While aggregate house price indices display time persistence, less is known about micro persistence. This article proposes that absence of micro persistence implies that an excessively high or low sell price in one transaction is not repeated in the next transaction. We exploit a unique Norwegian data set of publically registered housing transactions between 2002 and 2014 and follow housing units over time to see if excessive prices persist or revert. In a regression with time and unit-fixed effects of sell-price-to-predicted-price ratios on previous sell-price-to-predicted-price ratios, we reject persistence and find substantial reversion. We also test for possible arbitrage opportunities in the form of excess returns. Once we control for price increases that are due to home improvements, we document that there is little scope for profitable arbitrage in excess of the market return. The overall impression is that the Norwegian housing market is relatively micro efficient.