Learning about Housing Cost: Survey Evidence from the German House Price Boom

This paper uses new household survey data to study expectation formation during the recent housing boom in Germany. The cross section of forecasts depends on only two household characteristics: location and tenure. The average household in a region responds to local conditions but underpredicts local price growth. Renters make on average higher and hence more accurate forecasts than owners, although their forecasts are more dispersed and their mean squared forecast errors are higher. A quantitative model of learning about housing cost can match these facts. It emphasizes the unique information structure of housing among asset markets: renters who do not own the asset are relatively well informed about its cash flow, since they pay for housing services that owners simply consume. Renters then make more accurate forecasts in a boom driven by an increase in rents and recovery from a financial crisis.