increase in median price - change in distrubution?
Median closing price in King Co continues to climb, despite historically high inventory and historically low affordability numbers. Some have tried to explain this as an issue of a change in the distribution of sales. Perhaps lower priced homes aren't selling as well? Or maybe higher priced homes continue to appreciate at a higher rate...
There are lots of ways we could see an increase in median price while certain segments of the market simultaneously show weakness. The best way to assess this would be too look at the raw data, determine various statistical attributes, and see how these attributes change over time. The ideal situation would be to examine histograms of the raw data month to month over the years of interest, then measure skewness. (skewness is a statistical determination of how asymmetrical a certain distribution is relative to a bell curve.)
The distribution of housing sales is a classic example of positive skewness. Most of the homes sold in our area fall within a certain range, but some of the homes sold in King County are off the charts (Bill Gates has to live somewhere). This is why reporting average sales price doesn't really reflect the middle of the market; the really expensive homes skew the average away from the middle. Thus the focus on median price. because it truly reflects the middle.
Now I have tried all I can to get my hands on the raw data for sales in this area, but thus far I have had no luck. NWMLS reports some stats, and some of it very detailed. But they don't report skewness. You can roughly calculate skewness if you have mean, median, and standard deviation. NWMLS reports mean and median but no standard deviation. So I can't calculate a rough skewness.
At a minimum, though, you can get a sense that a distribution is shifting by comparing the mean to the median. I decided to do that for King Co for as far back as I could find data. Below is a fig that shows the ratio for mean:median over time.

Here is a fig for MOM change in % for the mean:median ratio:

And here is the YOY data. I don't have even a year's worth data, but it's pretty interesting:

YOY you can see that the ratio is greater than 0 for 6 of 8 months listed, and positive numbers indicate an increase in the mean:median ratio YOY.
In the first fig you can see that the mean has been approximately 21% higher than the median up until a few months ago. But the last few months have seen an increase in the ratio, which is especially obvious in the second and third fig. This suggests that the distribution is changing. You would expect to see an increase in the ratio as the distribution becomes more positively skewed. One way in which you might see this kind of change would be if there is an increase in higher priced homes. There are other explanations as well, but the only way to be sure is look at the distribution over time.
So I'd love to get my hands on the raw data if possible. And at a bare minimum I'd like to get mean and median data for King Co going back long before Oct 05, to see if the activity the past few months is just noise or if it's real. That shouldn't be too hard to find, but I don't know where to look. If anyone can help me out with either of these things, please let me know.
I'll be sure to keep an eye on this relationship and post updates.
There are lots of ways we could see an increase in median price while certain segments of the market simultaneously show weakness. The best way to assess this would be too look at the raw data, determine various statistical attributes, and see how these attributes change over time. The ideal situation would be to examine histograms of the raw data month to month over the years of interest, then measure skewness. (skewness is a statistical determination of how asymmetrical a certain distribution is relative to a bell curve.)
The distribution of housing sales is a classic example of positive skewness. Most of the homes sold in our area fall within a certain range, but some of the homes sold in King County are off the charts (Bill Gates has to live somewhere). This is why reporting average sales price doesn't really reflect the middle of the market; the really expensive homes skew the average away from the middle. Thus the focus on median price. because it truly reflects the middle.
Now I have tried all I can to get my hands on the raw data for sales in this area, but thus far I have had no luck. NWMLS reports some stats, and some of it very detailed. But they don't report skewness. You can roughly calculate skewness if you have mean, median, and standard deviation. NWMLS reports mean and median but no standard deviation. So I can't calculate a rough skewness.
At a minimum, though, you can get a sense that a distribution is shifting by comparing the mean to the median. I decided to do that for King Co for as far back as I could find data. Below is a fig that shows the ratio for mean:median over time.

