Page 1 of 2 12 LastLast
Results 1 to 15 of 16

Thread: Recent RE sales reports and why they DON'T necessarily speak to value

  1. #1
    Join Date
    Nov 2005
    Location
    Wayland MA
    Posts
    1,431

    Default Recent RE sales reports and why they DON'T necessarily speak to value

    A 2/13 Boston Globe article reported that the 2008 median selling price in Wayland had dropped 20% relative to 2007 while the same metric in Weston had increased by 7.5%.

    Bad news for Wayland, right?

    A 2/25 MetroWest Daily News article (attached) reported that the January 2009 median selling price in Wayland had increased almost 26% over January 2008 while Weston dropped almost 36%.

    Great news for Wayland, right?

    On both counts, not necessarily. While these reports are factually correct, the speak only to the median selling prices of homes in a given period. Without additional information, one can infer nothing about real estate value.

    Here's why. Imagine that in 2007, the median home that sold in Wayland was 3,000 square feet on an acre of land. Now, in 2008, imagine that the median selling home was 2,500 square feet on 3/4 of an acre. In a flat real estate market, which median home would you expect to have a higher value? The 2007 home, of course.

    In fact, one might expect the latter to sell for on the order of 20% less than the former. The 20% "drop," however, is not necessarily a drop at all. It might be, but without doing a thorough comparison of the homes to "normalize" them, the difference may well simply be that we aren't comparing equivalent groups of homes.
    Attached Images Attached Images

  2. #2
    Join Date
    Mar 2008
    Location
    Wayland MA 463 Old Conn Path
    Posts
    382

    Default Statistics or Observation?

    A well known citizen lawyer came to the FinCom microphone on 2-Feb-2009 and provided a story and a warning.... have a look

    http://www.youtube.com/watch?v=b5cdaogLvuI

    If you want to debate statistics, fine but lets not delude ourselves into thinking all is fine and Wayland is no worse than other communities. The Warren Group stands by its comparative mean analysis and they told me that directly. I believe them, they say they have 135 years of experience doing it. Don't debate with me, call them directly.

    Best advice:
    1. If you refi your house make sure your home equity credit line stays intact because the re-appraisal isn't going to look pretty
    2. If your going to get divorced, wait till after the depression is over.

  3. #3
    Join Date
    Jan 2006
    Posts
    117

    Default Worst Kind of Lies

    Not quite statistics but you get my point.

    I have been away from the message boards for awhile because of my kids sports schedules but I recently had a chance to read through all of the posts I have missed. I must say I have found them quite entertaining, especially the fact free posts and the ones that present factoids as hard data.

    A very kind person sent me the real estate sales listings for Wayland by year. This data was provided for a different purpose but is very interesting in light all the postings on sales price decline in Wayland.

    I put together a spreadsheet that had all of the selling prices for 2007 and 2008. There actually was very little difference in number of home sales (unless of course you are someone with your home currently on the market or are a realtor) for 07 and 08 (137 vs 130). The big difference is in the type of homes that sold. I used $750,000 (because someone mentioned that figure on the Towncrier website) as a data point. In 2007 48 homes sold for >750,000 in 2008 it was 26 homes. In 2007 24 homes sold for <400,000 while in 2008 it was 34.

    There can be a number of reasons for this difference but to draw a conclusion that it is the high tax rate does not seem to work. The proponent of the claim that the high tax rate is hurting Wayland real estate sales has also stated that it is significantly harder on the lower priced housing stock. The fact that over 40% more homes with lower prices sold in 2008 certainly is counter to that argument. The lower priced homes also started with lower listing prices so it is not due to a huge drop from expectations.

    Since the MLS data came to me with a boiler plate disclaimer for its accuracy I too can not claim it is totally accurate and would welcome corrections. My data did not replicate the 20% decline that has been quoted and attributed to the Warren Group but it is close and it is interesting none the less.

    I have attached a spreadsheet with 2007 and 2008 sales and associated MLS# for your review.
    Attached Files Attached Files

  4. #4
    Join Date
    Mar 2009
    Posts
    13

    Default Observations on statistics - part 1

    The Boston Globe, Boston Magazine and Metro West Daily News have all published statistics on median home prices from the Warren Group. Here are some observations:

    1.All statistics have a margin of error/accuracy associated with them, yet I don't see this reported anywhere. One should make an allowance for this before one makes inferences based on the data -especially if the data comes from different sources. Interestingly, last year's Boston Magazine article on a similar subject appears to have slightly different data for 2007 prices than this year's article. I could not find a mention of the source of the data for the 2007 article, so maybe they use different sources. Differences appear to range from 1% to almost 10% depending on the town.

    2.Understanding the nature of the distribution of the underlying data along with the median is important to prevent the "comparing apples & oranges problem" that Jeff is referring to in his post. Unfortunately, that information is not provided along with the Warren Group Data. As a "quick and dirty" test I used the data provided in the post by BTDowns to see if I could create his 2008 data set from his 2007 data set. The results were a lot closer than I expected, matching for most of the price range, but showing some divergence in the price band around $1M. This suggests the "apples and oranges problem" exists, but is not big enough to invalidate the data. See chart 4 of the Excel attachment and let me know what you think.

    My conclusion is that the reported "20% drop in median prices" overstates the actual trend somewhat, but there is a big drop across the price range with the price band around $1M getting hit the hardest.

    3.Assuming there is a big drop, it needs to be put in some context before drawing any conclusions. Luckily, the online version of this year's Boston Magazine article provides price data for 1998 and 2005 as well as 2007and 2008. (Note that there is no data for 2006) I loaded the data for Sudbury, Wayland, Weston and Concord into a spreadsheet to calculate the appreciation over time (see sheet 2, chart 6 and chart 9 of the Excel attachment)

    What the data shows is that Wayland and Concord had almost identical growth rates with both peaking in 2007, but then Wayland dropped significantly more than Concord from 2007-2008. Sudbury grew faster through 2005 (it's peak) and dropped in both 2007 and 2008. It is interesting to note that Wayland and Sudbury have both dropped about 20% from their peaks, but Sudbury's drop was spread over two years rather than one. Concord has only dropped about 5% from it's peak. Weston has yet to drop. Over a 10 year period Weston was #1, Concord was #2, Wayland #3 and Sudbury was #4.

    My overall conclusion is that the Boston Magazine article paints an overly negative picture of Wayland due to the limited breadth of their analysis. This does not mean that we should dismiss what happened to our property values over the past year, but it does mean we only need to look as far as Sudbury rather than all the way to Lowell to find company. Likewise for the Globe.
    Attached Files Attached Files

  5. #5
    Join Date
    Nov 2005
    Location
    Wayland MA
    Posts
    1,431

    Default

    Gary, thanks much for the thoughtful analysis. The insight that Sudbury and Wayland have experienced a similar effect when looked at over two years instead of one is truly illuminating.

    In the 3/26 Town Crier, Wayland resident and real estate broker Barry Nystedt (whom I don't know) wrote the following.

    Quote Originally Posted by Barry Nystedt
    With the MLS Property Information Network as my source, I downloaded the sales and the living area square footage of the sales over the last two years ($262 per square foot is the median for 2008, $280 per square foot is the median for 2007). The percentage change on a cost per square foot analysis is only 6 percent lower than 2007. The median house sale in 2007 was actually 10 percent larger. That is a far cry from 20 percent quoted in the articles – 333 percent off the mark. That is in line with Standard & Poor’s S&P/Case-Shiller Home Price Index change from December 2007 versus December 2008, a 7 percent decline for the entire Greater Boston region.

    Wayland is not outside the norm of what is happening in our region! In fact, the four-bedroom home sales in Wayland from 1,500 to 2,500 square feet increased 2 percent in 2008 versus 2007 on a cost per square foot basis. A similar property in Sudbury (most comparable community to Wayland) was 13 percent lower than Wayland for 2008.
    Gary, the one thing that I don't understand in your analysis is how you "used the data provided in the post by BTDowns to see if I could create his 2008 data set from his 2007 data set." Absent descriptions of the property (if available, I might have added land area and location to the house square footage that Mr. Nystedt used), it seems to me that there's still a "looking at the price of apples in 2007 and the price of oranges in 2008" problem.

  6. #6
    Join Date
    Mar 2008
    Location
    Wayland, MA
    Posts
    52

    Default Ahh...a breath of fresh spring air

    So nice to hear some more down-to-earth news than from "the sky is falling" woman on the Town Crier boards.

  7. #7
    Join Date
    Mar 2009
    Posts
    13

    Default Clarifications

    Jeff,

    I did not mean to imply that there was no "apples and oranges" problem. A better statement would be that the "apples and oranges" problem doesn't appear to be any greater than what one would normally encounter in an experiment based on population sampling techniques (across most of the price range). This indicates that one should be able to use the median data the same way one uses the results of other population sampling techniques, such as election result polling (i.e. carefully). A difference is that poll results are usually reported with a "margin of error" to warn people, whereas this data is not.

    Here is my "quick and dirty" method/reasoning:

    1. The "2007 BTDowns" data represents a non-random sample of the prices of the entire 2007 population of Wayland homes. It has an associated sampling error.

