Which brings me to two points:
1) Economic predictions of housing gyrations have been poor, and they continue to be so.
2) The numbers behind the headlines show something a bit different.
The seasonally adjusted numbers are what always get reported. In the latest report they give actual sales for 2005, as well as annualized numbers from the monthly snapshot for 8 different months. The annualization process has a lot of slop in it, and I have previously urged readers to discount it and just look at the actual numbers. Another way to use the annualized numbers is to put error bars on them. The actual number of homes sold in 2005 was 1.283M. The monthly annualized numbers from May through December vary from 1.236M to 1.367M. That says that the error bar on an annualized number should be at least 0.09M, or 7%.
That says that any variation in housing numbers that is less than 7% is just noise in the annualizing algorithm, and signifies nothing.
Similarly, if one looks at the error bars in the regional numbers, the variations due to the annualizing algorithm are always less than the error bars. All the sturm und drang over the housing market is noise in the annualizing algorithm.
For 2006, the latest annualized is 1.234M houses sold, which really means something between 1.148M and 1.320M, so 2006 home sales could be less than or greater than 2004 and 2005 homes sales.
Without seasonal adjustments, 2006 year-to-date home sales are down 11% from 2005 sales, but that means almost nothing as the summer sales season is not incorporated in that data yet, and month-to-month sales vary significantly due to exogenous (weather, gas prices, world events, etc.) and random factors.