We apply a marginal structural modeling (MSM) strategy to investigate the relationship between neighborhood poverty and BMI level among U.S. black and white adults. This strategy appropriately adjusts for factors that may be simultaneously mediators and confounders (e.g., income, health behavior), strengthening causal inference and providing the total (direct and indirect) neighborhood effect estimate. Short and long-term neighborhood poverty were positively associated with being overweight for both black and white women. No link was found for either black or white men. Socioeconomic and behavioral factors do not appear to be strong mediators. Sensitivity analyses suggest that the direction of point estimates is robust to unobserved confounding, though 95% confidence intervals sometimes included the null, particularly for white women. Compared to previous cross-sectional and longitudinal analyses, MSM results provide stronger evidence for a causal link between neighborhood poverty and body weight among women.