Aims: To assess the prospective association between alcohol consumption and self-rated health: in particular whether there is a relationship between stable alcohol intake and health; whether health is affected by changes in alcohol consumption; whether having a chronic condition alters the relationships between stable and changing alcohol intake and health; and whether the health of longer-term abstainers is different from more recent and intermittent abstainers. Design: Longitudinal analysis of a prospective, population-based study. Setting: Australia. Participants: A total of 13 585 randomly selected 45–50-year-old women surveyed in 1996, of whom 9396 (69%) were resurveyed in 1998, 2001 and 2004. Measurements: Estimates for the General Health subscale of the SF-36 for different levels of alcohol intake adjusted for having a chronic condition, depression, smoking and other factors. Findings: Longitudinal models of consistent alcohol intake showed mean scores for general health of moderate drinkers were significantly better than that of non-drinkers [mean difference = 4.3, standard error (SE) = 0.61], occasional drinkers (mean difference = 3.1, SE = 0.52) and heavy drinkers (mean difference = 2.1, SE = 1.00). Among moderate drinkers, a decrease or variation in alcohol consumption was associated with a significant decline of three to four points in general health. Similar results were obtained when women with an existing chronic condition were excluded from these models. The health of recent abstainers and intermittent drinkers was the same as longer-term abstainers. Conclusions: Consistent moderate drinkers had the best health even after adjustment for having a chronic condition, depression and life-style factors. Poorer health was associated with decreased alcohol intake among occasional and moderate drinkers.