The American Community Survey (ACS) will be a large continuing survey of the U.S. population using a Arolling sample@ design of the type described in Kish (1990). It is a mail survey with follow-up by telephone and personal visit of a sample of nonrespondents. The Census Bureau's ongoing Master Address File operation provides the frame. The ACS collects information about the same topics as the census Along form@ content sample questionnaire. After a period of testing and comparison to the 2000 census long form, the ACS sample will increase to about three million mailouts each year starting 2003, leading to replacement of the content sample in the 2010 census. This paper describes the ACS survey design, objectives, and estimation procedures along with the considerations that led to them. The major methodological issues in conducting the survey and assessing the quality of the data are outlined. Initial results from the 1996 tests in four counties are presented, along with plans for future research and testing.