September 18, 2006

The Numbers Games in Rex Nutting’s September 7 Report on the Housing Market

Nutting’s September 7 MarketWatch article is here (link requires free registration).

Even if your eyes glaze over at the mere mention of numbers, I encourage you to stay tuned for just the next nine paragraphs, as you’ll learn a lot, and pretty quickly.

The following sentences in Rex Nutting’s September 7 report on the direction of home prices initially caught my attention (eighth paragraph; “OFHEO” is the Office for Federal Housing Enterprise Oversight):

The average gain in home prices predicted by the economists in the survey was 0.4%. The OFHEO index has never shown an annual decline in its 30-year history. The smallest gain ever was 1.3% in 1991. See full story on the latest OFHEO release.

If you go to that link from September 5 (also requires free registration), you will see the following relative to annual home-price increases:

Home prices are up 10.6% (Nutting meant 10.06% — Ed.) in the past year, OFHEO said, compared with a 12.8% gain in the first quarter. It’s the fastest deceleration in the index in its three-decade history, OFHEO said.

If you really hate numbers and want to stop soon, I’ll give away most of the ending and guide you to exactly what you need to look at on the chart below:

  • The September 5 report uses numbers (with the red box around them) that come directly from OFHEO data, while the 1.3% for 1991 noted in the September 7 report (with the green box around it, rounded) uses an averaging formula that, while it is apparently commonly used, is inaccurate (i.e., it doesn’t describe actual calendar-year results, because it partially drags a portion of prior-year results into the calculation).
  • The fact that the September 5 article cites quarterly results while the September 7 deals with calendar-year results should be irrelevant.
  • Usually the differences between the correct (September 5) method and the incorrect (September 7) method are so small that they don’t matter much. But, as you see from the 2006 items below, if home prices don’t go up at all in the 3rd and 4th quarters of this year, the correct answer to what 2006 home price growth was will be 3.40% (blue box), while the incorrect answer favored by Nutting and some others will be 8.2% (purple box).
  • The economists quoted in the September 7 article came up with their consensus estimate of 0.40% based on the column where the blue and red boxes are (true 4-quarter appreciation), but Nutting compared their predictions to the “Annual Average,” where the purple and green boxes are. It’s not an apples-to-apples comparison, and he changed the measurement standard in the middle of the paragraph. It significantly affects the impact of what he reports, because the economists’ prediction of 0.40% home-price growth for 2007 is slightly above 1990′s full year appreciation of 0.26% (highlighted in yellow), and NOT below what Nutting said was the lowest year ever (1991′s 1.3%). This means that 2007, if it turns out as predicted, will NOT be the worst ever for home-price appreciation, as one would infer from Nutting’s report as it is currently written.

If you understand things at this point and don’t want to go further, see ya — no hard feelings. For those who want or need more detail, read on.

Nutting’s 1.3%

The averaging method Nutting used to report 1991′s home-price inflation doesn’t reflect reality, because it including the following items in the calculation:

  • 0.62%, the 1991Q1 inflation, representing the period from the 2nd quarter of 1990 to the 1st quarter of 1991.
  • 1.10%, the 1991Q2 inflation, representing the period from the 3rd quarter of 1990 to the 2nd quarter of 1991.
  • 0.72%, the 1991Q3 inflation, representing the period from the 4th quarter of 1990 to the 3rd quarter of 1991.
  • 2.59%, the 1991Q4 inflation, representing the period from the 1st quarter of 1991 to the 4th quarter of 1991.

The average of the four quarterly items is 1.26%, and was rounded by Nutting to the 1.3% used in his report.

But 2.59% was the real inflation in 1991. A house selling for $100,000 at the beginning of 1991 was selling for $102,590 at the end of the year, NOT $101,300. The reason for the difference is that, as you can see from above, the averaging calculation drags in results from 1990, which was a year with lower and actually negative home-price growth.

A Method Used by Others? Apparently, But They’re Still Wrong

Rex has correctly pointed out that at a Senate hearing last week (go to “Tables and Charts” Page 8 of Richard Brown’s testimony at this link; Brown’s testimony is a PDF; graph is actually at Page 26 of the PDF), an OFHEO official used a graph that employed the averaging method Rex used to calculate his 1.30% for 1991, and not the actual 4-quarter results as reported by OFHEO.