Here is a fig for MOM change in % for the mean:median ratio:

And here is the YOY data. I don't have even a year's worth data, but it's pretty interesting:

YOY you can see that the ratio is greater than 0 for 6 of 8 months listed, and positive numbers indicate an increase in the mean:median ratio YOY.
In the first fig you can see that the mean has been approximately 21% higher than the median up until a few months ago. But the last few months have seen an increase in the ratio, which is especially obvious in the second and third fig. This suggests that the distribution is changing. You would expect to see an increase in the ratio as the distribution becomes more positively skewed. One way in which you might see this kind of change would be if there is an increase in higher priced homes. There are other explanations as well, but the only way to be sure is look at the distribution over time.
So I'd love to get my hands on the raw data if possible. And at a bare minimum I'd like to get mean and median data for King Co going back long before Oct 05, to see if the activity the past few months is just noise or if it's real. That shouldn't be too hard to find, but I don't know where to look. If anyone can help me out with either of these things, please let me know.
I'll be sure to keep an eye on this relationship and post updates.
Comments
Good work. I had a similar thought and posted a slightly different look at this on this thread
http://seattlebubble.com/forum/viewtopi ... ight=#2051
I am hoping to have the county level data for median and average over the weekend. I'll keep you posted if I get it.
D
Here is what I have for the 3 counties I have been tracking for week of 6/15/07 relative to a year ago:
King Co - YOY inventory up 60%
Pierce Co - YOY inventory up 54%
Snohomish Co - YOY inventory up 56%
So why if inventory is up so much does the median continue to climb?
The answer is obvious:
1) Sales in the lower half of the market have tanked as the only people who kept this part of the market up were people who took out suicide loans and infestors.....both of which are no longer part of the game.
2) Sales in the upper half (especially the upper 10%) continue briskly as the equity locusts continue to chomp the high end crops.
3) Result...a completely skewed median that proves once and for all that Seattle is special...if you are a millionaire.
BTW....I also keep track of target areas in KingCo. As expected the biggest increases in inventory are Des Moines, Kent, and Federal Way.
The top/bottom disparity is related directly to what is happening in America...income polarization of the haves and the have nots. Problem is that the have nots will get angry and hungry....
I missed that post! I'm glad to see that your analysis jibes with mine; increases in the mean:median ratio.
Thanks. I'm looking forward to it. I wouldn't think that it would be hard to find just median and mean data going back a long ways, but I can't find anything older than Oct 2005.
I had thought this might be happening. In fact, it was this idea that spurred me to do this analysis. But what I see here, and what DJO showed in the thread he linked to does not bear this out.
If there was a drop off in the lower end market, then you would expect a decrease in the mean:median ratio, not an increase. A decrease in the left tail of the distribution would make the data less skewed, not more.
Just to be clear, there may indeed be a relatively greater decrease in sales in the lower end market, as you suggest. I would have expected as much, given the issue of affordability. But if there is an increase, then it would have to overshadowed by huge appreciation and/or many more sales in the higher end market. This is the only way you could see an increase in this ratio. Perhaps this is happening, but I would need to see the raw data to be sure.
But maybe I'm not thinking about this clearly. Regardless, an increase in the mean:median ratio indicates some fundamental change in the market. I'd love to look at the data from other bubbly markets, especially the Bay area.
My post is based on observation and "gut feel" (digesting many articles/data/etc.)
As an engineer I will always defer to hard empirical data.
That being said...
1) Nice areas (Madison Park, Leschi, Madrona, Kirkland, etc.) continue to list and sell homes (those "sold" signs are still going up pretty fast).
2) The other areas (pretty much everything south of Seattle) have "for sale" signs that sit and sit and sit......
3) Inventory is rising much faster in south king country (I track these weekly)
4) Brokers I talk to say the CA equity locusts are not exhausted yet and there are still a lot of people looking to buy in the high end who don't really care if we are "at the top" or not.
All of which lead me to the conclusion that continue high end sales are skewing the statistics and providing a false positive on the King Co market.
Keep up the data mining. One of these days the "great Seattle crash" will no longer be debateable....it will be historical fact.