We examine the pattern of price depreciation in Japanese land values subsequent to the 1990 stock market crash. While all land values fell heavily, the data indicate that Japanese commercial land values fell much more quickly than residential land values. Using an error-correction specification, we confirm that Japanese commercial land prices exhibited faster convergence to steady state values than residential land prices. We then develop an overlapping-generations model with two-sided matching and search to explain this disparity. In the model, old agents own real estate and are matched each period with a young agent endowed with an unverifiable idiosyncratic service value for the old agent’s real estate. When fundamentals decline, the old agents optimally "fish" for high service flow young agents by pricing above average valuation levels. This leads to higher illiquidity and default in times of price decline, as well as price persistence which is increasing in the variance of average service flows. As we would posit that the variance of service flows would be higher for residential real estate than for the commercial real estate market, this model matches the Japanese experience.