Case-Shiller Tiers: Price Declines Slow Slightly in High and Low Tiers

Let’s check out the three price tiers for the Seattle area, as measured by Case-Shiller. Remember, Case-Shiller’s “Seattle” data is based on single-family home repeat sales in King, Pierce, and Snohomish counties.

First up is the straight graph of the index from January 2000 through February 2009.

Case-Shiller Tiered Index - Seattle

All three tiers dropped again in February, but in a reversal from last month, the high tier fell the slowest, dropping just 0.6% month-to-month. The low tier has rewound to April 2005, the middle tier to June 2005, and the high tier to June 2005.

Here’s a chart of the year-over-year change in the index from January 2003 through February 2009.

Case-Shiller HPI - YOY Change in Seattle Tiers

The low tier inched closer to matching its October 2005 +19.6% performance but in opposite magnitude. The high tier actually improved slightly from a year-over-year basis, for the first time since March 2006. The low tier again kept its title for largest YOY decline, with the high and middle tiers swapping places. Here’s where the tiers sit YOY as of February – Low: -18.1%, Med: -14.9%, Hi: -14.6%.

Lastly, here’s a decline-from-peak graph like the one posted yesterday, but looking only at the Seattle tiers.

Case-Shiller: Decline from Peak - Seattle Tiers

What’s interesting is that the decline softened somewhat month-to-month for the high and low tiers, but continued at roughly the same rate for the middle tier. It will be interesting to see if this is just a one-month outlier or the beginning of some sort of pattern.

(Home Price Indices, Standard & Poor’s, 04.28.2009)

0.00 avg. rating (0% score) - 0 votes

About The Tim

Tim Ellis is the founder of Seattle Bubble. His background in engineering and computer / internet technology, a fondness of data-based analysis of problems, and an addiction to spreadsheets all influence his perspective on the Seattle-area real estate market. Tim also hosts the weekly improv comedy sci-fi podcast Dispatches from the Multiverse.

22 comments:

  1. 1
    deejayoh says:

    I am not a big fan of the C-S tiering methodology. given that they use fixed price bands, isn’t it possible that properties move between tiers as price drops? Seems like it is not pure apples to apples comparison over time.

  2. 2
    patient says:

    RE: deejayoh @ 1 – Good question. If for example high tier drops into mid tier you would think the former high tier properties price decline is now part of the low tier average decline even though it’s a high tier price fall. If that’s how it works, or if c/s counts the source tier belonging for each transaction?

  3. 3
    kfhoz says:

    It is not fixed-price bands. Price breakpoints between low-tier and middle-tier properties and price breakpoints between middle-tier and upper-tier properties are DYNAMICALLY computed for each period, so that there are the same number of sales in each of the three tiers.

  4. 4
    patient says:

    RE: kfhoz @ 3 – The problem kind of remians the same though. Do you know if a transacation is put in the source or target tier should it move from to another between the previous a current sale?

  5. 5
    Nathan says:

    I think what we are seeing is that a majority of the houses in the middle tier should be selling in the bottom tier. If that is the case, and the middle band houses are over prices, then it doesn’t matter if the bands adjust dynamically, the middle band will still drop the fastest, the lower band will drop slower, since it is absording the new listing from the middle band that are now in the top of the lower band.

    someone feel free to slap me around if I’m way off base here.

  6. 6
    jon says:

    The tiers are based on the current sale transactions rather than the housing stock. Right now there are very few sales at the high end, so the CS high tier consists of mostly mid priced homes.

  7. 7
    The Tim says:

    From the Case-Shiller methodology pdf:

    The Division of Repeat Sales Pairs into Price Tiers

    For the purpose of constructing the three tier indices, price breakpoints between low-tier and middle-tier properties and price breakpoints between middle-tier and upper-tier properties are computed using all sales for each period, so that there are the same number of sales, after accounting for exclusions, in each of the three tiers. The breakpoints are smoothed through time to eliminate seasonal and other transient variation. Each repeat sale pair is then allocated to one of the three tiers depending on first sale price, resulting in a repeat sales pairs data set divided into thirds. The same methods used for the Metro Area Indices are applied separately to each of these three data sets to produce the Low- Tier, Medium-Tier and High-Tier Indices.

    Note that the allocation into tiers is made according to first sale price. Individual properties may shift between price tiers from one sale date to the next. We use only the tier of the first sale, ignoring the tier of the second sale. This allocation was chosen so that each of the tier indices closely represents a portfolio of homes that could be constructed on each date using information actually available on that date. Thus, the tier indices are essentially replicable by forming a portfolio of houses in real time. The Low-Tier index for a metro area is an indicator of a strategy of buying homes falling in the bottom third of sale prices (while the High Tier Index as an indicator of a strategy of buying homes in the top third of sale prices) and holding them as investments for as long as the homeowner lived in the home. The trend of home price indices in each of the three tiers reflects the outcome of such an investment strategy.