    2. The "2008 BTDowns" data represents a non-random sample of the prices of the entire 2008 population of Wayland homes. It has an associated sampling error.

    3. We are hypothesizing that the relationship between the entire 2007 population of Wayland homes and the entire 2008 population of Wayland homes is that the 2008 price is 16% lower than the 2007 price. (note 16% is the difference between the 2007 and 2008 median prices in the BTDowns data - the Warren Group difference in medians is 20%) Presumably, there is an error associated with this hypothesis (price shifts are usually not constant across all price bands).

    4. The combination of the two sampling errors is what causes the "comparing apples to oranges" effect. (i.e. if one were comparing the entire populations, then one could compare apples to apples and oranges to oranges.)

    5. Based on this, the 2008 data sample (represented as either an ordered set or a graphical image) should be equal to the 2007 data sample (represented either as an ordered set or a graphical image) times .84, with any difference being attributed to the sum of the sampling errors (call it the "apples to oranges error") and the hypothesis error.

    The "quick and dirty" test is just multiply the 2007 data set by .84 and compare it to the 2008 data set and measure the difference.

    The statement "see if I could create his 2008 data set from his 2007 data set" more literally meant "see if the difference between the actual 2008 data set and the 2007 data set multiplied by .84 (i.e. the predicted 2008 data set) was within the bounds of acceptable experimental error". The implication would then be that the "apples to oranges error" and the hypothesis error would be within the bounds of acceptable experimental error.

    I cleaned up the spreadsheet and graph from the last posting and attached it. See Sheet 3 for the ordered set and Chart 2 for the graphical image. The result is that the combined error is within 5% over most of the price range and within 2-3% over much of it - which is better than I expected.

    The $250K and under price band is 10-20% higher than predicted, but it only has 3 samples, so it is likely to have a large "apples to oranges" (sampling) error.

    The roughly $300K-$400K price band is about 10% higher than predicted and the number of samples is reasonable (about 25) so it may have experienced less depreciation relative to other price bands.

    The $1M-$1.5M price band is about 20% lower than predicted, but the sample size is small, so it is likely to have a large "apples to oranges" (sampling) error.

    In general, price ranges below the median ($505K) seemed to depreciate slightly less than above the median.

    In summary, the BTDowns data suggests that prices dropped about 15% with a margin of error of about 5%. There is a reasonable possibility that homes around the $300 to $400K range depreciated less than that and a slight possibility that homes around the $1M to $1.5M price band depreciated more than that, but there is not enough data for the $1M-$1.5M price band to be confident.

    I am somewhat bothered by the difference in median values between the BTDowns data (MLS based), Barry's data (MLS based) and the Warren Group data as it suggests a systemic error component not accounted for in the above analysis.

    Remember, all of the above should be taken in the spirit of probabilities (shades of gray) rather than absolutes (blacks and whites).

    I hope this helps to clarify my last post.

    The data is attached, and observations/questions/comments/critiques are encouraged.

    Gary

  8. #8
    Join Date
    Nov 2005
    Posts
    726

    Default another take at the analysis

    Gary,

    I did an analysis where I looked at the 2006 (and also 2005, 2007 and 2008) assessments of those properties that sold in 2007 v. those that sold in 2008. I figured looking at the assessments prior to any of the sales ensured that there was no bias from adjustments that might occur toward the actual sales price.

    I found that those properties that sold in 2007 had both an average and a median sale price about 10% higher than those that sold in 2008.

    What do you think about this approach?

    Kim

  9. #9
    Join Date
    Mar 2009
    Posts
    13

    Default Assessed vs. market pricing

    Kim,

    Thanks for taking the time to add to the public pool of data - we just need to sort it all out so we can make the best possible inferences. We have some important town policy decisions coming up!


    I am not sure I understand your method.


    When you said:

    "I found that those properties that sold in 2007 had both an average and a median sale price about 10% higher than those that sold in 2008."

    Did you possibly mean to say "median assessed price" rather than "median sale price" ?


    In any event, market pricing and assessments are just two different methods of determining the intrinsic value of a house. The two different methods will generate different results and one can debate the relative biases, but one would hope that they are correlated over time.

    Care needs to be taken with assessed prices though, because they introduce a somewhat subtle time shift to the data. This is due to the fact that assessed price is determined at a point in time before the actual sale price is. For example, if one assumes that assessments are performed every year (I think this is the case), they may be as much as a year old at the time of sale.

    So what you are reporting as 2007 to 2008 changes may actually be more indicative of 2006 to 2007 changes - it would be interesting to compare the two methods over time.

    Also care needs to be taken that the assessment methods can be changed. I don't remember Wayland's recent history, but I know assessment methodology is a subject that has been under discussion. A change in method would cause a discontinuity in the year to year data for a single year.


