The U.S. housing market exhibits an unusually high degree of volatility, which challenges traditional models that rely on large preference shocks to explain such fluctuations. In this paper, I argue that the expectation channel plays a key role in driving this volatility. I incorporate Diagnostic Expectations (DE) within a Two-Agent New Keynesian (TANK) model featuring housing and banking sectors. Using Sequential Monte Carlo methods to estimate the model, I find that DE reduce the size of the housing preference shock by more than one-third relative to Rational Expectations, while reproducing the housing market fluctuations. This result holds whether agents’ imperfect memory is based on recent or three-year past experiences. When the expectations channel is removed -i.e., when agents are rational- the model fails to generate the high volatility in house prices observed in the data. These findings highlight the importance of the expectation formation process in explaining a substantial part of unmodeled disturbances affecting the housing market and in shaping policy responses.