    A “value effect,” has been noted in the tier indices: low-tier indices have typically appreciated somewhat more than high-tier indices. Part of this value effect may be analogous to the effect that motivates value-investing strategies in the stock market. Individual homes’ prices have shown some tendency to mean revert, so purchasing lowpriced homes may have been an overall good investment strategy. We do not know whether this value effect will continue into the future, and the value effect has not been stable through time even in the historic sample that we have observed.

    The high-tier indices will tend to lie closer to the aggregate indices than do the low-tier indices. This is as we would expect, since the aggregate indices are value-weighted and hence the high-tier repeat sales figure more prominently in the aggregate indices.

  8. 8
    Kary L. Krismer says:

    I would think fixed would be much better than dynamic, or maybe something that is “sticky.” Perhaps high could be the lower dollar limit for the high tier the entire prior year, and it would only change once a year. (I should have read Tim’s post first–seemingly they do something different to accomplish what I was thinking.)

    If you look at the March NWMLS numbers, the mean fell dramatically, over $50k, with the difference between the mean and median being the smallest amount in at least three years, and roughly half what is typical. That means higher end properties simply are not selling at even the same depressed levels as other listings. So presumably for March though May, the C-S high end tier limit will be rather low. And as it drops, presumably the drop they show will be less than if they left the cutoff higher, because the tier will include houses in a lower, more popular, price range.

  9. 9
    deejayoh says:

    By jon @ 6:

    The tiers are based on the current sale transactions rather than the housing stock. Right now there are very few sales at the high end, so the CS high tier consists of mostly mid priced homes.

    I was thinking that too – but if they tiers are not fixed (as I have been corrected) then mid tier houses are then probably showing up as high tier, low tier as mid tier, and truly expensive homes are hardly represented at all

    which means I still have problems with using the data this way;…

  10. 10
    Scotsman says:

    Love that first chart- heading for the bottom… in 2014.

    In other news, the 10 year treasury has shot up 21 basis points to 3.11%, the highest in months, and
    all those new treasuries haven’t even hit the market yet. Expect interest rates to head up.

  11. 11
    Garth says:

    My problem with the tiers is that are spread out over the three counties and thus pretty useless as $400,000- 500,000 house in most parts of pierce or snohomish counties would be a higher end home there, while the same amount in seattle / king county buys a house that probably belongs in the lower tier.

  12. 12

    ANOTHER TSUNAMI OF TOXIC LOANS TO RESET SOON IN SEATTLE, JUST LIKE CALIFORNIA?

    And just at the wrong time, with home price crashes and unemployment making it far, far worse this time.

    I think The Tim would love the charts on this California website; they are very creative in my opinion.

    See the article:

    http://www.doctorhousingbubble.com/notice-of-financial-default-california-develops-a-mortgage-tsunami-patter-reminiscent-of-the-2007-subprime-collapse-alt-a-and-option-arms-unite/

  13. 13
    kfhoz says:

    RE: Garth @ 11 – Neighborhood location is part of what puts a house in a particular tier. If you narrow down to King County then a million dollar house in Medina is in the same tier as a million dollar house in Kent. If you only consider Seattle then a half-million dollar house in Skyway is still the same tier as a half-million dollar house in Broadmoor.

  14. 14
    One Eyed Man says:

    RE: deejayoh @ 9

    Admittedly, the CS Methodology pdf isn’t the easiest thing to interpret, but I think this issue is explained in the first sentence of the 2nd paragraph quoted by Tim. It says:

    “Note that the allocation into tiers is made according to first sale price. Individual properties may shift between price tiers from one sale date to the next. We use only the tier of the first sale, ignoring the tier of the second sale. This allocation was chosen so that each of the tier indices closely represents a portfolio of homes that could be constructed on each date using information actually available on that date. Thus, the tier indices are essentially replicable by forming a portfolio of houses in real time.”

    If I understand correctly, the property is assigned to a tier based upon which third the first sale occurred in. That sale could have been at any time over the past, up to about 6 yrs ago (the actual maximum number of years is specified in the pdf and newer sales are weighted more than older ones). The sale of any pair that took place in February is not considered when the pair are assigned to a tier. As a consequence each tier for February is not necessarily made up of the same number of paired sales. But each tier is made up of sales that would reflect a group of properties that were in that tier when the first transaction for the property occurred some years earlier. The idea is to model what would have happened if you bought a portfolio of properties all in one tier when you purchased them.

