October 27, 2006
Interpreting median house prices
"Home Prices Plunge by Most in 35 Years", declare the headlines. But those numbers don't mean what you might think.
A number of Econbrowser readers felt I made a mistake in discussing only the quantity data from this week's statistical release from the Census Bureau on new home sales, leaving it to other sources to relate the facts this way:
The median price of a new home plunged in September by the largest amount in more than 35 years, even as the pace of sales rebounded for a second month. The Commerce Department reported that the median price for a new home sold in September was $217,100, a drop of 9.7 percent from September 2005. It was the lowest median price for a new home since September 2004 and the sharpest year-over-year decline since December 1970.
A sharp decline it was indeed. I didn't find a breakdown of the monthly Census figures by region, but they do offer regional breakdowns for the quarterly data which are summarized in the table below. The median price of a U.S. home sold in 2006:Q3 was $232,300, which is a 1.7% decline from the median for 2005:Q3. Using the quarterly data unfortunately smoothes out much of the really dramatic drop that we saw in September in particular, but I think I can still use the quarterly data to call attention to an important point. Specifically, using the quarterly figures, where did that 1.7% drop come from? Certainly not from the Northeast, where prices were up 19%. Nor did it come from the Midwest (up 4.0%), nor the South (up 0.7%), not the West (up 1.6%).
| 2005:Q3 |
| 2006:Q3 |
| percent |
| % of 2005:Q3 |
| % of 2006:Q3 |
Hmmm....There are only four regions in the Census database, and the median price was up, not down, in every single one of them, (a point also noted by my esteemed colleague over at Macroblog, who rarely misses much).
But how, you might ask, can house prices have risen everywhere and yet the national median shows a decline? The answer comes from composition effects. The number of homes sold in the South fell by less than it declined in the West. Thus homes in the South represented 49% of the 2005:Q3 sample and 55% of the 2006:Q3 sample. Because the median home in the South costs just a little over half as much as the median home in the West, any shift in the fraction of recorded home sales that are coming from the South would translate into a reduction in the U.S. national median sales price, even if the price of every single home in America had gone up.
Now, for the record (before I get yet another batch of eagle-eyed Econbrowser readers letting me know the error of my ways), I do not believe that the price of every single home in America has gone up. As I noted above, these quarterly calculations miss much of the very interesting action that characterized the Census September figures in particular. For what it's worth, my guess is that, if we did perform the ideal calculation of taking every single home in America, tracking the September-to-September change in the price of that home, and calculating the median of those numbers, we might well arrive at a negative number. But it is also my opinion that this negative number would be far more mild than the 9.7% drop that results from the far less ideal calculation of finding the median price of all the homes sold one year, and comparing that with the median price of all the homes sold the following year.
As I mentioned in my original post, I definitely agree that the change in home prices is potentially one of the most important factors in determining the dynamics of what comes next. But the problem is, it's extremely difficult to infer too much about that very important magnitude from the numbers that accompanied this week's Census data.
Posted by James Hamilton at October 27, 2006 04:01 PMdigg this | reddit
Listed below are links to weblogs that reference Interpreting median house prices:
» Homes Prices Down--Except They're All Up from Businomics Blog
James Hamilton at Econbrowser has a great post about the median home price report.using the quarterly figures, where did that 1.7% drop come from? Certainly not from the Northeast, where prices were up 19%. Nor did it come from the [Read More]
Tracked on October 28, 2006 01:38 PM
» Housing Price Decline Data from TruePress
The folks at Econbrowser take apart the recent headline “Home Prices Plunge by Most in 35 Years.” As you can see from the chart they provide below, when looked at on a regional basis median housing prices declined nowhere in the country.... [Read More]
Tracked on October 30, 2006 04:55 AM
I like to use the OFHEO's HPI deflated by CPI ex-shelter for real housing prices. It's a pretty laggy indicator, though -- you'll have to wait for December 1 to get 06Q3.
The last twelve quarters of that measure on a year-over-year basis read like an equity index in a bull market: 4%, 6%, 7%, 6%, 10%, 8%, 10%, 11%, 6%, 9%, 9%, 5%, ..Q3?
With for-sale-only vacancies at historic highs, my bet is that residential prices almost have to have started to roll over last quarter. We'll see soon enough.
Posted by: wcw at October 27, 2006 04:37 PM
Agreed 100%, wcw. That's the index I've primarily relied on (e.g., this post), and you're also right that it's main drawback is that it comes in so much later than the other data.
Posted by: JDH at October 27, 2006 06:27 PM
Very interesting point...
Of course, this effect should happen in both directions, and where were the skeptics when fantastic price appreciation was going on? But I concede it is good to understand the composition of such a complex statistic.
I might also point out that it is possible the same effect was in play 35 years ago (I haven't checked), and we know that was a portent of ill things to come, so this is likely to be the case again.
And while we are critiquing the statistics, we might as well point out the lack of inclusion of sales incentives is a huge upward bias.
Anyway, this report did jibe with the evidence I've seen "on the ground", which was nominal price decreases in September in the 6.5% ballpark. In fact the across-the-board regional price increase data borders on the unbelievable.
Posted by: Aaron Krowne at October 27, 2006 07:11 PM
It is also worth noting that, in an important sense, methodology of the Census statistic is perfectly valid: it gives you the median sales price of homes in the US in general for some time interval (in fact I would say this interpretation has primacy). As regions are defined in a purely arbitrary way, one must question the validity of a regional sales price breakdown as an "important" statistic.
In fact, I'll bet you can more-often-than-not construct a regional decomposition of the sales data to the effect of having disjoint regions that all show median price increases, even when the overall change is negative.