    /Gary

  10. #10
    Join Date
    Nov 2005
    Posts
    726

    Default

    Gary,

    I did mean to say that the median ASSESSMENT had changed. And this was finding was consistent from year to year, so it did not seem to be dependent upon the methodology or formulas used in the assessment process. It seemed to me to indicate a difference between the housing stocks that sold in the two years.

    Kim

  11. #11
    Join Date
    Jan 2006
    Posts
    117

    Default Sample Size

    Gary,

    You have clearly put a great deal of thought and knowledge into your analysis. Thanks for setting such a standard. I can not actually contribute at that level.

    That being said I do think that the real problem in trying to use the samples to calculate statistically relevant metrics is the sample size. I believe the limited number of sales in any given year means that the calculation of median or average does not give a reliable answer. The fact that the number of sales are roughly comparable from year to year does not mitigate the unreliability of the metric. I know we all would agree that comparing one month's sales to another month is completely unreliable and I do not think one years worth of data creates a large enough data set.

    You thoughts?

  12. #12
    Join Date
    Mar 2009
    Posts
    13

    Default Dinosaur Tails

    Ben,

    Here is a better explanation of my approach along with my thoughts:

    Approach:

    Statistics have a reputation for lying, so the prudent thing to do is subject them to a "lie detector test" before believing anything they say.
    Such a test usually consists of hooking up a supportive witness to a polygraph and subjecting the witness to interrogation. If the witness shows strong character, and their story is consistent and focused under interrogation, you have a tendency to believe them.

    In this case, I am taking the sample sets and manipulating them in various ways and observing the shape (character) of the graph, its consistency and whether or not the data points are focused around the expected shape. The fact that the sample sets have almost the same number of members is of no interest other than the fact that it makes the comparisons much easier.
    Sampling error (the source of many lies) is usually shows up as weak shape, inconsistency and lack of focus (lots of scattered points).

    After the initial interrogation, your data showed weak shape, but relatively strong consistency and focus. This was perplexing. Then under further interrogation (changing the scale of the graph to logarithmic) a stronger shape emerged.

    What this means is that your data sets have a lognormal rather than a normal distribution. Note that lognormal distributions are not "abnormal" distributions and they are relatively common. They look like normal distributions (which are bell-shaped) with a dinosaur tail. The problem is that if you unknowingly apply basic statistics to a lognormal distribution, you are going to get distorted results (that could be interpreted as sampling error). No matter how big you make your sample, the "tail" and its effects will not go away, it will just become more distinct. The resultant error is due to the violation of a basic assumption the statistic relies on (that the data follows the bell curve) rather than the size of the sample.

    As long as you know the data is log normally distributed ahead of time you can account for the "tail effects" when you are making inferences from the statistics.

    Thoughts:

    With the caveat that your data may not be representative of all data, it appears that one year's worth of data can yield reasonable accuracy (margin of error of about 5%) as long as you are careful to account for the "tail effects" of the lognormal distribution and recognize that there can be points that lie outside the margin of error.

    I would not trust any single home price statistic derived from a sample gathered over less than a year without studying the sample data. Data sampled at intervals less than a year can be useful for analyzing trends over the course of a year if presented as a continuous graph (see attachment). The graph in the attachment indicates a downward trend in Wayland Home prices over the course of 2008. Note that large sampling error is indicated by the volatile behavior of the “1 week” graph, and it becomes less evident, but still present in the “90 day” graph.
    Attached Images Attached Images

  13. #13
    Join Date
    Dec 2005
    Posts
    44

    Default

    Gary,

    I have no real background in statistics and am having a bit of trouble following your analysis. It would greatly help if you would answer the following question: After doing the analysis that you have described in your posts, do you conclude that, based upon the median sale prices in Wayland in 2007 and 2008, real estate values in Wayland decreased by approximately 20% from 2007 to 2008 with a 5% margin of error? Thank you.

  14. #14
    Join Date
    Mar 2009
    Posts
    13

    Default 10yr Warren Group Data Analysis

    Here is a spreadsheet that makes it easier/safer to interpret the Warren Group Data. I included the 2009 "year to date" data as a pick me up - but be careful with it, it is only an early indicator and has a very high margin of error.

    Kim,
    I wanted to get this data before commenting more on your approach. Is your data in a format that it could be added to this spreadsheet and compared? I am concerned about drawing conclusions from just two years of any approach.

    /Gary
    Attached Files Attached Files

  15. #15
    Join Date
    Nov 2005
    Location
    Wayland MA
    Posts
    1,431

    Default

    So that we don't end up with two threads on the same topic, it may make sense to continue this discussion here.

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •