We study empirically and theoretically the patterns of home equity withdrawal among retirees, using a model where retirees are able to own or rent a home, save, and borrow against home equity, in the face of idiosyncratic risks concerning mortality, health, medical expenditures, and household size and observed house price changes. The estimated model is found to replicate successfully the patterns of home ownership and saving/borrowing decisions of retirees. We use the estimated model for several counterfactual experiments. There are three main findings. First, the model predicts that a house price boom suppresses home ownership and increases borrowing, while a future decline in house prices would have the opposite effect. Second, costs of home equity borrowing are restricting the borrowing of retirees, and thus reduction of such costs (e.g. lower costs of reverse mortgage loans) might significantly raise home equity borrowing. Third, there are two implications for the retirement saving puzzle. Although the cost of home equity borrowing affects the borrowing of retirees, it does not affect the total asset holding, implying that equity borrowing costs do not seem to offer a quantitatively significant contribution to resolving the retirement saving puzzle. On the other hand, the magnitude of the retirement saving puzzle might be exaggerated, as a sizable part of "retirement saving" is due to house price appreciation.