Predicting home prices

Housing prices in the US rose by about 50% in real terms over the four year period from February 2002-February 2006. In two brilliant posts, Edward Glaeser brings out the problems in explaining housing bubbles and predicting its trajectory.

In a model similar to the "technical analysis" of stocks, Joseph Gyourko and Edward Glaeser find key statistical regularities in housing price changes — long-term reversion to the mean and short-term momentum. They write,

"If an area’s prices go up by an extra dollar over one five-year period, then that area’s prices on average drop by 32 cents over the next five years. This pattern is long-term mean reversion... But at higher frequencies, like one year (or one month), momentum is the rule. If prices went up by an extra dollar during one year, those prices rise by 71 cents, on average, during the next year. One possible explanation for this fact is that people may base their expectations about future price growth on the price growth during the recent past.

Short-term momentum and long-term mean reversion together can produce cycles. At the start of the cycle an initial positive jolt then generates more growth because of momentum. Prices rise until the point where long-term mean reversion becomes powerful enough to cause prices to drop, and then momentum keeps the price drop going."