|
Bay Area Real Estate Market Newsletter
("BAREMN")
as of September 5, 2003
We anticipate being able to resume the monthly data and the analysis of that
data in a more timely manner. We hope to have September's data and analysis
posted by September 14th. Below brief summary of some of the challenges that we
have endured during the past two months as the MLS changed computers, vendors
and software simultaneously.
The MLS "transition" that started on July 20th was and
still is a
complete disaster. We used to perform 1235 searches nightly in 4 hours downloading
4-5 megabytes of information. As a result of the "transition" we now only
do 343 searches, take 6 hours and downloads about 50 megabytes of information most
of which is overhead. The previous MLS provided 40 detailed listings per minute
compared with the new MLS that only provides 4 detailed listings per minute.
Previously, a
detailed listing was about 4,800 bytes compared with the new MLS at
approximately 93,000 bytes. A primary reason
for these changes is the MLS changed from a text only format to a graphic user interface.
A secondary reason is the new servers must find all the data before they start
to download the information. The old servers started downloading the information
as soon as the first listing was found. In other words, the MLS changed from a
faster parallel process to a slower serial process. After it became clear that
these were permanent opposed to temporary issues we started making changes to
help compensate for the drastic reduction in information throughput. We
purchased a new computer that is dedicated to downloading MLS data 24x7.
We also drastically reduced the data we collect. We completely dropped San
Bentio County and made several adjustments to our price ranges. For example for
condos we now combine all properties below $250,000 and combine properties above
$1,000,000; for houses all properties below $350,000 are combined; and for multi-unit
properties below $500,000 or above $1,000,000 are combined. We also stopped
tracking currently available properties and off-the-market properties because
the MLS no longer indicates when properties were originally listed and no-longer
indicates which properties have been re-listed. Lastly, we were also forced to
eliminate our quality control searches.
We
have also noticed and corrected numerous discrepancies with the data.
Unfortunately, we know that we have not caught all of the issues. For example,
It is difficult to find and therefore count a listing that has sold when the
sales date is blank or 1/1/2000. The good news is we believe we found the
majority of these errors. But making the manual correction increases the chances
of mistakes and occasionally requires us to make assumptions. Of course it took
time to learn the nature of these database errors in order to find them. We
believe that our data will get more accurate as time elapses since July 20th as
we learn how to deal with them. This
explains why the data near July 20th is less complete. In the near future we
expect to establish procedures for finding the bulk of the errors in order to
restore the level of confidence we had prior to the transition.
One
issue with the database that will impact our analysis for the foreseeable future
is the MLS accidentally dropped the continuous list date. Because the data is no
longer obtainable it is impossible to calculate the days on the market. Instead
of the days on the market, the new MLS reports the days since the last
re-listing. This data is completely meaningless. For example, DOM for initiated
sales for SCC dropped from 34 to 24 in 5-weeks at a period of time when the
market was flat to slowing definitely not improving rapidly. This will make the
DSI data all that much more valuable. Even if the MLS returns the data soon it
will take time (6-months) for the majority of the listings to have accurate
continuous list dates. This explains why we dropped our analysis of available
and off-the-market properties.
Because
of all these changes we would greatly appreciate it if you could alert us to
anything that seems unusual. This will allow us to take a closer look at the
issue. Although there are likely minor discrepancies we believe that this data
is a reasonably accurate representation of the actual market conditions.
|
