by CalculatedRisk on 11/20/2011
As many readers may recall, over the last year and a half I have noted numerous times that the NAR’s estimates for existing home sales appear to have understated the decline in existing home sales since 2006, with the “gap” increasing from 2007 through 2009. The basis for that assertion was that existing home sales based on property records in some key states declined materially more than did the NAR’s estimate of existing home sales in those states. In addition, CoreLogic’s estimates of existing home sales based on property records in its database (which covers “over 80%”of the US housing market) show materially larger declines since 2006 than do the NAR’s estimates.
The NAR is aware of these “discrepancies” and has been since at least 2009, but changing its methodology is not a trivial task. However, reportedly the NAR (working with others) has been looking into this issue, and is exploring whether it needs to change its methodology to get better estimates of “actual” existing home sales.
And from CoreLogic in February 2011: CoreLogic: NAR’s 2010 existing home sales are overstated by 15% to 20%
Historically, the CoreLogic existing sales data have covered about 85% to 90% of all NAR’s existing home sales data. However, in 2006 NAR’s sales data became elevated relative to the CoreLogic, MBA, HMDA and Census sales related data, and that trend has continued and become more pronounced through 2010. There are several reasons for the divergence, including benchmarking drift, more sales going through MLS systems due to consolidation and a lower share of for sale by owners (FSBO) home sales. Net, NAR’s existing home sales data are overstated by about 15% to 20%.
Apparently the NAR is getting close to releasing the new methodology for estimating sales, from the NAR on November 11th:
NAR presently is benchmarking existing-home sales, and downward revisions are expected for totals in recent years, although there will be little change to previously reported comparisons based on percentage change. There will be will be no change to median prices or month’s supply of inventory. Publication of the improved measurement methodology is expected in the near future.
The most important aspect of this revision is the “improved measurement methodology” so that we will have more accurate information in the future. The next most important part of the revision is the level of “visible” inventory. According to the NAR release, inventory will be revised down for the last several years by the same amount as sales, keeping the months of supply the same as originally released. With the NAR revisions, I expect listed inventory to be at the lowest level since late 2005.
That means we can estimate the downward revisions to sales by looking at other sources of inventory data. I’ve been using the monthly inventory data from deptofnumbers (aka housingtracker) for 54 metro areas.
This graph adjusts the reported NAR inventory by the HousingTracker changes, using 2006 as the starting point.
The gap between the NAR reported inventory and the HousingTracker inventory steadily increased over the last 5 years. The “drift” was fairly gradual, but cumulative over the last 5 years or so.
Using the HousingTracker data, here is what the adjustment to the NAR sales would look like (this is NOT the NAR adjustment):
|Year||Sales, as Reported||YoY Change, as Reported||Adjustment||Sales, Adjusted||YoY Change, Adjusted|
|1An example of adjustment, this is NOT the NAR adjustment, 2estimate for 2011|
These adjustments might be a little high, but I expect the current year sales to be revised down by 10% to 15%. CoreLogic expects 2010 sales to be revised down by 15% to 20%.
Hopefully the revisions will be released soon.