Abstract Background Inter-relationships among built and socioeconomic environmental characteristics may result in confounding of associations between environment exposure measures and health behaviors or outcomes, but traditional multivariate adjustment can be inappropriate due to collinearity. Methods We used principal factor analysis to describe inter-relationships between a large set of Geographic Information System-derived built and socioeconomic environment measures for adolescents in the National Longitudinal Study of Adolescent Health (Wave I, 1995-96, n = 17,294). Using resulting factors in sex-stratified multivariate negative binomial regression models, we tested for confounding of associations between built and socioeconomic environment characteristics and moderate to vigorous physical activity (MVPA). Finally, we used knowledge gained from factor analysis to construct replicable environmental measures that account for inter-relationships and avoid collinearity. Results Using factor analysis, we identified three built environment constructs [(1) homogenous landscape; 2) development intensity with high pay facility count; 3) development intensity with high public facility count] and two socioeconomic environment constructs [1) advantageous economic environment, 2) disadvantageous social environment]. In regression analysis, confounding of built environment-MVPA associations by socioeconomic environment factors was stronger than among built environment factors. In fully adjusted models, MVPA was negatively associated with the highest (versus lowest) quartile of homogenous land cover in males [exp(coeff) (95% CI): 0.91 (0.86, 0.96)] and intensity (pay facilities) [exp(coeff) (95% CI): 0.92 (0.85, 0.99)] in females. Single proxy measures (Simpson's diversity index, count of pay facilities, count of public facilities, median household income, and crime rate) representing each environmental construct replicated associations with MVPA. Conclusions Environmental characteristics are inter-related. Both built and SES environments should be incorporated into analysis in order to minimize confounding. Single environmental measures may be useful proxies for environmental constructs in longitudinal analysis and replication in external populations, but more research is needed to better understand mechanisms of action, and ultimately identify policy-relevant environmental determinants of physical activity.