This study uses the longitudinal features of the American Housing Survey (AHS) to identify the characteristics of units that change from being owner occupied to renter occupied as well as the reverse. Using 2001 and 2003 AHS national data, the study begins with a univariate analysis of the incidence of tenure change by 21 housing unit, household, neighborhood, and geographic characteristics. The study then develops logistic regression models to predict the probability of tenure change in one direction (either renter to owner, or owner to renter). Another modeling effort identifies the characteristics of units in the market segment that is likely to change tenure, without regard to direction. These analyses show that single-family attached units, units in smaller structures, and older units are more likely to change tenure. A surprising result is that many units changed tenure without changing households—the same people lived in the housing units before and after the change in tenure. The study explores the details of this phenomenon but does not reach a conclusion as to its cause.