by Calculated Risk on 6/17/2011 04:09:00 PM

CR Note: The California Association of Realtors (CAR) released their May home sales data this week: May sales and price report

Economist Tom Lawler writes:

Using a “reasonable” estimate of the California Association of Realtors’ (CAR) seasonal factor for May (I have some history of CAR’s NSA numbers), the CAR’s May home sales report would imply an estimate of about 45,000 existing SF detached home sales on an unadjusted basis last month. That number is about 26.6% higher than Dataquick’s (DQ) count of new and existing sales of SF homes and condos in California last May, indicating that the CAR’s existing home sales estimates continue to vastly overstate overall existing home sales in the state.

CAR, of course, uses a methodology similar to the National Association of Realtors (NAR) to estimate existing home sales – it estimated total existing home sales (including sales outside of MLS) in 1999 based on data from the 2000 decennial Census and from the 2001 Residential Finance Survey, and then assumes that total existing home sales growth has been the same as MLS-reported home sales growth since then.

While CAR’s and DQ’s estimates of existing SF home sales in California were reasonably close in 2005 and 2006, they started to diverge significantly in 2007 and by even more so in 2008 and 2009 – with CAR’s estimate materially exceeding DQ’s estimates. Similar “gaps” emerged in other states between the NAR’s estimate of state sales and sales based on property records. Of course, I noted this growing gap back in 2009, and discussed some of the reasons why it was occurring. The gap gained media attention earlier this year following a piece by CoreLogic on the growing disparity between NAR estimates and property records data.

The NAR has pretty much acknowledged that its methodology has probably resulted in an overstatement of existing home sales over the past several years. Normally the NAR would wait for detailed Public Use Microdata Sample (PUMS) data from the decennial Census, as well as data from decennial Residential Finance Survey. These data come out with a long lag, and the NAR’s benchmark revisions for 1999 existing home sales didn’t come out until 2004.

The 2010 decennial Census, however, did not include the “long-form questionnaire” that provided many of the data elements needed for NAR’s decennial benchmarking exercise (the benchmarking methodology, seemed “OK” for principal residence purchases but seemed “iffy” for investor/vacant home sales). As such, the NAR cannot use the same benchmarking methodology as it has done – but just once a decade – in the past.

Recent media attention on what appears to be a growing gap between the NAR’s estimates and “actual” home sales has put pressure on the NAR to accelerate some sort of “rebenchmarking, leading the NAR in February to issue a set of “Q&A’s” on the topic, including this one.

Q: When will the new benchmarking take place?

A: In 2010 Census, a long-form questionnaire was not used. Therefore, the Census no longer asked about whether people moved and bought a home. So another brand new benchmarking process is needed. NAR has already been in contact with all key housing economists in the industry and government agencies and a few in the academia about finding a new benchmarking process. We expect a new clean, agreed-upon benchmark figure by the summer of this year.

In addition, we will be determining a new way to re-benchmark on a more frequent basis, possibly annually to lessen any drift that can accumulate over time. This frequent re-benchmarking, rather than wait every 10 years, is needed since the Census no longer collects the long-form questionnaire. As with all benchmarking, we will be working with various outside housing economists to develop a new-agreed upon method.

Since I have had discussions in the past with the NAR on this topic, a number of folks have asked me “what’s up” with the NAR’s benchmarking, and when will it release new existing home sales estimates? My answer is, “I don’t know,” and I’m not even sure what methodology the NAR will use.

One approach NAR had been looking at was whether it could use data from either the American Community Survey (latest 2009) or the American Housing Survey (latest 2009, and small sample), to try to estimate existing home sales (note: the 2009 ACS/AHS would be used to estimate home sales over the second part of 2008 and the first part of 2009). Aside from the fact that (1) these data sources aren’t consistent; and (2) the housing stock estimates are not updated, these data sources also have nothing on investor/second homes. In my view this approach is fatally flawed.

A second approach favored by myself and most other housing economists is for the NAR to use property records wherever available, and to develop methodologies where property records are not available to derive sales estimates in these other areas. Relative to when the NAR first developed its “benchmarking” approach decades ago, there has been a dramatic increase in the availability of/access to property records across most of the country. As a result, in many parts of the country one can directly measure total and existing home sales – though there are some “technical” issues associated with excluding non-arms-length transactions and/or properties acquired by a lender/servicer via foreclosures, as well as in some states identifying new vs. existing homes.

This second approach, however, is not a trivial task. In many states property records are not updated in a timely fashion, and in some cases the lags from sale to recordation in publicly-available files can be pretty long. E.g., CoreLogic’s first published estimate of “non-distressed” existing home sales for the first half of 2010 (published in September 2010) was 1.191 million, but it’s latest count for the first half of 2010 (published in May 2011) was 1.352 million. Last August CL;s count of non-distressed existing homes for 2009 was 2.485 million, but this May’s count was 2.607 million.

There are also often issues on data quality/accuracy, file/data formats, etc., that require significant resources/attention. (The NAR could, of course, contract out this process.)

In addition, estimating sales in areas where no data are available is not a trivial task, and requires certain assumptions that may or may not be valid. Still, this approach is head and shoulders better than any approach using either the ACS or the AHS.

So ..(1) is the NAR going to follow the suggestions of competent housing economists and develop a benchmark based on property records? Don’t know. (2) Will the NAR come out with a new benchmark figure this summer? Don’t know, but if they do I’d bet it would not be based on property records, mainly because I doubt if they have had time to develop robust estimates (unless they have hired a third-party vendor to so do). (3) When they come out with a “benchmark” revision, what year will the benchmark number be for? Don’t know, but if they do it is likely because of data availability to be for either 2008 or 2009, and not 2010. And (4) if/when any benchmark revision comes out, will they show sizable downward revisions? For the 2008-2011 period, absolutely, regardless of what methodology the NAR employs.

CR Note: Now that summer is almost here (starts next week), maybe it is time for the NAR to provide an update on when the benchmark revision will be available - and perhaps some discussion of the new methodology. It is important to understand that the NAR estimates of home sales were probably too high over the last few years (perhaps 10% or more too high last year), and that their estimate of inventory was probably also too high.

This makes it difficult to analyze or use the data - and might have led to policy errors and also means the BEA probably overestimated the Brokers' commissions portion of Residential Investment (slightly) over the last few years!