Sunday, June 08, 2008

The right House Price Index for Analyzing Fannie/Freddie's Finances

In this month's Atlantic, Bob Shiller has a thought-provoking piece on how to attenuate the mortgage meltdown. In that piece, however, he juxtaposes Freddie Mac Chief Economists' Frank Nothaft's statement that the company has modeled the company's financial health assuming a 13.8 percent decline in house prices with his own that real house prices have fallen by 15 percent from peak to trough.

From the standpoint of mortgage performance, it is nominal, not real, house prices that matter, because mortgage balances do not adjust to changes in the price level. So long as nominal prices rise, the incentive to default is low, because home equity will be increasing.

Just as problematic, Shiller is applying the Case-Shiller Index to make conclusions about Fannie/Freddie performance. But for evaluating Fannie/Freddie, the OFHEO purchase index is best, because it only contains loans that Fannie/Freddie actually fund, and because it is based only on transactions. This index has fallen 3.1 percent from peak so far. This is nothing to celebrate, but it is well within the realm of what Freddie modeled.

4 comments:

Anonymous said...

The universe of transactions may favor the OFHEO index over the Case-Shiller, but the weighting scheme favors the use of Case-Shiller. OFHEO is unit weighted, while C-S is dollar weighted, so that C-S gives a lot more weight to California. And the failure of a $300K California loan costs you twice as much (roughly) as the failure of a $150K Ohio loan.

Also, the OFHEO index might be more applicable to their normal business, but C-S might be more applicable to their non-agency portfolio investments.

Of course, nominal makes a lot more sense than real.

Richard Green said...

Except that Fannie's non-agency business is less than 4 percent of its total business. See http://www.fanniemae.com/ir/pdf/monthly/2008/043008.pdf . And Freddie's non-agency business is about 10 percent. See http://freddiemac.com/investors/volsum/pdf/0408mvs.pdf.

On the other hand, the point about dollar weighted vs. unit weighted is a good one.

Anonymous said...

Richard: Your main point is well-taken, but Freddie's figures look a little different from the OFHEO figures you cite. Here's the first sentence of Freddie's May 29, 2008, press release ("NATIONAL HOME VALUES FALL IN FIRST QUARTER"):

"Freddie Mac (NYSE: FRE) announced today that its Conventional Mortgage Home Price Index (CMHPI) Purchase-Only Series registered a 10.4 percent drop in U.S. home values during the first quarter of 2008 on an annualized basis, following a downward revised 9.9 percent annualized drop in the fourth quarter."

Don Coffin said...

One thing to note (in re Mark O.'s comment. A 10% annualized decline in the 4th quarter is really a 2.5% decline during the fourth quarter. Similarly, the 9.9% annnualized decline in the first quarter is a 2.5% decline in the first quarter. That's a 4.9% cumulative decline in the fourth and first quarters...not a 20% decline.