Forecasting Model — Not!

[sdrealtor]

It was clear to most of us that the party changed in July 2004. Prior to that anything would sell quickly with multiple offers. However, the time when this was happening was concentrated between October 2003 and May 2004. I think alot of it was emotional and sheer lack of inventory that made things so crazy. People were desparate to find a home and when inventory was in the 3,000 it was nearly impossible to get one. At least one that you wanted.

As I’ve said many times b4. i dont think complex models will be needed. I think it will be pretty obvious to anyone who has been following things fairly closely.

[davelj]

While multi-variate regression is a great tool, in many instances often simpler ocular regression does just fine. Ocular regression suggests that the trendline in median price in SD housing will be re-establised about 30%-35% below the peak. I think Occam’s Razor needs to be applied to some of this analysis.

[powayseller]

So the “ask a professional realtor” model seems to be the most time sensitive. sdrealtor has been saying all along that the market turned in spring/summer 04. I didn’t believe him at first, and thought it was in August 05 (per piggington data), which is the time Campbell gave his sell signal. So sdrealtor was one year ahead of Campbell’s model and the piggington data.

A quantifiable model, while lagging one year to the “ask the realtor” model, would remove the personal factor. But which model would you trust?

Remember that Campbell told people to sell in 12/01. If you had listened, you could very well have been priced out 13 months later when his model changed its mind and told you to get back in.

Following Campbell’s model would have resulted in you missing the biggest real estate boom in California history. If you sold in 12/01, or just didn’t buy because of the Sell signal, you could very well have been priced out when he issued the Buy signal in 1/03.

I am working on getting data for a model, but it may not work either. I tried some things with a national model, and it was a dud. So in the meantime, the “ask the realtor” model is the most reliable. sdrealtor will know when the market turns long before any quantifiable model will tell us. So will a pickup in sales, a decrease in foreclosures, less seller incentives.

Now that the internet makes data so easily accessible, we don’t have to rely on these models as much. Back in 1990, how would anybody know the market turned without a model of some sort? There was no blog where you could ask a realtor.

[Author]

Posts