As it has proved difficult to explain the recent US house price boom on the basis of fundamentals, many observers have emphasised the role of speculation. This kind of argument is, however, indirect, as speculation is treated as a deviation from a benchmark. Our paper identifies house price expectation shocks directly, using a VAR with sign restrictions. House price expectation shocks are the most important driver of the US house price boom. We also show that a model-based measure of changes in price expectations leads a survey-based measure. Our baseline specification leaves the question of whether expectation shifts are realistic or unrealistic unanswered. In alternative specifications, we provide evidence that expectation shifts during the boom were largely unrealistic.