Counterfactual scenario 1 (Figure 12) shuts o§ the housing preference shock such that ut = 0 for all t. The RE models now exhibit no signiÖcant run-up in housing value. This result conÖrms the importance of the housing preference shock in ìexplainingî the run-up under rational expectations. In contrast, the MA models still exhibit a sizeable run-up in housing value, particularly in the version with long term mortgage debt which continues to be hit by positive income growth shocks and loosening lending standards. The run-up in housing value now starts earlier than in the data, however. This pattern is driven by the series of persistently positive income growth shocks during the late 1990s (Figure 7). From the top right panel of Figure 8, we see that the housing preference shocks in the MA models are slightly negative from 1995 to 2000ó the period of ìirrational exuberanceî in the NASDAQ stock market index. One interpretation of these results is that the positive income growth shocks of the late 1990s helped fuel a run-up in stock prices rather than house prices. Since there is only one asset price in our model, the only way to delay the rise in housing values until after 2000 is to postulate the existence of small negative shocks to housing demand during the late 1990só a period when investors were paying more attention to the stock market than the housing market