A Simple and Honest Proposition

Here is a simple and honest proposition:  Data interpretation should lead to conclusions, not the reverse.

If an analyst believes that an indicator is important, and trumpets news about that indicator, it is intellectually dishonest to abandon the measure when it moves the other way.

If one endorses the Baltic Dry Freight index, Dr. Copper, or the inverted yield curve, then one should be willing to change forecasts when those indicators reverse.

Fair enough?

Some of these have recently rebounded, with little attention.  We shall follow up in more detail.

The Application to Housing

Nearly everyone, even those who are not equity investors, is interested in the housing market.  Calculated Risk (a site we feature, along with everyone else!) covers this like a blanket.  There are several key articles today.  Our advice is to read them all.

It is a great job of telling us all what is at stake and where things stand.

Seasonal Adjustments

We are a bit confused by the analysis from The Big Picture, another featured site.  Barry Ritholtz correctly observes that there are important seasonal factors in home sales.  This is, of course, the reason that everyone else uses the seasonally adjusted data to compare one month to another.  Barry maintains [RealMoney article not yet available, but promised] that we should use unadjusted year-over-year data for comparisons.  We do not understand the advantage of this approach.

Everyone agrees that things are much worse than a year ago.  The seasonally adjusted pace of sales was up 2.9% last month and down 2% this month.  The year-over-year was down 19.9% this month, actually a bit better than last month.  Does that tell us something?  Is it an improvement?

If that is the real test, the year-over-year is going to start looking better in September or October, just because last year was so bad.  Will Barry call a turn in housing if year-over-year flattens out?  That would seem to follow from his analysis, even if the month-to-month seasonal data shows a decline.

We still do not see the advantage of this approach over looking at the seasonally adjusted data.  When one spots a big seasonal effect, as Barry demonstrates, doing the adjustment seems routine.  (We note that he criticizes the WSJ article from last month as not recognizing seasonality, even though it employed the seasonally adjusted data, a 2.9% increase.  Had they used the raw data, the increase would have been 12.3%.  (Check out the data for yourself here).

Questions for further Review

What does it mean for a home to be in “inventory?”

We have not yet seen a good answer to this question.  Let us offer a simple comparison that everyone will understand.  We hold various stock positions.  Each day there is a point where we would buy more and a point where we would sell.  We enter stock offers “away from the market.”  Are these offers part of inventory?

Furthermore, if the stock price were to move higher, more stock would be offered.  If it were to move lower, more bids would appear.  The market clearing mechanism involves price discovery, where the market-clearing price is found.  This affects both price and volume.

Applying this to Housing

In our neighborhood, there are people, empty-nesters since this is a kid-friendly town, who are willing to sell at a price that is not realistic.  These homes are part of “inventory” but the offers are not really serious and the sellers are not motivated.

Meanwhile, CNBC reported today that there was a survey as part of the report [UPDATE: here].  18% of homes offered were in foreclosure.  We may assume that these are motivated sellers, as are those who have a job transfer or other personal needs.

Our point is that the concept of “inventory” in existing home sales is a bit elusive.  If prices were to move higher, the “inventory” might increase dramatically.  Things would be worse than we think.  Meanwhile, the existing “inventory” may not really measure the motivated sellers.


We do not have a firm conclusion — only questions.

We would like to see more analysis where economists looked at shifting demand and supply curves and talked about market-clearing prices.  Instead, nearly all market commentators (mostly non-economists) view both supply and demand as black and white.  This does not recognize that a buyer whose credit score does not qualify at one price may be good enough at another. Perhaps that buyer has learned how to apply for credit card with no credit and is building a strong credit score which would be accepted in the near future. This affects the demand curve.

Another question relates to the effect of government programs.  We know that the liberalization of the conforming loan limit on jumbo’s has shifted the demand curve.  How much?

We also know that efforts to help those threatened by foreclosure will shift the demand curve.  How much?

No one is analyzing the problem in this way.  At some point — who knows when? – there will be some bottoming action.  What indicator should we follow to observe this effect?

What is the difference between OFHEO prices and Shiller?  Why?

It will be interesting to see if those who have been the most aggressive in pointing out problems use their methods and indicators that signal the onset of the solutions.

You may also like


  • Bill aka NO DooDahs! April 22, 2008  

    Um, I don’t “feature” Calculated Risk.
    I’ve covered the difference between OFHEO and S&P C-S data several times. Here’s ONE example.
    Both have flaws, but I believe the limited geographic coverage and bias in weighting towards larger homes makes the S&P C-S less reflective of the reality, for the majority of homeowners.

  • David Merkel April 22, 2008  

    My blog crashed. I hope I can get it back up. I want to write more about this, but I want to give you a little now.
    In early 2004, I wrote a piece for RealMoney called “The Fundamentals of Market Tops.” It was a hit. I trotted out a bunch of ways to (sort of) determine a market top using fundamental measures. I concluded that we weren’t at a top (which angered my bearish boss, but that’s another thing).
    In May 2005, I did a piece that was an exact parallel, except I applied it to the residential real estate markets. I said we within a year or so of the top, and in October of 2005, my googlebots confirmed for me that we were at the top. I had set up a bunch of them to track specific phrases, and the tone of the chatter shifted dramatically in the prior month.
    I’ve had a number of people ask me to write a piece called “The Fundamentals of Market Bottoms,” and I may do that. One thing that will be different about the two articles is that in my experience top indicators are different from bottom indicators. If a high value of a given variable is a top indicator, it does not mean that a low value is a bottom indicator. A low value would say “not top,” but that is different than a bottom.
    I plan on setting up a new batch of googlebots for the housing bottom, but they will focus on different phrases than the last set. There will be more pain from hidden inventory, and a lot of false starts before the residential housing market gets going again.
    Oh, and one last note… at the last place I worked, we didn’t care much for the OFHEO data… there is too much of a lag there.

  • Bill aka NO DooDahs! April 23, 2008  

    Once a quarter is too much lag for a geographically comprehensive and properly-weighted index? I believe they’re working on monthly updates now.
    NAII FastTrack or Actuarial Life Expectancy tables, and ISO Loss Cost and Vehicle Symbol data all have far more lag. Did they think much of that data where you used to work?
    Which reminds me, I do think year over year (YOY) is better than month over month (MOM) in terms of reducing seasonality, but YOY has as much noise in it as MOM. When looking at series like NAII FT, we often remove seasonality by looking at summations of year ending quarters (YEQ) data, so that each data point is a full year. You could probably do 12-month moving average on raw monthly data and get a very good seasonal adjustment, as well as smoothing out some randomness.
    Back to the weighting issue. Imagine 11 homes, 10 of which are $100K and 1 which is $1mil.
    Imagine the $1mil home sells for $800K but several of the $100K homes sell for $100K. The S&P C-S weighting registers a 10% drop. The OFHEO weighting registers a 1.8% drop (the WEIGHTING method, note that OFHEO is flawed in missing coverage of higher values). Which is more “accurate?”
    Imagine now that all of the $100K homes sell for $90K, but the $1mil home sells for $1.1 mil. The S&P C-S index weighting registers no change. The OFHEO weighting would register a 9.2% drop. Which is more “accurate?”
    And what kind of innumerate putz calls a 10-city index “nationwide?”

  • VennData April 23, 2008  

    …those know-nothings at the Chicago Mercantile do.
    …and those putzes at S&P do, too
    Most folks live in the major SMSA’s: 205M of the 300M live in the top twenty five SMSA’s
    And speaking of the article, where were you when the CS S&P index was a-risin’, DooDahs?

  • Bill aka NO DooDahs! April 23, 2008  

    Venn, I was in the same place I’ve been since Sept 2006. Before that, I had a blogspot blog (going back to 2005) and another at Nodoodahs.com that got wiped out. I don’t particularly care about housing data as a whole, or the economy, I’m playing the market, and that information just isn’t necessary to play the market. 🙂
    I point out the errors in relying on a flawed metric (S&P C-S) as a FAVOR to those individuals who may be following the fearmongers. Not that OFHEO isn’t flawed as well, missing the top end of the market, but I’m honest about presenting both metrics, their pros and cons, and why I prefer one to the other. Which is more than you can say for the fearmongers, who present C-S as if it were fact.
    I started presenting my real estate analyses as a favor to the retail investor, specifically to counteract the damage being done by the larger pundits. Maybe a few will respond to logic, even though most will be consumed by emotion.
    Hey, where’s Houston on the C-S?
    Why don’t you Wiki Houston’s population size, and see if it doesn’t fit on the 20 largest cities?

