Okay, the bad news is that we're going to talk a lot about statistics.  The good news is that we're also going to talk a little bit about Derek Jeter. 

Last week, NAR came out with its monthly Existing Home Sales Report, which projected a seasonally-adjusted annual rate of 3.83 million sales, down 27.2% from last month and 25.5% from last year.  The report garnered its normal share of media attention, which was typically hysterical and superficial. I happened to be on the "Stuart Varney & Co." show on Fox Business to talk about the numbers as they came out (filling in for my brother Greg, who is a regular), and the people on the show were understandably ready to pull out nooses so they could hang themselves.

Here's the thing: the numbers are nonsense.  As much as I admire and respect NAR, and am a proud member, NAR is basically committing malpractice when it puts out a report using the flawed methodology they use at a time when the market is in flux, particularly since we all know that the media and industry observers are going to read too much into the reports.

If you can't read all the way to the end, then just know this: home sales across the country are actually up year to date by about 6.7%.  Sales in the past 12 months (i.e., August 2009 through July 2010) are up 10.7%.  The real estate market is fine, or at least the housing market is no worse than the general economic conditions in the country.

Okay, with that said, let's talk about why the NAR Existing Home Sales Report is a dangerously misleading pile of statistical nonsense:

The NAR Existing Home Sales Report Methodology
The problem with the report is simple: it is biased by a limited sample size consisting of one month's worth of home sales, which are then extrapolated for a 12-month projection. I'm not particularly fond of that methodology, since I think that one month's worth of numbers is not a sufficient sample size to project a full year ahead, but even if the methodology was generally sound it was folly to apply it blindly to the July 2010 sales figures.

The sample size problem is clearly demonstrated in July.  July's numbers were weak in part because of a hangover from the Home Buyer Tax Credit that expired in April.  Lots of people who would normally buy in the summer accelerated their plans to take advantage of the credit. Combined sales in April, May, and June, representing tax credit transactions, were 1,604,000, which was an increase of 16.0% from the same three months in 2009. Then sales in July were 389,000, which was down 26.8% from July of last year.  But if you take the last four months all together, we get 1,992,000 sales from April-July 2010, and 1,913,000 in April-July 2009.   Basically the same numbers over the past four months, just distributed unevenly.

Then NAR compounds the problem by seasonally adjusting the data to account for the fact that real estate sales tend to go up in the warmer months and down in the colder months.  So they don't just take the July sales and multiply by 12.  Essentially, what they do is look at the average sales in July (or whatever month) as a percentage of the sales over a given year, and then assume that if July is down X%, then every month in the next year will also be down that same X%.

Again, I agree with the idea of correcting for seasonality, but I don't know that you can apply a seasonal adjustment without considering the context. It makes no sense to apply seasonality blindly when we have an unrepresentative month like July 2010.  This July is much, much weaker than last July because we now have a hangover from the tax credit, and last July we actually had a tax credit in effect. So to think that July's numbers are representative, and then to use traditional July seasonality to project a year in advance, is foolish.

And it's not just this month.  We had the same problem going in the other direction in the past few months, with NAR over-estimating annual sales based on an unrepresentative and tax-credit-bolstered May and June.  For example, the April existing home sales report projected annual transactions of 5.79 million, May was 5.66 million, and June was 5.26 million.  So we literally went from a projection of 5.26 million in one month to 3.83 million the next month. 

How Derek Jeter Figures Into This
Okay, I promised we'd talk about Derek Jeter, of my beloved Yankees. My point is that NAR's methodology, extrapolating a year's worth of sales from a one-month sample, and then seasonally adjusting that sample, is like trying to project Derek Jeter's seasonal batting average from a one-month sample.

So let's say that we were going to predict Jeter's batting average for the 2009 season based on his performance in April 2009, the first full month of the season. At the end of April 2009, we'd see that he was batting .287.  Now, given that in April 2009 he had played 21 games, about 13% of the season, we could be confident that we had a pretty good sample size (certainly better than what NAR gets from one month out of 12, since 1/12 is 8.3%).  

But rather than just simply extrapolate that his 2009 batting average would be .287, the same as his April sample, we'd seasonally adjust the numbers to account for the fact that he might be a particularly good or bad hitter early in the season.  In fact, he has a career .306 batting average for April, which means that April 2009 represented a poor omen for Jeter's season: after all, his April average was 19% below his career average.  So if we did a simple seasonal adjustment, we'd say that his average for the season would be .254, which is 19% below his career seasonal batting average of .314.

Essentially, we'd be predicting that Derek Jeter would hit .254 for the 2009 season, well below his career .314 average, because he hit only .287 in April, a month where he normally hits .306.  

And we'd be way, way wrong.  Because contrary to the indications from the small sample size, and defying normal seasonality, Derek Jeter hit .334 in 2009, and led the mighty Yankees to a World Series Victory.

That's why I hate the NAR report.  Because it takes too small a sample size, extrapolates too much into it, and blindly uses seasonality without putting the data into context.

NAR's Projections Over Time.
I did some quick analysis of the NAR Existing Home Sales reports for the past few years to see how close the forecasts have been.  And they're not that close, particularly when the market is in a state of flux. (I have put my spreadsheet here in case you want to check my math.