Fine. I want to see this official come into a hearing in early 2007 and, assuming that home prices didn’t change at all in the final two quarters of 2006, bring in a graph showing that 2006′s home-price inflation was 8.20% (purple box above) and not 3.40% (blue box above). It’s not, going, to, happen. Once the difference between the two methods becomes obvious, they’ll have to move to the correct actual results method. Until then, it seems that the phrase “close enough for government work” would apply.

Back to the September 7 Article

But let’s go back to the first sentence of the first excerpt above, namely “The average gain in (2007) home prices predicted by the economists in the survey was 0.4%.”

I can guaran-flipping-tee you that the economists quoted were estimating how they thought prices on December 31, 2007 would compare to prices on January 1, 2007, i.e., they were using the column with the red and blue boxes. They were NOT employing some kind of averaging method analogous to what Rex Nutting used, or that the OFHEO used at the Senate hearing last week. In other words, Nutting changed the measurement standard (from the economists to OFHEO “average”) in mid-paragraph!

Is it important? I should say so. Using the measurement method consistent with the economists, it would still be true that “The OFHEO index has never shown an annual decline in its 30-year history,” — but just barely so. 1990′s true calendar-year home-price increase was 0.26% (highlighted in yellow above). That is slightly less than the economist’s consensus, as opposed to almost 1% above it. Nutting would not have been able to implicitly predict that 2007 will be by far the worst year ever for home-price inflation (comparing 0.4% to 1.3%), because 1990′s real result of 0.26% is even worse than what economists are predicting for 2007. The paragraph, if written based on the correct statistics, would have lost most of its gloom-and-doom pop.

Correction Not Coming

Rex Nutting has already indicated that he thinks no correction is necessary to the item discussed here. Based on the obvious mid-paragraph inconsistency, I insist that the paragraph in question as written is just, plain, wrong.

Regardless of whether or not a correction is made, here’s the important takeaway — If the calculation method discussed here is typical of what goes into reports on home-price changes, it’s going to be difficult take any of those reports at face value without a lot of further digging.

And what other business reports are making what I believe should objectively be seen as errors like the one I’ve just reviewed here?

To have to tell readers, as I am doing, that you’ll have to do basic research and perform independent calculations before you can trust a business news report from a mainstream business press outlet is disappointing, and disconcerting.


ADDENDUM 1 (modified from original at 11:00 AM): I also believe that This paragraph (the fifth one) is incorrect as written in that same September 7 article was originally disputed by me:

Existing-home prices have risen at an average of 9.6% annually in the past four years, well ahead of the inflation rate. New-home prices rose 13.3% in 2004 and 9% in 2005.

From OFHEO data, which I believed was the frame of reference, I could not determine how the first number was calculated, and I believed that 2004 and 2005 were reversed. Rex Nutting responded with support from different sources that I would assume is correct. I will look at it later and if there is a problem I’ll note it (otherwise, you can assume there is none). I believe Nutting should have indicated where the information came from, but the information itself is correct. ANSWER: Rex’s e-mail reply indicated that he used data from Haver (a company whose work was used in a previous post), and I don’t have any reason to doubt that he reported the Haver data correctly. But he didn’t tell us where the data came from. The realtor data discussed in previous paragraphs is from their “monthly economic outlook,” which I believe is a press release. If it’s the same as the Haver information, it wasn’t identified as such. I believe that an average reader of the article would have assumed that the excerpted paragraph above came from OFHEO, or perhaps from some unidentified place. OFHEO was the only specific data source cited in the article.

ADDENDUM 2: There were another oddity in the September 7 article that I could not explain (actually there others besides this one, but I’m worn out). For some reason, Nutting looked at an unusual time period for this sentence:

From 1968 through 2000, median sales prices rose about 6.2% annually, while the consumer price index rose at a 5.1% annual rate.

Why would that period have been selected, when data is available through 2005? Is it to tell us what happened before the runup began? If so, why not say that? For what it’s worth, consumer prices rose by only 4.76% from 1973-2005, and I believe (but don’t know, because I don’t know where the data from the 1970s is) that 1973-2005 home-price appreciation would be greater than 1968-2000′s 6.2%.


No Comments

No comments yet.

RSS feed for comments on this post.

Sorry, the comment form is closed at this time.