    Given the complexity of the method, I think you’d probably have to do some fancy statistical analysis of the data to prove whether the tiers are a valid sample or pure BS. I would imagine Shiller’s probably done the analysis and decided the method has value for the purposes stated. That purpose being to model price change of a pool of properties in a specific tier when previously purchased).

  15. 15
    Garth says:

    kfhoz,

    Your last example pretty much illustrates my issue, the “low tier” does not even exist in most of seattle, and there are no $500,000 homes in broadmoor, though if there was one, it could not be compared in a meaningful way to one in skyway..

    Working with data for my job, I am always suspicious of summaries of data that is already a summary like the CS index, as the result has too many different methodologies at work to say the result is valuable.

  16. 16
    Kary L. Krismer says:

    It’s much more useful to know things like in 2009 only about 30 houses over $1M have sold per month, and that there are over 1500 active listings at such a price.

    But then again, I’d say the same thing about C-S in general. I’d much rather know that the median in February for King County was $375,000 and the mean almost $475,000 than know the C-S number for three counties was 152.12. For that number to be useful, I’d have to look up the January 2000 values for King county–the point which C-S set as being 100.

  17. 17
    EconE says:

    Well…anyhow…

    I just Zillowed the good Doctor’s house.

    Funny that it’s only been viewed 30 or so times…you know…being a “person of public importance” and all.

    It would be interesting to ask the Doc a few questions.

    1. What he thinks of the “Zestimate” value? (exactly 10x purchase price in 08/89 per Zillow)

    2. Does he feel that he bought during a bubble? (His claim to fame chart says yes)

    3. Did he pay “market price” for his home at the time? If not, what were the circumstances?

    4. What should his house be “worth” in accordance with his own methodologies?

    5. What upgrades has he done and when?

    6. What value does he place on his upgrades? (if any)

    7. Aren’t those taxes kinda high?

  18. 18
    Nwerner says:

    I have a question about the CS index that has been nagging me. I have read on here may times about the imperfect metric that is median home price as measured by the MLS but the CS index strikes me as a potentially skewed indicator as well. While I concur that a paired sales analysis of the same house’s repeat sales is a good indication, during the peak of the boom when there was a spate of ‘flipping’ wouldn’t the paired sales analysis have been potentially overstated by the numerous upgraded houses that were flipped?

  19. 19
    The Tim says:

    RE: Nwerner @ 18 – Again quoting from the Case-Shiller methodology pdf, page 7 this time:

    Price Anomalies. If there is a large change in the prices of a sales pair relative to the statistical distribution of all price changes in the area, then it is possible that the home was remodeled, rebuilt or neglected in some manner during the period from the first sale to the second sale. Or, if there were no physical changes to the property, there may have been a recording error in one of the sale prices, or an excessive price change caused by idiosyncratic, non-market factors. Since the indices seek to measure homes of constant quality, the methodology will apply smaller weights to homes that appear to have changed in quality or sales that are otherwise not representative of market price trends.

  20. 20
    Kary L. Krismer says:

    RE: Nwerner @ 18 – There is no perfect system for judging the value of houses collectively. C-S tries to do away with some of the problems of mean/median, which are related to the mix changing over time. But it has its own problems too.

    If you really want to know the value of your house you need a CMA and/or appraisal, and that’s only likely to be accurate within +-5% if you’re lucky. That at least should only use comps of houses that are relatively similar to yours and in your same general area.

  21. 21
    Kary L. Krismer says:

    RE: The Tim @ 19 – I wonder what they’re going to do with all the short sales? I have two clients looking for low priced properties down in the Kent/Renton area, so yesterday I did a little tour of such properties. There was only one property that was in even average condition, to be generous, and it’s problem selling was likely that it was next to a very busy road. That is often overlooked in a hot market, but not in a market where there’s a ton of inventory.

    Anyway, I suspect the paired sale method for those properties will show a substantial decline due to condition, because I doubt many of those properties were in as bad of condition when purchased. Sort of the opposite of the remodel situation. So realistically, C-S should throw those out.

    Anyway, seeing those properties sort of supports my theory that expanding the home ownership percentage isn’t a good idea because there are a lot of people that simply are not able, for one reason or another, to maintain a property. They should be tenants, and God help the landlord that rents to them.

  22. 22
    pa.pow says:

    First timer post … There seems to be a lot of interest and commenting on this particular group of data (meaning C-S in general ) which is arguably considered to be the best measure of price change, albeit 2 months in the past. To help in understanding the current market conditions it would be useful if this lag could be reduced. Is there a reasonable way to approximate what the C-S results will be ahead of the 2 month lag? (I know it’s only one piece of data that the community considers but it seems to be the measure used in the end to either validate or discredit what predictions were previously made).

Leave a Reply

Use your email address to sign up with Gravatar for a custom avatar.
Your email address will not be published.

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

Please read the rules before posting a comment.