Cycles of boom and bust in the housing market seem to give support to the 'bubble theory' of the house price where the house price is essentially the sum of the fundamental price and the bubble. We test the theory using the cross-sectional time series data of 50 US states from 1985 onwards. We measure the bubble using the percentage changes in median house prices for the 2004:1 to 2006:2 period as well as the deviation from the long run PIR (price-to-income ratios). We find that states with a larger bubble experience a larger subsequent price downfall. Our data suggests that the house price bubble speculation is quite widespread, but geographically confined to 3-4 geographical regions. We document that the excess liquidity measured by the subprime mortgage and the speculative activity as measured by the variation of house prices explained by the past increases in prices cause the bubble formation. Our data also shows that the states with the greater use of subprime mortgages experienced a greater bubble and a greater subsequent bust suggesting that the disintermediation of the mortgage financing contributed to the most recent housing market bubble.