“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.