To identify the impact of socioeconomic status on incident impaired glucose metabolism and type 2 diabetes and to investigate the mediating role of health behaviours on this relationship using national, population-based data.The Australian Diabetes Obesity and Lifestyle (AusDiab) Study is a national, population-based, longitudinal study of adults aged 25 years and above. A total sample of 4,405 people provided complete baseline (1999-2000) and 5 year follow-up (2004-2005) data relevant for these analyses. Fasting plasma glucose and 2 h plasma glucose were obtained from an OGTT, and demographic, socioeconomic and behavioural data were collected by interview and questionnaire. Multinomial logistic regression examined the role of socioeconomic position in the development of diabetes and mediation analyses tested the contribution of health behaviours in this relationship.Highest level of education was a stronger predictor of incident impaired glucose tolerance and type 2 diabetes (p = 0.002), compared with household income (p = 0.103), and occupational grade (p = 0.202). Education remained a significant independent predictor of diabetes in fully adjusted models. However, the relationship was attenuated by the health behaviours (smoking and physical activity). Mediation analyses indicated that these behaviours were partial mediators (explaining 27%) of the socioeconomic status-diabetes relationship.Smoking and physical activity partly mediate the relationship between low education and type 2 diabetes. Identification of these modifiable behavioural mediators should facilitate the development of effective health promotion campaigns to target those at high risk of developing type 2 diabetes.