Aim: To determine whether there are systematic differences in physical activity between residential areas after extensive control for sociodemographic factors at the individual level. Methods: Multi-level regressions of walking/bicycling, sedentary activities, household work and exercise were carried out in a representative sample of 68,303 adults in 39 residential areas in Stockholm County, first adjusting at the individual level for country of birth, sex, age, education, occupational class and income. The type of housing was then considered at the individual level or, for walking/bicycling and exercise, at both the individual and area levels (as a measure of area density). Results: After adjustment for sociodemographic factors, differences between residential areas remained in walking/bicycling, corresponding to 0.27 SD, or 50 min/week between the most and least active areas. Forty per cent of this difference could be explained by the type of housing at the area level. For sedentary activities and household work, respectively, much of the variation that remained after adjustment for sociodemographic factors was, in turn, explained by the type of housing at the individual level, leaving a difference of 0.16 SD (80 min/week) and 0.13 SD (60 min/week), respectively. For exercise, the corresponding difference was 0.11 SD (11 min/week, not sensitive to housing). Conclusions: Area level factors may influence walking/bicycling. High area density was associated with more activity. However, high density also comes with a type of housing (apartments) that is associated with less household work and, surprisingly, more sedentary activities, introducing a challenging trade-off. The differences in exercise were smaller than for all other types of activities.