PURPOSE: Test for a causal relationship between neighborhood design and physical activity within the neighborhood by controlling for self-selection. DESIGN: Cross-sectional and quasi-longitudinal analyses of residents of selected neighborhoods. SETTING: Eight Northern California neighborhoods. SUBJECTS: Random sample of 1682 adults stratified by movers (moved within 1 year) and non-movers (moved >1 year ago) responding to self-administered mail surveys (24.7% response rate). MEASURES: Self-reported number of days in last 7 days of moderate to vigorous physical activity somewhere in the neighborhood and self-reported change in physical activity in the neighborhood from prior to moving (for movers) or from 1 year ago (for non-movers). ANALYSIS: Zero-inflated Poisson regression for cross-sectional analysis (n = 1497); ordered probit model for quasi-longitudinal analysis (n = 1352). RESULTS: After we controlled for physical activity attitudes and neighborhood preferences, selected neighborhood design characteristics were associated with physical activity within the neighborhood and changes in selected neighborhood design characteristics were associated with changes in physical activity within the neighborhood. CONCLUSIONS: Both cross-sectional and quasi-longitudinal analyses provided evidence of a causal impact of neighborhood design. Improving physical activity options, aesthetic qualities, and social environment may increase physical activity. Critical limitations included self-report measures of physical activity, lack of measures of duration and intensity of neighborhood physical activity, lack of measures of total physical activity, and limited measures of preferences related to physical activity.