Way to take Langley's idea and hit out of the park. Interesting post.

Great graphs. Mohican (who runs langley financial planning) has been doing these scatter plots for a few months. The x intercept at 6 months is strikingly similar for this US data you post and what Mohican found for Vancouver.

Who knows, maybe this is a universal constant.

Interestingly, Vancouver, which is the very last bubble city in the world still holding out, has recently hit just about 6 months of inventory. MOM price changes have slowed, but we'll soon see if they turn negative!

VHB

To various commenters: Stop worrying about the correlation. Why should we expect a linear relationship between these two variables?

The key thing to acknowledge is that there is a relationship of decreasing prices and months of supply. The data is very sparse out in the 8+ months of supply range, so we can't extrapolate the expected price declines very well. We could perhaps draw some 95% confidence intervals but they would be very large due to the lack of data.

However, we can confidently predict large declines in price. Great plot CR!

Since VHB left us, mohican has been doing a bang up job with langley financial blog here in Vancouver B.C.. I'm really glad that he's getting a bit of well deserved recognition for his work. If you read his blog and see what's up with "the bubbliest city in N.A." as R. Shiiller called Vancouver back before he sold out, you will realize why I hang out here so much. I want to know what's coming down the pike at us before it gets here.

How do I read it?

the existing homes Months of Supply hit 11.2 months, and will probably be over 12 months this summer. This suggests nominal price declines of over 5%
With all due respect, I would say "The existing data suggests months of supply does not predict price changes."

"Why should we expect a linear relationship between these two variables?"

We don't expect a linear relationship. This is merely a regression exercise, nothing more. In fact it is likely the case that negative price drops will behave differently than price rises because different effects start taking hold; months' inventory can go extremely high if its denominator, sales, is close to zero but this does not portend -100%+ price drops.

"I would hypothesize that price declines will lag months of supply by a quarter or so."

The strongest correlation is likely based on prolonged high MOI, not necessarily a lag. It's probably true that properties that sit for a few months with low sales will be softer on price and will mark the true state of the market. I'd bet CR's correlation would increase if 3 month moving average of MOI is used.

OT- gas prices

I just got back from 160 mile trip to Hershey, PA. I was great! Minimal traffic. I have made this trip for the last 4 years and this was BY FAR the least traffic I've seen.

It made me have a new appreciation for high gas prices.

Very nice work, CR. I have not seen this graphed before.

Now, the question becomes: what are the predictors of months of inventory?

Me, I like fundamental economic variables, and still am a fan of my three factor model -- defaults, employment, and resale home sales -- for predicting when San Diego resale home prices may trough (see the four graphs):
Free Inflection Point Forecasting Model | Piggington's Econo-Almanac | San Diego Housing Bubble News and Analysis

My model says that low point for home prices is far, far away, as defaults are still rising, employment is still falling, and home sales are still declining.

Rob,CR -
I view the graph as two parts.
The left is a bullseye, the classic mark of "no correlation".
The right (7+ months) is a small set of outliers, not something to draw conclusions from.
Now, the right (7+ months) is all negative, which IS suggestive.
HOWEVER, they are all from the down leg of the cycle. When the RE cycle bottoms (someday), there will be a lot of inventory, right? Right?
Then prices will head up slowly.
So post-bottom, there will be data points with large inventory, and POSITIVE price changes.
So after the cycle completes, the right side of the graph will also look like a bullseye.
So no correlation.
So curve fitting really isn't appropriate, IMO.

CR, great post. Where do I find the data to construct an area-specific analysis of my own? Or, do you know of similar analyses on the San Francisco Bay Area? (I'm a stats person, new to real estate, trying to figure out when to buy)

I agree with jus me. The timing of the downtrend has only started...there is not enough time involved to make a correlation. Btw, my area is going up up and away.

I don't mind gay marriages but I don't like to see them be able to adopt. It's not a natural (normal) environment for the kids to grow up into; It'll be unfair to the child. Handicapped people should not be allowed to adopt either. Anything that is identified as not normal should not be able to adopt. The outliers to the left or right of the curve should not be allowed to adopt. Congrats to the gay people; you shouldn't be singled out for your genetic abnormality...you are still human...just a little unique. So...we male, female, and other - which bathroom should the "other" be allowed to use?

I hope I don't offend most of the people on here.

You hope off-topic bigoted rants aren't offensive?

Hope again.

I'm an engineer and I love stats. But I know when human psychology is a trumping wildcard variable that also is very reactive to other variables. Thus, I think JG's point is key: to model 'predictors' of months of inventory, you have to model homeseller behavior ...even better, model what most influences a homesellers’ behavior. The behavior in question are 1) taking unsold houses off the market out of frustration and returning them out of desperation; 2) Setting selling prices based on the outcome of their emotional/mental battle over what price "I deserve" vs. what price will sell my d#ng house. So what influences such emotional behavior? Well the employment and financial circumstances of homesellers is key, circa 2008. So JG writes "I like fundamental economic variables, and still am a fan of my three factor model -- defaults, employment, and resale home sales"

Add to this the semi-anecdotal evidence that banks are not filing foreclosing docs on all homeowners whose defaults warrant foreclosures, thus depressing REO listings. Such bank behavior is a result of internal battles over bank foreclosure staff shortages and uncertainty over a veto-proof foreclosure bailout from congress, amongst other things. Ignoring these variables is a slippery slope that leads to bizarre concepts like CPI that don’t include items (food & energy) that inflate; and unemployment data that excludes people who don’t have jobs. It makes my ‘inner engineer’ cringe.

Basically the graph shows the rate ofchange of price increase/decrease using the ratio of MOI.

Simplistically, how many straws can the camel hold on his back before the load breaks his back?

And of course during the transition from seller's to buyer's market the market passes through the "balanced market".

CR,

Not meaning to cast any stones, but R^2 values of .54, .60, .66 don't strike me as very 'robust'; particularly to call the root of the equation based on them. Does the data picture clarify at all if you shift it - for example, to account for the lag in reporting prices at closing but inventory as current listings?