Tuesday, August 21, 2018

How well do weekly purchase mortgage applications forecast monthly housing permits and sales?


 - by New Deal democrat 

Tomorrow purchase mortgage applications will get their regular weekly update. Their 4 week average went negative YoY last week, for the first time in over three years, after having made a new post-recession high only two months ago.

Of course, I have no idea if this declining trend will continue or not.  But if it does persist, how reliable a portent is that for the housing market as a whole?

The purchase mortgage index has had some hits and misses in the past. [Note: Unfortunately the Mortgage Bankers Association does not permit FRED to publish their data, so the below graphs aren't as tidy as I would like them to be.]

First, here's an oldie from the early days of the Calculated Risk blog, showing purchase mortgage applications from 2001 through 2006:



Purchase applications had notable troughs roughly once a year before peaking in about June 2005.  Let's compare that with single family and total permits:



Both measures of permits appear to have made equivalent troughs and the overall peak between 1 -3 months later.

But mortgage applications seriously lagged the big decline in housing in 2006-97 by about half a year, and then did not rebound until at least 2013. as shown in this graph of purchase mortgage applications compared with total new plus existing home sales, from the Yardeni blog:



The Mortgage Bankers Association put down their problem during the housing bust to compositional issues, i.e., not adequately picking up newer, non-bank issuers.

Indeed, since 2013, mortgage applications have coincided very well with new+existing home sales, increasing in 2013 with low rates and then decreasing in 2014 in response to the "taper tantrum." Last year - 2017 - both mortgage applications and permits had a downturn during the summer that resolved positively in autumn.

So if the MBA is correct, then if the current downdraft in applications continues, it should be mirrored in the monthly reports on permits and home sales within the next several months. Because "new home sales" are *very* volatile and *heavily* revised, I can't give any kind of specific forecast for this week's report, but in the spirit of making some sort of testable prediction, the argument here suggests that most likely the report will be higher YoY (above 558,000), but below the best levels of earlier this year (672,000).