We characterize the dynamics of relative house prices, housing sales, construction rates and population growth in response to city-specific income shocks for 106 US cities. We then develop a dynamic search model of the housing market in which construction, the entry of buyers, house prices and sales are endogenously determined in equilibrium. Our theory generates dynamics that are qualitatively consistent with our empirical observations and a version of the economy calibrated to match long-run features of the housing market in U.S. cities offers a substantial quantitative improvement over similar models with no search. In particular, variation in the time it takes to sell a house (i.e. the house's liquidity) induces house values and transaction prices to exhibit momentum, or serially correlated growth.