Background To examine the association between socioeconomic status (SES) and risk of frailty, and to assess whether behavioural and clinical factors (BCF) mediate this association.

Methods Cohort of 1857 non-institutionalised individuals aged ≥60 years recruited in 2008–2010 and followed through 2012. Education, occupation, and BCF were ascertained at baseline, and incident frailty was assessed at follow-up with the Fried frailty criteria.

Results Men showed no differences in frailty risk by education or occupation. Compared with women with university education, the adjusted OR (aOR) adjusted for age and the number of frailty criteria at baseline for incident frailty in women with primary or lower education was 3.02 (95% CI 1.25 to 7.30); once fully adjusted for BCF, the OR was 2.00 (95% CI 0.76 to 5.23). No alcohol intake (vs light–moderate), longer time spent watching TV, less time spent reading, and a higher frequency of obesity, depression and musculoskeletal disease in those with primary or lower education accounted for most of the decline in OR. BCF explained 50.5% of the excess frailty risk associated with lower education. The aOR of frailty incidence for manual versus non-manual occupation was 2.24 (95% CI 1.41 to 3.56) versus a fully aOR of 2.05 (95% CI 1.24 to 3.37). BCF explained 15.3% of the association, with individual mediators being similar to those for education-related differences.

Conclusions A lower education or a manual occupation was associated with higher frailty risk in older women. These associations were partly explained by lower alcohol consumption, higher sedentariness, and higher obesity and chronic disease rates in women with lower SES.