LAUSR

Abstract To examine whether residential self-selection bias influence the associations between neighborhood physical activity (PA) facilities and respondents' PA behavior. This study uses a natural experimental research design. Three waves of China Family Panel Studies data were used. The outcome variables were the weekly duration and frequency of PA. The availability of neighborhood PA facilities was measured using the presence of PA facilities and the number of PA facilities within residential neighborhoods. Multilevel regression models were used to examine the relationships between the availability of neighborhood PA facilities and respondents' PA behavior, adjusted for covariates. Associations were stratified by two groups (those living in private housing vs. public housing). Residents of private housing can choose their residential location, whereas those in public housing have little freedom to do so. Therefore, comparing these two groups can help us determine whether residential self-selection bias exists. The results show that both the presence and number of neighborhood PA facilities are positively associated with the frequency and duration of PA among residents of private housing. However, the evidence does not support an association between the availability of neighborhood PA facilities and PA behavior among residents of public housing. Residential self-selection bias may have a pronounced effect on the findings of studies of the association between neighborhood PA facilities and individuals’ PA behavior. Therefore, it is important to control for residential self-selection bias when examining built environment–PA associations.