Home Sales Data: Lots of Huffing and Puffing

There's the key: the margin of error in both data points is greater than the estimated increase in new home sales. An increase of 13% in October with an error of +/- 17.7% means that the number has no statistical significance; the Census Bureau economist I spoke to agreed with this interpretation. So too for the 9% rise in September new-home starts, which also had a margin of error greater than the data: +/- 18.2%.

But for the moment, let's ignore that statistical aberration. Let's assume that the statistical error is not fatal. When looking at the historical monthly new-home sales numbers from Census, we see some interesting items.

The highest recorded month-to-month increase in new-home sales measured by the Census Bureau was in April 1963 -- 31.2%. The second-biggest increase was 26.8% in May 1980, and the third-biggest was 26.43% in July 1963. The biggest drops were February 1963 (-21.49%), April 1980 (-21.94%) and January 1994 (-3.77%).

So our most recent number from October was the 23rd highest out of 514 data points. While that's roughly in the top 4.7%, it's not so outrageous as to be false prima facie. And according to the numerologists who work with this data, revisions to these double-digit numbers tend to soften them a bit, but they usually stay somewhat robust after all the data are in. So while October's report may be revised down somewhat, I don't expect it to be cut very aggressively.

Lastly, looking back over the past 15 years of data, we see that a regression to the mean has followed nearly all double-digit monthly gains. The subsequent month's data were significantly lower -- flat to negative -- in nearly every case.

Months With 10%+ Gains in New Home Sales Months with high increases in new home sales have historically been followed by flat to negative months.
Jun-03	10.70%
Jul-03	(-2.1%)
Dec-00	11.70%
Jan-01	(-4.8%)
Jul-00	11.90%
Aug-00	(-4.4%)
Nov-98	11.40%
Dec-98	(-4.6%)
Jan-98	10%
Feb-98	(-0.7%)
Mar-95	10.2%,
Apr-95	0.8%.
Feb-94	10.82%
Mar-94	8.89%
Apr-93	16.45%
May-93	(-10.70%)
Sep-93	12.56%
Oct-93	(-3.03%)
Jan-92	21.15%
Feb-92	(-5.47%)
Mar-92	(-13.30%)
Source: U.S. Census Bureau

In each of these months, the subsequent month's data were significantly lower. The one exception was a strong March 1994 following a double-digit gain in February 1994. Mean reversion fans should note that this exception followed the all-time worst drop in new-home sales in January 1994: -23.77%. So with the positive two-month February/March 1994 combo on top of that awful January, our mean reversion thesis remains intact.

---

Barry Ritholtz is chief market strategist for Maxim Group, where his research and market analysis are used by the firm's portfolio managers and clients in the U.S., Europe and Japan. He also publishes The Big Picture, his macro perspectives on the economy and geopolitics, entertainment and technology industries, and is a member of the board of directors of Burst.com, a streaming media software company. At the time of publication, Ritholtz had no position in any securities mentioned in this column, although holdings can change at any time. Under no circumstances does the information in this column represent a recommendation to buy or sell stocks. Ritholtz appreciates your feedback; click here to send him an email.