Even if selection bias cannot be ruled out (e.g. those that agreed to participate could have healthier behaviours than nonparticipants), the data were obtained from a relatively large, random sample of older adults sampled from the Australian Electoral Commission electoral roll (voting is compulsory for person aged 18 years and over), which limits the risk of self-selection bias. The cross-sectional design and inability to determine causality of effect is also a limitation.