The WSJ asked UBS to prepare an analysis of bonds backed by Countrywide option ARMs, and has a big story about the results today. It's not entirely clear what information was available to UBS that wasn't available to the WSJ, but it does leave the poor reader not really knowing what to think: at one point, UBS and Countrywide are flat-out contradicting each other, and the WSJ gives no indication of which one is more plausible.
By 2006, nearly 29% of the option ARMs originated by Countrywide and packaged into mortgage securities had a combined loan-to-value of 90% or more, up from just 15% in 2004, according to UBS.
Of all Countrywide's option ARMs, including those kept by the bank as investments, fewer than 5% have had a combined loan-to-value ratio over 90%, a spokesman said.
There are two ways this discrepancy could conceivably be resolved. The first would be if Countrywide kept most option ARMs on its own books, and securitized only the most toxic. But earlier in the article we have already been told that Countrywide has $27.8 billion of option ARMs, while it securitized $122 billion.
The second possibility is that Countrywide wrote an enormous number of option ARMs with relatively low LTVs prior to 2004, then saw the proportion with high LTVs rise to 15% in 2004 and eventually 29% in 2006. But if you add up all the prior years, it's still less than 5% overall. This explanation doesn't really hold water either, since option ARMs were very much a niche product before 2004.
So we're left with a question mark hanging over the UBS analysis. And once it's there, other numbers spring out, and we ask ourselves whether we should believe them or not:
Of the option ARMs it issued last year, 91% were "low-doc" mortgages in which the borrower didn't fully document income or assets, according to UBS, compared with an industry average of 88% that year. In 2004, 78% of Countrywide's option ARMs carried less than full documentation.
All these numbers are enormous: 91% of option ARMs were basically stated-income loans? That's crazy, if it's true, and would prove that Countrywide was a particularly lax lender – but now I'm not sure how much I can trust these UBS numbers.
I'd love to see more detail on all of this. Specifically, since the UBS report was commissioned by the WSJ, I'd like the WSJ to simply post the report on its website, so we can read it and judge for ourselves, rather than having to rely only on the WSJ's journalists' précis. Any idea why the WSJ seems to be keeping the report to itself?
(Thanks to Mark Gimein for calling my attention to the article.)