UC Davis

Residential self-selection bias is a concern in studies of neighborhoods and health. This bias results from health behaviors predicting neighborhood choice. To quantify this bias, we examined associations between pre-move health factors (body mass index, walking, and total physical activity) and post-move neighborhood factors (County Sprawl Index, Census tract socioeconomic status (SES)) in the Nurses Health Study (n = 14,159 moves from 1986-2008). Individuals in the highest quartile of pre-move BMI (BMI > 28.4) compared to the lowest quartile (BMI < 22.5) moved to counties that averaged 2.57 points lower on the sprawl index (95% confidence interval -3.55, -1.59) indicating that individuals moved to less dense counties; however, no associations were observed for pre-move walking nor total physical activity. Individuals with higher pre-move BMI tended to move to Census tracts with lower median income and home values and higher levels of poverty. Analyses examining the change in neighborhood environments after a move demonstrated that healthy pre-move behaviors were associated with moves to worse socioeconomic environments. This type of self-selection would bias results downward, underestimating the true relationship between SES and physical activity. Generally, the magnitudes of associations between pre-move health factors and neighborhood measures were small and indicated that residential self-selection was not a major source of bias in analyses in this population.