Binary logit models are used to predict coverage, knowledge and perceptions of superannuation on the basis of individual demographic, socioeconomic and financial characteristics. The data is drawn from the 2003 ANZ Survey of Adult Financial Literacy in Australia and relates to 3,548 respondents. Knowledge of superannuation is defined, amongst other things, in terms of understanding superannuation fees and charges and statements, recognising the voluntary and compulsory nature of additional employee and employer contributions and the recognising the lower taxation of superannuation compared to other investments. Factors examined include gender, age, ethnicity, occupation, educational level and family structure, along with household income, savings, and mortgage and non-mortgage debt. The evidence suggests that knowledge of superannuation is unevenly spread across respondents, even those with superannuation funds. Such knowledge is generally lowest for females, those from a non-English speaking background, those with low levels of secondary education and persons aged less than thirty. Knowledge is generally better for professionals, those aged over forty or retired and the university educated. The models best predict the overall coverage of superannuation and for those with superannuation, the compulsory nature of employer contributions and the ability to read and understand superannuation statements.