I looked at 29 forecasts from March 2007 (the earliest I could find data) through July 2009 (the last month that we have 12 months of closed transactions to check against the projection):

• Out of the 29 projections, NAR was off by more than 10% in its projection 12 times, which is 41.3%.  So NAR was off by a LOT over 40% of the time.
• Of those 12 poor projections, NAR was off six times on the upside (estimating that the market would be stronger than it was) and six times on the downside (estimating that the market would be worse than it was. So NAR wasn't hyping the market: it was unduly pessimistic just as much as it was unduly optimistic.

To NAR's credit, it was within 4% 12 times, which is not unimpressive.  But what you see when you look at the reports as a whole is that NAR's predictions do just fine fine when the market is in a stable environment, but any time that the market undergoes a shift, NAR's projections are dramatically off.

Here's what I mean.  Starting in March 2007, as the market started to decline after years of a seller's market, NAR was off significantly for six straight months: 13.5%, 11.8%, 13.8%, 12.6%, 14.2%, and 10.9%.  Essentially, by extrapolating from the months comprising the last gasp of the seller's market, NAR significantly overestimated sales over the next 12 months.  The projections started to get closer by the end of 2007, as the monthly sales started to stabilize.

So starting in September 2007, NAR's projection for sales in the next 12 months was within 4% of the actual sales in 11 of the next 12 monthly existing home sales reports. Why?  Because the market starting in September 2007 settled into a relatively stable market for about a year. 

Then in September 2008, we had the financial crisis precipitated by the Lehman collapse, which caused a significant slowdown in the real estate market, and NAR's numbers got wacky again.  Starting in November 2008 through June 2009, NAR's monthly projection for sales in the next 12 months was too low by the following percentages: 11.1%, 8.1%, 13.2%, 9.3%, 13.6%, 13.3%, 13.5%, and 10.9%. Basically, NAR was reading too much into the weak sales figures in the six month's following the financial crisis, making the market seem worse than it ultimately proved to be.

So essentially we have a report that makes relatively good predictions when the market is completely stable, but ends up being ridiculously inaccurate when the market is in any state of flux.  But it's exactly when the market is in a state of flux that people look to the leaders in the industry for some dependable analysis of where the market is going. Don't analyze he robust April numbers and inspire us to sing a happy tune, and then analyze the weak July numbers and tell us that we should start burying gold coins in the back yard.

Everyone reads the NAR Report, and everyone reads too much into it. So this week we now have a media meme that the real estate market is in more trouble than it really is, scaring buyers out of the market. We're even seeing reports that the government might resurrect the Home Buyer Tax Credit, since the market is so weak.

Now, I love the Home Buyer Tax Credit. I created a whole website around the tax credit, and a blog, and I'm proud to know the tax credit better than just about anyone in the country. I love me the tax credit, is what I'm trying to say. But even I think it's nuts to discuss bringing back the tax credit just because of the July numbers. 

Think about how crazy that is:

• The tax credit distorts the market by pulling summer transactions into the spring.
• As a result, sales in the spring and early summer are huge.
• Because lots of buyers accelerate their home purchases, the July sales are lower.
• The slow July numbers, when extrapolated and seasonally adjusted by NAR, create an impression that the housing market is in free fall
• So the government thinks about creating a new tax credit to start the cycle again.....

Isn't that just crazy nuts loony?  We might actually see a billion-dollar change in government policy based on a NAR analysis that takes one unrepresentative month's worth of data and predicts a year's worth of misery.

Again, I'm a proud member of NAR, and the research they do is extraordinarily impressive.  But the NAR Existing Home Sales report is dangerous nonsense any time the market is shifting.  It just amplifies the results in a relatively small sample size of data, and moreover exacerbates the prevailing national mood: when we're feeling confident, it makes us overconfident, and when we're concerned, it makes us panicky.

The Real Numbers.
So let's look at hard numbers.  No projections, no seasonal adjustment, but using the real numbers from the NAR report:

• In the last rolling year ending in July 2010 (i.e., the last twelve full months), we've seen 5.35 million home sales.  In the rolling year ending in July 2009, we saw 4.83 million home sales.  That's an increase in the past rolling year of 10.7% year-on-year.
• So far year to date, we've seen 2,996,000 home sales.  Last year at this time, we had 2,807,000 home sales.  That's an increase of 6.7% over last year.

Those are real numbers.  Activity in the market so far this year is up 6.7% from last year.  And over the past 12 months, we've seen an increase in activity of about 10.7% from the prior 12 months.

Does that mean we're in recovery? No, not really.  We still need to see what the market looks like after we get through this tax credit hangover. My read on the market is that the biggest challenge we have right now is economic uncertainty.  Prices have come down, rates are amazingly low, and inventory levels provide a lot of choice, so the housing market fundamentals are poised to drive increasing levels of demand.  But people generally don't start looking for a new home if they're worried about their job or their salary, so we might need to see better economic news before we see a full housing market recovery.

But I would be happy to wager whatever you want that in the next 12 months we'll see at least 10% more sales than the 3.8 million projected in the NAR Existing Home Sales Report for July.