Friday, May 2, 2008

Fishy

There's a lot of confusion floating around out there about the ultimate extent of losses on sub-prime mortgages, and accordingly about whether we are (were?) seeing a liquidity or a solvency crisis in the credit markets. The debate often seems to center around whether the big bank write-downs reflect economic reality -- loans people will not be able to repay -- or whether these write-downs are just a temporary fiction created by the Basel II rules for mark-to-market accounting in banks. For example, I have heard, repeatedly, that the situation is overblown because, "the write-downs that people are taking are based 100% of the mortgages defaulting and the lender only being able to recover 50% of their money through foreclosure". The latest salvo in this debate is an article in the FT which cites a recent Bank of England study:
In its latest financial stability report, the Bank suggests that the price of some AAA securities only makes sense if you believe that the ultimate loss ratio on securitised subprime mortages will be 38 per cent. However, that loss can only occur if three out of four households defaulted on their loans and house prices tumble dramatically too.
This intrigued me of course (who doesn't want to buy under-priced AAA securities, and so, despite being totally unqualified to do this, I tried to wade into the report in question to see what some of these numbers mean. The upshot is that I think the BOE should put the crack-pipe down and back slowly away from the discount window.

The key section of the report (and what the FT was citing) is the box on pages 17-19. The idea is that there are two ways to value sub-prime mortgages.

First, you can start with an estimate of what percentage of the homeowners will default and multiply this by what percentage of the loan value you can expect to recover. That is, you can guess. You can make an educated guess based on the history of defaults for these types of loans (which have existed for all of 4 years), and a forecast for housing prices. The BOE does this and comes up with $193b in defaults, which is around 20% of the $1 trillion of total sub-prime toxic sludge that's out there. The headline of the FT article is based on this number, which turn out to be approximately the level at which the triple A 'super-senior' tranche of the typical MBS would start to see some losses.

Second, instead of guessing, you can mark these loans to market by asking people at how much of a discount they would be willing to buy them. The problem with this is that nobody in their right minds wants to buy these things right now because they'd have to guess how much they're worth (in addition to the fact that these things are really complicated). So in lieu of having any bids, what people have done is indirectly estimate what the market for these loans would be like, if it existed, by asking the market how much it wants to charge for insuring these loans against default, and extrapolating the default rate from that (this is what the ABX index in the credit default swap market represents). This is admittedly a pretty roundabout way of doing things, but if you try it, you come up with a number somewhere around $375b, or close to double the first estimate. These numbers should be seen in light of the total write-downs realized to date of around $200b.

Now, to find out which of these estimates is closest to the truth you need to dig into the numbers a bit more. And the wheels simply fall of the first estimate when you find out how they did it. It is just an extrapolation of the default rates that have already occurred on these loans, coupled with the assumption that 75% of loans that go into default result in foreclosure and 50% of the initial amount is recovered. These assumptions are ludicrous though. The past default history of these loans is less than useless, as up till now house prices were rising and credit was available to refinance these loans. The future is almost guaranteed to be much worse than this. In addition, even in the best of times the lender only gets 50% of the value of a home back during foreclosure, after taking into account the legal fees, commissions, and missed maintenance costs. Are you really going to get 50% back if prices fall another 30% from where you made the loan? Finally, throw in the negative equity phenomenon where people have absolutely no reason not to walk away from their non-recourse mortgage, and you have every indication that the $200b figure you come up with via this approach is a serious underestimate. So the BOE and the FT jaw-boneing the over-reaction of marking to market, when in fact the market has so far only been marked to a serious underestimate of the total damage, is, to say the least, wishful thinking. They are pretending, along with the rating agencies, that anything stamped with a triple AAA will suffer no losses, and it just seems completely implausible.

Should we then conclude that the higher estimate provided by the market is correct? In general I have no problem accepting that markets do strange and irrational things over the short-term, so I don't necessarily take it for granted that this estimate is correct. Unfortunately, the $400b number is in the ballpark of what other people have been talking about for a while now based on admittedly sketchy, but much more convincing, methodologies for estimating the ultimate aftermath of the great housing bubble. So the conclusion of all this is that the next time you hear someone tell you that banks have exaggerated their write-downs and that the whole credit crisis is due to silly regulations about mark-to-market accounting -- the next time someone tries to convince you to buy AAA sub-prime bonds (sounds like "fresh-frozen" or "military-intelligence" doesn't it?) because, "their priced at 100% default and 50% recovery" -- the next time someone claims that the write-downs are over and the worst is behind us -- you should remind them that when you loan people 100% of the value of a property at the peak of the market, you shouldn't expect much more than half back.

No comments: