Evidence on loss aversion and the endowment effect suggests that people evaluate outcomes with respect to a reference point. Yet little is known about what determines reference points. We conduct two experiments that show that reference points are determined by expectations. In the first experiment, we endow subjects with an item and randomize the probability they will be allowed to trade it for an alternative. Subjects that are less likely to be able to trade are more likely to choose to keep their item, as predicted when reference points are expectation-based, but not when reference points are determined by the status quo or when preferences are reference-independent. In the second experiment, we randomly assign subjects a high or low probability of obtaining an item for free and elicit their willingness-to-accept for it. Being in the high probability treatment increases valuation of the item by 20-30%.