BACKGROUND As the United States population ages, an unprecedented proportion of the population will be aged 70 and older. Knowledge of alcohol use and its consequences in this age group is not well known. In light of the disparate findings pointing to negative outcomes with excessive drinking yet also benefits of moderate drinking, the true risk of alcohol use in late life needs more investigation. METHODS This study examined the correlates and 2-year health outcomes related to alcohol use in 7,434 elders aged 70 years or older. Data was collected as part of the Assets and Health Dynamics of the Oldest Old (AHEAD), a nationwide health and economic study of elders. Data from Wave 1 and Wave 2 of AHEAD are presented. Frequency and quantity of drinking was assessed by self-report as was health status, lifetime alcohol or psychiatric problems, presence of chronic illness, functional impairment, and depressive symptoms. Cognitive status was assessed using a brief measure. RESULTS Approximately 44% of the sample reported any alcohol use in the past three months, with the majority of drinking less than daily. Daily drinking was associated with being Caucasian, married, in relatively good health, and having good affective and cognitive status. Drinking was not associated with negative health outcomes two years later and was protective against stroke and functional impairment. Decline in drinking between Wave 1 and Wave 2 was strongly associated with poor health. CONCLUSION This study offers no evidence of negative health outcomes for drinking moderately and confirms the U-shaped curve often found in studies of alcohol and health. Nonetheless, cessation of drinking was associated with poor health suggesting the health benefits of moderate drinking may result from selection of a healthy group of people capable of sustained moderate drinking. Public health recommendations for moderate drinking must take this phenomenon into account.