- NAR’s monthlies are annualized and seasonally adjusted. Numbers of homes sold in winter are so low, in many places less than half of spring sales, that percentage changes are magnified.
- Seasonal adjustment can be inaccurate; neither weather nor interest rates nor the economy are consistent from one winter to the next.
- Trying to collect all the sales in a consistent manner each month is a challenge.
- If TRID did slow November closings, it pushed them into December, which should elevate next month's results.
The National Association of Realtors’ monthly series on sales of existing homes should come with a data-translating thesaurus. NAR does not mean to mislead, but I think its data format conceals reality. The data are useful, but only as a broad, multi-month guide to trends.
Each monthly figure gets a lot of headline ink, and the 10 percent decline in November got plenty — and some editorial commentary by NAR.
But each monthly is not what it seems. I know that few readers are fond of math, but here it comes.
NAR’s monthlies are annualized and seasonally adjusted. “Annualized” means the actual monthly sales figure multiplied by 12. The seasonal adjustment tries to take into account…seasons, which are a profound variable in home sales.
The annualized monthly result is then trumpeted as a percentage change from the prior month’s annualized adjusted figure.
Technologists refer to GIGO — garbage in, garbage out. NAR is not that bad, but the weaknesses stand out: first, the season. Numbers of homes sold in winter are so low, in many places less than half of spring sales, that percentage changes are magnified.
Far better to compare non-adjusted non-annualized November 2015 to the same for November 2014, and get a good year-over-year trend.
Second, seasonal adjustments themselves are wildly inaccurate: neither weather nor interest rates nor the economy are consistent from one winter to the next.
The third deficient approximation is just trying to collect all the sales in a consistent manner each month. Most of us struggle to get it right even in our metro areas, trying to smooth MLS data compared to recording data at county courthouses.
That inherent error in NAR data is plus or minus 5 percent at least (is that transfer at the county a sale, or a correcting deed from one person to another person with a new spouse and surname?). The alleged 10 percent drop in sales from October to November is at the outer edge of statistical scatter, possibly trendless.
The last element, editorializing, claims that TRID — new in October — slowed November closings. If it did, it pushed them into December, which should elevate those results, unless we believe that the real estate, title and mortgage industries did not get better at TRID with experience.
I like the idea of blaming my competitors for being slow on the uptake, but they’re not that slow.
Lou Barnes is a mortgage broker based in Boulder, Colorado. He can be reached at firstname.lastname@example.org.