Background: Physical activity (PA) in older adults is influenced by a range of environmental, demographic, health-related, social, and psychological variables. Social cognitive psychological models assume that all influences on behaviour operate indirectly through the models constructs, i.e., via intention and self-efficacy. We evaluated direct, indirect, and moderating relationships of a broad range of external variables with physical activity levels alongside intention and self-efficacy. Methods: We performed a cross-sectional survey of a representative and stratified (65--80 and 80+ years; deprived and affluent) sample of 584 community-dwelling people, resident in Scotland. Objectively measured physical activity and questionnaire data were collected. Results: Self-efficacy showed unique relationships with physical activity, controlling for demographic, mental health, social, environmental, and weather variables separately, but the relationship was not significant when controlling for physical health. Overall, results indicating support for a mediation hypothesis, intention and self-efficacy statistically mediate the relationship of most domain variables with physical activity. Moderation analyses show that the relationship between social cognitions and physical activity was stronger for individuals with better physical health and lower levels of socio-economic deprivation. Conclusions: Social cognitive variables reflect a range of known environmental, demographic, health-related and social correlates of physical activity, they mediate the relationships of those correlates with physical activity and account for additional variance in physical activity when external correlates are controlled for, except for the physical health domain. The finding that the social cognition-physical activity relationship is higher for participants with better health and higher levels of affluence raises issues for the applicability of social cognitive models to the most disadvantaged older people.