Most tenure choice models using cross sectional data have used either a sample of recent movers or a sample comprising all households. There are problems with estimating both types of models in cross sectional data. A sample of recent movers oversamples renters, and a sample of all households will yield estimates based on household decisions made in the past. This research designs a method to correct for sample selection in a sample of recent movers. There are large differences in the importance of age and immigrant status and duration as a predictor of homeownership. At the same time, income effects are similar across models.