that less sprawl increases physical activity – men moving to less sprawling areas showed point estimates of decreased physical activity. And, congruent with this, these men significantly increased their BMI, compared to men with unchanged sprawl levels. Because of small numbers, we could not meaningfully examine changes in the proportions meeting physical activity recommendations between 1988 and 1993 with corresponding changes in urban sprawl. Second, the assessment of sprawl may have been too imprecise – counties are large geographic areas – to observe an association. However, expected associations were observed in cross-sectional analyses; thus, the measure of sprawl was at least precise enough for these analyses. Third, finer features of the built environment (e.g., streetlights, traffic, etc.) were not assessed; nonetheless, this study still provides important information for a “macro” level of the built environment.
Fourth, occupational physical activity was not assessed; however, Harvard alumni likely had white-collar jobs involving primarily walking and stair-climbing, which were assessed. Additionally, walking for exercise versus travel, which is more likely to be influenced by the environment,8,12
was not differentiated. This, however, unlikely caused a major bias because expected associations between urban sprawl and overall walking were observed in cross-sectional analyses. Fifth, the reasons for, and timing of, moves to a new residence were not ascertained. Moves may have occurred because of ill-health, and a short duration at the new residence may be insufficient for change in physical activity to occur. Finally, while the homogeneity of the population limits generalizability to other groups; younger men may have a wider range of activity, allowing for better detection of any differences, and their behavior also may be more strongly influenced by the environment.46
In conclusion, the longitudinal data from this study do not support the hypothesis that the built environment, as measured by county-level urban sprawl, can increase physical activity; however, statistical power was limited. More research is needed before dedicating vast sums of money to urban redesign with the sole intention of increasing physical activity. Studies of longitudinal design among diverse populations, with adequate sample sizes among persons who move, and which employ objective and precise characterizations of the built environment, are needed.
This is report No. XC in a series on chronic disease in former college students.