- The structure of the Annual Housing Survey (AHS) data set has unique characteristics which complicate application of the standard methodologies of duration analysis. This paper derives a general likelihood function for estimating duration models using the AHS that incorporates these unique characteristics. A duration model is then estimated using separate samples of renters and owners from the AHS. With respect to household characteristics and mobility behavior, our results largely confirm previous findings by Henderson and Ioannides (Journal of Urban Economics, 26, 212-230 (1989)). In addition, we estimate completed lengths of stay that are 40 to 50% larger than the popular stereotype. 1992.