Recently much progress has been made in developing optimal portfolio choice models accomodating time-varying opportunity sets, but unless investors are unreasonably risk averse, optimal holdings include unreasonably large equity positions. One reason is that most studies assume investors behave as expected utility maximizers with power utility. In this article, we provide a formal treatment of both static and dynamic portfolio choice using the Disappointment Aversion preferences of Gul (1991). While different from the Kahneman-Tversky (1979) loss aversion utility, these preferences imply asymmetric aversion to gains versus losses and are consistent with the tendency of some people to like lottery-type gambles but dislike stock in-vestments. By calibrating a number of data generating processes to actual US data on stock and bond returns, we find very reasonable portfolios for moderately disappointment averse investors with utility functions exhibiting low curvature. Disappointment aversion preferences affect intertemporal hedging demands and the state dependence of asset allocation in such a way as to not be replicable by standard expected utility functions with higher curvature. Furthermore, it is easy to reconcile the large equity premium observed in the data with disappointment aversion utility of low curvature and reasonable disappointment aversion.