Hitwise Intelligence - Bill Tancer - US
« Google's Weekly Search Numbers | Main | Social Networks & News - Another MySpace Phenomenon »
August 24, 2006
Unexpected Decline in Existing Home Sales? Not Exactly.
On Wednesday, the National Association of Realtors (NAR) released their existing homes sales figures for July. Headlines covering the release talked about the surprising drop in this economic indicator. Once again, leveraging the power of our search term data (as we did here in our post about the monthly unemployment number), we weren't surprised by the news; we actually anticipated this drop over a month ago.
In the chart below, I've superimposed the monthly data from the NAR on a Hitwise search term volume chart for the search query "homes for sale."
This is an excellent example of how search term volume on a term related to an economic indicator, can act as a leading indicator, helping us to predict future indicator movement. In the case of existing home sales, this is possible for three reasons:
1)The search query "homes for sale" is directly related to the subject indicator, the purchase of a home
2)Consumers interested in purchasing a home, likely search for "homes for sale" in advance of actually purchasing a home, from the data above, that gap can be anywhere between two weeks to a month.
3)The National Association of Realtors, and other organizations and government agencies require several weeks (three in this case) to compile and release their number, our search term data is based on weekly data, allowing us to chart volume every Monday for the previous week.
Aside from the excitement of discovering this indicator, the other good news is, barring any unusual circumstances over the next week, our search term data indicates a reversal in the downward trend reported by NAR yesterday. The August numbers, scheduled to be released September 25th, should reflect that reversal. Mark your calendars.
Postscript: Reuters article on this post here.
Posted by Bill Tancer at 02:59 PM
Posted to economics
Trackback Pings
TrackBack URL for this entry:
http://weblogs.hitwise.com/movabletype/mt-tb.cgi/301.
Comments
This is very interesting analysis, Bill. I am curious about 2 things. Since the NAR data represents closed transactions (not opened ones), and homes are usually in escrow for some period of time -- maybe 1-3 months -- then shouldn't your data lead the NAR data by longer? I assume (1) buyer searches for "homes for sale"; (2) buyer takes 2-4 weeks to decide (I'm surprised it isn't longer); (3) buyer opens escrow and is in escrow for 1-3 months; (4) buyer closes transaction in month x; and (5) NAR reports data for monthx in month x + 1. Second, since the NAR data is seasonally adjusted at annual rate, is your data seasonally adjusted too? Obviously fewer people are searching for homes in November say than in March. Thanks!
Posted by: Carl Reichardt at August 24, 2006 06:58 PM
Carl,
Thanks for your very thoughtful questions. When I first discovered this phenomenon, I also wondered why the lead-time was not greater, the only explanation I've come up with is that perhaps people continue to search for "homes for sale" throughout the purchase process and perhaps all the way up to closing.
Regarding your second question, the NAR began using the X-12-ARIMA model (used by the Census) to adjust for seasonality in 1989. When checking fit for different queries, I did not adjust for seasonality and found that my unadjusted query data was the best fit the seasonally adjusted NAR existing home sales number, go figure.
Bill
Posted by: Bill Tancer at August 24, 2006 07:25 PM
Dear Bill,
The correlation you've discovered is very interesting, but questions
arise because home purchases are extremely seasonal, and you've only
shown one year of data. Before reaching your conclusion, that home
sales are going to go up, I'd want to know whether the correlation
exists in previous years.
It's possible that you only went back one year because the data
provided in the spreadsheet on Realtors.org only goes back one year.
I guess they must have some marketing reason for this, because you
can go back through their old press releases and dig out the data for
previous years.
