Using proprietary micro data on the complete set of housing transactions between 1993 and 2009 in 99 metropolitan areas, we investigate whether contagion was an important factor in the last housing cycle. We define contagion as the price correlation between two different housing markets following a shock to one market that is above and beyond that which can be justified by common aggregate trends. Our estimates deal with the following empirical challenges: (a) defining the timing of local housing booms in a non-ad hoc way; (b) addressing specification search bias that arises when only one aggregate series is used to estimate both the timing of the housing boom and the magnitude of price volatility during that period; and (c) controlling for common variation in economic conditions. We find strong evidence of contagion during the housing boom, but not during the bust. These effects appear to arise mostly from the closest neighboring metropolitan area, with the price elasticity ranging from 0.10 to 0.27. This is large enough to account for up to 30% of the jump in prices at the beginning of local booms, on average. Estimated elasticities are greater when transmitted from a larger to a smaller market, and also more important for the most elastically-supplied markets. Finally, local fundamentals and expectations of future fundamentals have very limited ability to account for our estimated effect, suggesting a potential role for non-rational forces.