Posted by: Aaron Krowne at October 27, 2006 07:18 PM
One interesting thing that came to my mind as I read your posting is that this seems to show strong demographic shifts in population away from urban to rural, or from high cost of living to low cost of living, or from high tax states to low tax states, or from blue states to red states.
Posted by: Dick at October 28, 2006 05:17 AM
Dick: Partisan politics has infected your brain. You look out at the world and only see red and blue. :-)
I don't see what you're seeing at all. Sales have dropped more dramatically in the more expensive parts of the country. Sales have declined less quickly or stagnated in the less expensive parts of the country.
This makes practical sense. As JDH said in his post: the mix of sales is changing. That changes the national median. It says nothing about people moving.
Posted by: T.R. Elliott at October 28, 2006 08:58 AM
T.R., you're projecting. Why do you think the eastern and midwestern industrial states are losing congressional seats while the south and sunbelt are picking up seats?
Posted by: Rich Berger at October 28, 2006 09:18 AM
Rich: Arizona and Nevada have the fastest growing populations in the US.
The national association of realtors reported back in august that
"The Realtors survey showed that the biggest declines occurred in states that had been enjoying red-hot sales during the five-year housing boom. The five biggest declines this spring compared to the April-June period of 2005 were Arizona, down 26.9 percent; Florida, down 26.7 percent; California, down 25.3 percent; Virginia, down 23.9 percent, and Nevada, down 23.5 percent."
Nevada and Phoenix are the fastest growing states and are now experiencing very sharp declines in sales. Therefore, in these two cases, we see the complete opposite of what Dick said.
Posted by: T.R. Elliott at October 28, 2006 10:00 AM
T.R. Elliott - "Nevada and Phoenix are the fastest growing states and are now experiencing very sharp declines in sales."
They're running out of water.
Have you seen Lake Mead lately?
And how long has Phenix been a state, Elliot?
Posted by: x-ray vision at October 29, 2006 10:21 PM
Well, but we expect during a downturn sales to fall the most in regions with the largest declines in prices. People get all fussy about their home and refuse to sell when being offered prices too far below what they expect.
Posted by: Barkley Rosser at October 30, 2006 12:44 AM
Arizona and Nevada have the fastest growing populations in the US.
Thanks for this. It makes my point. A huge portion of the growth of Arizona and Nevada has come from California. The government of California is creating a third world economy there and many productive people are leaving the state. Also, government restrictions make it almost impossible to build new housing in California unless you are one of the top 5% wage earners, so median and lower housing is stagnant. But in both Arizona and Nevada beautiful homes with great vistas are being built that are affordable to the middle class.
Now it is possible that with the influx of people who have lived under central planning state governments that they may create a shift in the politics of their new home states, so I do not know whether this is a good or bad situation.
Posted by: Dick at October 30, 2006 05:39 AM
Dick: The original JDH post highlighted the changing mix of sales. The changing mix of sales does not show a change in demographics. It shows a change in sales which are not correlated with the change in populations. The fact that populations are growing faster in some states does say something, but that is not acquired from the original post. You read that into it from your existing knowledge about population trends. But the original post did not have anything to say on this topic, and in fact said the opposite in the cases of Nevada and Arizona.
Posted by: T.R. Elliott at October 30, 2006 07:12 AM
If I understand, you are implying that housing prices are going down because purchasers are moving into lower priced homes in their own states, that there is no demographic shift. Is this what you are saying?
To me the facts as shown by the professor have to lead to the conclusion that purchasers are actually moving to markets where they can get a better house for either the same or less than they are currently paying. There have to be demographic shifts and as Rich noted the census seems to confirm this.
An anecdote to illustrate this, I have a lot of friends who are retiring and then moving to states like Tennessee and Alabama where they can build a home on a large tract of land for less than the price of their current home in the city.
Posted by: Dick at October 31, 2006 04:39 AM
Dick: There are well known demographic trends in the United States (e.g. California continues to grow, as do other states). That's not what we are necessarily looking at. We're looking at buyers and sellers sitting on the sidelines.
That's all you can say. Home sales are declining just about everywhere. They are declining more rapidly in those markets that were more bubble like. That says nothing about demographics. It probably says a lot more about speculative second home purchases, people moving up, etc.
For the last twenty years at least, a large number of people have been retiring to cheaper locals. California, for example, had a large influx of people at the end of WWII, those people were employed, they purchased homes, they realized large gains, and now they've moved--as my parents and their friends--from expensive locals such as Los Angeles to areas like Oregon in which they can purchase large plots of land, build homes, etc.
This is a long-term trend. What we are seeing right now is a huge drop in sales, with the larger drop in expensive areas bringing down the median national price.
Posted by: T.R. Elliott at October 31, 2006 09:57 AM
Great, informative post! Aggregated data, like national median home prices and national median temperatures, can be useful in talking about long-term macro trends, but aren't necessarily informative about local trends at every level. The Bay Area, where I live, is famous for micro-climates in both weather and real estate -- fog in one neighborhood, sunshine in the next one; prices dropping in one town, and edging higher in the next. Real estate is a distinctly local phenomenon.
Posted by: Kevin Boer at November 5, 2006 10:41 AM
You are right. It doesn't jive. I was working for a homebuilder in the Northeast and we were imagining that we saw fewer and fewer qualifying customers. We also imagined that we were selling homes for less and less and that we were seeing more cancellations.
I'm going to call and get my job back. The're just firing all of the old workers, so they can hire all Mexicans. Do I sound like a crazy person, or is it just you?
Posted by: Barry at November 18, 2006 02:04 PM