This study aimed to investigate gender, race/ethnicity, education, and income as moderators of relations of perceived neighborhood crime, pedestrian, and traffic safety to physical activity. Participants were from two samples: adults (N = 2199, age = 25-65 yr) and older adults (N = 718, age = 66+ yr) from high- and low-walkable neighborhoods in the Washington, DC, and Seattle, Washington, areas. Neighborhood safety and transportation and leisure walking were assessed via survey, and moderate to vigorous physical activity was assessed using accelerometers. Sociodemographic moderators were investigated using interaction terms and follow-up within-group tests from mixed-effects regression models. Overall direct effects of safety on physical activity were not found, with one exception. Seven interactions were found in each sample. Interactions were found for all physical activity outcomes, although total moderate to vigorous physical activity was involved in more interactions in adults than older adults. Half of the interactions revealed significant positive relations of pedestrian and traffic safety to physical activity in the more affluent/advantaged group (i.e., high education, high income, and non-Hispanic white) and null associations in the less affluent/advantaged group. Race/ethnicity was a moderator only in older adults. One-third of the interactions involved gender; half of these involved crime safety. Interactions involving crime safety showed nonsignificant positive trends in the more affluent/advantaged group and women and nonsignificant negative trends in the less affluent/advantaged group and men. Sociodemographic moderators of neighborhood safety explained some of the variation in adults' and older adults' physical activity. Patterns suggested positive associations between safety and physical activity in participants with more affluent/advantaged sociodemographic characteristics, although some patterns were inconsistent, particularly for gender. More refined conceptualizations and measures of safety are needed to understand if and how these constructs influence physical activity.