  • Bill aka NO DooDahs! April 23, 2008  

    Should have added this, forgot, sorry.
    List the largest 10 metros, and count how many in are in the Putz’s 10-city index. Then do the same for the largest 20 metros. See the flaw?
    Don’t use a trading index for economic analysis. That’s the same stupid mistake people use on the dollar, see my post http://www.billakanodoodahs.com/2007/10/tale-of-two-dollars/
    Now, calculate the percentage of housing units in the 10-city and 20-city metro. Calculate the percentage of housing units covered by the OFHEO, which uses Fannie and Freddie data. Compare. Which is broader?
    All of this is IN ADDITION TO the flawed weighted algorithm in the C-S, which was illuminated above in my first comment, and in a recent post.
    Bottom line: people making economic projections based on S&P C-S are deeply misguided.

  • VennData April 23, 2008  

    Case Shiller’s SMSA total pop’l is over 100 million, and is representative of the top 25 and top 100 (e.g. by including Charlotte.)
    OFHEO includes a smaller slice of the housing market (conforming SFH) and their much narrower skew decreases the percentage change, a priori.
    The example of one $1M house and 10 $100K houses is not representative of the distribution. Here are the LA percentiles:
    …meaning, the CS weighting would not create outcomes using real world examples using the bimodal distribution example(s) of 1 $1M and 10 $100K.
    From the OEFEO site:
    The house price indexes published by OFHEO — hereafter referred to collectively or individually as the “HPI” — are based on a modified version of the weighted-repeat sales (WRS) methodology proposed by Case and Shiller (1989)
    The constructions are different. Rejecting one because Houston’s not in it is not valid reasoning when the other misses a multitude of housing stock formation (besides CS has Dallas.)
    I appreciate your trading using other tools, but to say using CS to make economic projections may be misguided can only be proven using a regression on predicitive models. I’d like to see how it stacks up against OFEHO, NAR, or your local realtor before judging.
    Given its broad nature and realistic coverage, I’d put CS at the top of my list as a tool to understand the change in the price of dwellings.

  • Tim Plaehn April 23, 2008  

    My problem with YoY housing data is that we are still ratcheting down from excess of 2004-2006. The second chart on this article:
    shows more regular sales in the 5-6 million range for several years before the runup. And of those huge numbers in 04-07, how many were the same home flipped several times?

  • Bill aka NO DooDahs! April 23, 2008  

    Venn, your money, your trade. Enjoy.

  • Lord April 23, 2008  

    OFEHO offers broad national coverage but offers poor coverage of expensive markets and is subject to shifts between and within markets. One doesn’t know whether prices have increased or activity shifted to higher priced markets. CS is much more limited in coverage but better as a function of locales covered, especially higher priced markets. Since real estate is local, CS is much more pertinent to buyers and sellers in these markets, while OFEHO is more relevant to middle America. While there are foreclosures in middle America, the economically important ones in magnitude of loss are in metro areas.
    The advantage of YOY is it gives not just a measure of change but the magnitude over time. An increase from say -20.0 to -19.9 would be an improvement, but not much of one. It might represent a stabilization though. A YOY value will lag naturally.

  • RB April 24, 2008  

    In following housing, much of what the bulls said seemed to center around “affordability does not matter” — wanting to buy a home does not equate to a demand for buying one. The one market that I follow regularly is our local Orange County, California market — housing costs as a proportion of income are currently 39% against a long-term average of 32%. There is at least another 22% to correct merely to return to the average at current interest rates, while the chances are that we see a 30% correction producing an overshoot, most likely with a combination of nominal declines and flat home prices for a few years.