Friday, April 24, 2015

Housing Tax Policy, A Series: Part 28 - Some more dogs not barking

I was reading through some old housing posts, and was reminded of this article from Yuliya Demyanyk at the St. Louis Fed, where she describes some research by she and Otto Van Hemert regarding the subprime market in the 2000s.  They find that teaser rates, cash out loans, high LTVs, and poor borrower creditworthiness were not significant causes of delinquency.
"(T)he subprime crisis did not confine itself to a particular market segment, such as no-documentation loans, hybrid loans, cash-out refinance loans, etc. It was a (subprime) market-wide phenomenon. For example, borrowers with mortgages that carried a fixed-interest rate—the rate that will not reset through the entire term of a loan—had very similar problems to borrowers with hybrid mortgages. Borrowers who obtained a subprime mortgage when they bought a home had the same problems in 2006 and 2007 as those who refinanced their existing mortgages to extract cash. Borrowers who provided full documentation and no documentation followed the same pattern."
Here is a chart from the article of relative delinquencies, by year of origination and FICO score.  Delinquencies increased across the board.  And, remember, by 2006, the Fed Funds rate was already at its peak level.  The loans which were part of the spike in delinquencies had been initiated at the top of the rate cycle.  These delinquencies couldn't have had anything to do with rising rates after a teaser period.  Also, note that the number of subprime loans topped out at just under 2.3 million in 2005.  There were still about 1.8 million issued in 2006, and in 2007 this dropped to 316,000.  So, in terms of quantities of defaults, the 2006 vintage is the heart of the crisis.

Demyanyk and Van Hemert find that the overwhelming cause of delinquencies in the 2006 and 2007 vintages was falling home prices.

The last home buyers to have homes with an average market value above the purchase price 12 months after they bought it were the February 2006 buyers.  Buyers in December 2006 were sitting on losses two years after their purchases of 17% nationwide and 27% in the cities in the S&P/Case-Shiller 10 city index.

Demyanyk ends the article with doubts about the predictive value of FICO scores.  But, it seems obvious to me that FICO scores were doing a pretty good job of describing credit risk until a financial tsunami hit in 2006 and after.  In 2Q 2007, an unprecedented and sharp divergence between housing prices and real long term bond prices began.  At that point, when home prices really began to drop, home prices weren't falling back to a normal level.  They were moving away from longstanding equilibriums.

For decades, returns on homes (net rent as a percentage of home values) roughly matched 30 year mortgage rates minus an inflation premium (because homes are a real asset that gains value over time and mortgages are in fixed nominal terms.)

In 1995, homes were earning about 4% Net Rent/Price, expected inflation was about 3.5%, and 30 year mortgages were running about 8%.

In 2005, homes were earning about 2.5%, expected inflation was about 2.5%, and 30 year mortgages were running about 6%.

You get the idea.  Home returns and inflation generally added up to something in the ballpark of 30 year mortgage rates.  The returns in 2005 were a little on the low side, but within historical ranges.  (I suggested in this post that the chronic shortage of housing has led to a long term trend of above normal rent inflation, and that this could mean the inflation premiums for houses are higher than inflation premiums for, say, TIPS bonds, and that could explain the slightly lower home rent yields.)

Since 2010, homes have earned about 4%, inflation expectations have been about 2%, and 30 year mortgage rates have been about 4%.  Homeowners are earning as much on a real asset as the lenders are earning on a nominal asset.

This simply can't represent an efficient price level.  I mean, I suppose you could argue that because volatility in home prices has been so high, home owners are now earning a risk premium.  After all, stocks are real securities and they earn a premium over nominal bonds.  But, that's a bit of a circular argument.  To make that argument, you have to say that home prices had to fall and that the new low prices are justified by the fall itself.

So, the story that the fall in home prices was inevitable and reasonable has two elephants in the room that it must disregard.  (1) High rent inflation during a time period where homes were supposedly over-supplied, and (2) A home price level during the supposedly inevitable bust which is completely outside the realm of historic ranges or theoretical norms.


  1. There is no tension between rent inflation and an over supply of homes. Very few homes are built in locations and with layouts (or even with loans) that allow an easy transition between selling and renting. A new development in Arizona is barely going to impact rents in DC, nd will only marginally impact rents in Phoenix (the type of person who rents generally doesn't want to rent on the outskirtss of city). The resources that go into building that developement (labor, bricks, lumber, copper ect) could have been used to build new rental units in DC, and so over building of houses can lead to increased rents as new rental units are more expensive to build. It is very simple to model a situation in which the nation has an oversupply of houses and increasing rents.

    1. While it is true that renter inflation has been higher than owner equivalent rent inflation during the boom and bust, both have been high. I have a post that I should have up by Monday with BEA data which shows that nominal housing expenditures were level during the boom, but real housing expenditures (including imputed rent of home owners) have been declining sharply, as a portion of total consumption.

  2. I'm not too sure that this 2008 paper is useful, since it turned out that Fannie and Freddie were deliberately hiding the quality of the loans they were buying by mischaracterizing them on their balance sheets. Something the FM executives admitted to when the SEC sued them.

    Well explained by Peter Wallison in his recent book 'Hidden In Plain Sight'.

    1. Their data is loan level data from CoreLogic.