By Robert Freedman, senior editor, REALTOR® Magazine
In response to media coverage about the calculation of existing-home sales, NAR Chief Economist Lawrence Yun yesterday walked reporters through the association’s methodology for calculating its monthly EHS figures as part of his regularly scheduled press conference in Washington.
NAR each month collects home-sale data from a sampling of MLSs around the country. That monthly data is compared against the previous month’s data to derive the upward or downward shift in the home sales pace. Thus, if the participating MLSs report 5 percent fewer home sales, when all of their data is tabulated on a national basis and seasonal variations are accounted for, then NAR’s monthly EHS report reflects that.
To anchor its data, NAR every 10 years compares its figures to the findings of the decennial census, which up until 2000 sent out a “long-form” questionnaire to U.S. households to generate rich data sets on household activity, including home buying. In 2000, that long-form census identified the home sales figure at 5.2 million, and NAR “rebenchmarked” its EHS data sample based on that number.
Such periodic adjustments are standard among researchers to maintain the accuracy of any data series that relies on extrapolations from a baseline. In 2000, NAR adjusted its data by 13 percent to bring it into alignment with the census data.
NAR is now developing a new way to rebenchmark its data, because the Census Bureau no longer sends out its long-form questionnaire as part of its decennial census, leaving NAR without that data set to use as its rebenchmarking standard.
Yun said he is convening a panel of some 20-30 economists and others who follow home sales on a regular basis for their input as NAR puts in place its new rebenchmarking procedure. Depending on the procedure that’s used, NAR could rebenchmark its MLS data samples more frequently, possibly as frequently as every two years or even every year.
More frequent rebenchmarking could help minimize the statistical “drift” that occurs in any regular sampling of data. Yun said he can’t predict how much, or in what direction, EHS data has drifted since the 2000 rebenchmarking, but that some degree of drift is expected. Some MLSs have consolidated, markets are seeing fewer for-sale-by-owner transactions, and some sales might be appearing in more than one jurisdiction because agents are increasingly listing homes in more than one MLS: all of these are possible variables that could lead to drift in EHS data trends, Yun said.
Similar kinds of “statistical noise” leads to drift in home-sale data tracked by other entities, Yun said. For example, data samples that rely on FHA, Fannie Mae, and Freddie Mac closings don’t capture all-cash transactions, which today comprise more than a third of home sales. Mortgage purchase applications, though a very useful metric for directional movement on a week-to-week basis, had a huge upward drift in the 1990s, which would have implied quadrupling of home sales during that decade. And data that relies on court house recordings could be showing a downward bias because of the lag in recordings of auction sales, short sales, and foreclosure sales, among other things.
In the video above, Yun talks about NAR’s EHS calculation methodology and the development of a new benchmarking procedure.