Well, it turns out that I did exactly that for an article on the
bursting of the housing bubble for my own web site (
http://GenerationalDynamics.com ). Here is the sales data back to
the beginning of 2003:
** Sales of existing homes, annualized, seasonally adjusted,
thousands of homes:
1-Jan-03,6100
1-Feb-03,5860
1-Mar-03,5530
1-Apr-03,5850
1-May-03,5850
1-Jun-03,5830
1-Jul-03,6130
1-Aug-03,6460
1-Sep-03,6680
1-Oct-03,6350
1-Nov-03,6050
1-Dec-03,6370
1-Jan-04,6000
1-Feb-04,6130
1-Mar-04,6480
1-Apr-04,6630
1-May-04,6810
1-Jun-04,6920
1-Jul-04,6720
1-Aug-04,6550
1-Sep-04,6760
1-Oct-04,6760
1-Nov-04,6920
1-Dec-04,6810
1-Jan-05,6820
1-Feb-05,6820
1-Mar-05,6870
1-Apr-05,7180
1-May-05,7140
1-Jun-05,7350
1-Jul-05,7130
1-Aug-05,7210
1-Sep-05,7200
1-Oct-05,7050
1-Nov-05,7030
1-Dec-05,6750
1-Jan-06,6570
1-Feb-06,6900
1-Mar-06,6900
1-Apr-06,6750
1-May-06,6710
1-Jun-06,6600
1-Jul-06,6330
I wonder if you might re-run your correlation with this additional
data?
Sincerely,
John
John J. Xenakis
E-mail: john@GenerationalDynamics.com
Web site: http://www.GenerationalDynamics.com
Posted by: John J. Xenakis at August 25, 2006 09:50 AM
I run a real estate brokerage just outside DC, and can address the lead time question, at least for our local market. About 1/3 of our business closes the same month as the contract is written, and I have found that to be accurate for a number of years. Meaning today is the 28th of August, the last week of the month. We still have about two weeks that contracts written will settle in September.
I have not gone back and looked at this year, but I suspect 80% of contracts written settle within 30 days. Americans are in to immediate gratification. If there was a way to be able to look at a house, and move in that night, buyers would do it.
Posted by: Chris Guldi at August 28, 2006 06:07 AM
I own a small mortgage firm in here in Orlando (a market that is different from most). In addition to mortgages we also help investors purchase forclosed homes throughout Florida and Texas.
I agree with what the gentleman before me posted (John) with regards to time frame of less than 30 days - but speaking for the foreclosure business we NEVER take that long. I wonder what the percentage of home sales are foreclosure sales vs. traditional sales? While the percentage of foreclosure sales is small to be sure, they close within a week, sometimes in two days. Enough of that type of activity would skew the data pretty substantially.
Excellent post.
Posted by: Christopher Farrell at August 28, 2006 03:18 PM
I had an error on your trackback url, but posted some additional comments on my blog
http://kwmdrealestate.blogspot.com/2006/08/one-more-thing-to-track.html
Posted by: Chris Guldi at August 28, 2006 09:56 PM
I love this data, thank you for sharing it. Is there a way to continue seeing a "live" chart for this that is updated every monday along with your updates?
Thanks!
Posted by: Glenn at August 29, 2006 01:42 PM
August home sales were horrible, and September home sales will be too. To your point:
"2)Consumers interested in purchasing a home, likely search for "homes for sale" in advance of actually purchasing a home, from the data above, that gap can be anywhere between two weeks to a month."
That knife cuts both ways. Consumers interested in selling a home probably search for "homes for sale" just as much. The Sept 25th NAR numbers will be A-W-F-U-L.
Chris Zaharias
searchquant.blogspot.com
Posted by: Chris Zaharias at September 14, 2006 07:24 PM
Chris,
Thanks for the comment. While most economists would agree with you, there have been some signs (not expected by most) that economic conditions are looking up for August (the Sept. 25th announcement will actually be for August's existing home sales data).
So far, we have August data showing gas prices declining, retail spending is up, unemployment down. Also there have been some increases in the pending home sales index over the last several months which would bode well for homes sales.
I agree with you that there could be several search missions associated with "homes for sale" however if that was the case, I would expect that it would have interfered with the previous directional predictions from the term. I guess time will tell.
-Bill
Posted by: Bill Tancer at September 14, 2006 07:36 PM
This data released today shows a very weak August:
Posted by: Chris Zaharias at September 19, 2006 10:16 AM
We still have decline as of 07/2007
As Prices Rise, U.S. Home Sales Drop
National Association of Realtors sees no change in the snail’s pace of the real estate market with a 3.8% decline in sales of existing homes last June while prices barely increased. On that same month the seasonally adjusted rate for total existing home sales was 5.75 million units. Sales were behind by 11.4% the 6.49-million pace of June 2006 and less than the 5.98 million units in May. The average price for an existing home in June was $230,100 that is only 0.3% higher than the average rate of $229,300 in June of last year.
Posted by: Vanessa Doctor at July 26, 2007 04:29 PM
