Demand and Must-Sell Supply

Submitted by Rich Toscano on July 20, 2007 - 4:31pm

Back in April, I put up a long-term chart of the number of monthly home sales divided by the number of Notices of Default (NODs) in that same month. A summary of the reasoning behind the chart, explained at length in the original article, goes like this: price declines take place when there is a lot of "must-sell" inventory, meaning that sellers can't wait around for the price they want but rather have to take whatever price they can get. The number of defaulting homeowners can provide an admittedly rough indication of how much must-sell inventory is or soon will be on the market. Comparing sales volume to the number of NODs can tell us how much demand is likely to offset the potential must-sell supply, thus providing a clue about the direction of pricing pressure in the months ahead.

read more at voiceofsandiego.org

(category: )

Submitted by John from Doom on July 21, 2007 - 3:51am.

I think you intended "sellers" not "buyers" in the second sentence.

Submitted by admin on July 21, 2007 - 9:33am.

Good catch John -- it is fixed here and soon will be at Voice. Thank you.

Submitted by jhamon on July 21, 2007 - 9:39am.

I wonder, Rich, if it would be easier to look at the entire data series if it were semi-logarithmic. For example, it's pretty hard to appreciate the price index pullback in the 90s, but on a percentage basis, it was quite significant.

Submitted by admin on July 21, 2007 - 10:11am.

Couple problems there. Excel doesn't handle log charts well, especially with dual axes. And even if we could log the price, it's not appropriate to log the sales-per-nod because the raw number is what matters, not the rate of change like with prices.

This chart is really "about" the sales per nod... it's not intended to give an idea of the scale of the price decline, just to show that nominal prices have generally fallen while sales per nod was low and risen when it was not. For a price focused chart this would be a good start:

http://piggington.com/historical_home_pr...

Submitted by sdduuuude on July 22, 2007 - 9:41am.

jhamon - I agree.

I was going to post the same thing
(with the Schiller index remaining on a linear scale)

The graph will lose its punch if the picture gets worse and there end up being 2 or 3 NODS for each sale instead of the other way around. i.e. if it gets 3x worse, the graph will only show a small movement.

Submitted by Fearful on July 22, 2007 - 11:46am.

I am with Rich on this one. Unless there is some reason to think there is an exponential relationship, there is not much point to looking at the chart on a log scale. If we just want to look at the lower value detail, just omit the time periods with the larger values.

Rich, I am puzzled. I do not see much historical relationship between two sets of data, though it seems like there ought to be.

It actually looks like the Sales/NOD is correlated to the derivative of the price. High Sales/NOD => rising prices. What is the level at which prices are flat, and below that, prices will fall?

Submitted by pgsandiego on July 23, 2007 - 4:46pm.

Are the NOD/NOT comparisons to the last downturns skewed as "most" recent homeowners have first and second mortgages?

Basically, does today's homeowner defaulting on two mortgages instead of one get counted twice?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.