Not that I think everything was above-board, but to be fair, these are hugely correlated instruments: it's not reporting 93% of all AAA debt tanked, but 93% of all AAA-rated derivatives of subprime mortgages. Given the subprime-mortgage crash, it's not surprising that everything tied to subprime mortgages uniformly tanked too.
The way an AAA-rated derivative of a subprime mortgage could exist to begin with was via "tranching", where debt was sliced up into different buckets of repayment priority, which heightened the correlation. In a minor crash, the AAA securities would do okay, because they'd get the payments from the non-defaulting loans, and the lower-rated tranches would take the losses. But once you go over the threshhold where there isn't enough money to even repay the top tranches, then everything tanks all together. So really, near-0% and near-100% default rates were the likely outcomes.
The initial iteration is that they pool a bunch of loans together to aggregate the risk. They know from history (short term history - a vulnerability that turned into a failure mode) that mortgage defaults are quite unlikely to happen in the first, say 5, years, so they can sell paper that matures in 5 years and be confident that it has a low failure rate from the (flawed) historical perspective. The longer term paper on the loans will historically have a higher probability of failure, so the longer term paper has lower ratings.
Then the recursion occurs.
1) Homeowners kept refinancing their houses (encouraged by the insatiable appetite for new mortgages to slice and dice into AAA-rated paper), so their loans never aged, they just kept getting reset.
2) The bankers took the remaining less-than-AAA-rated aggregated debt and used to create a new debt vehicle and then claimed all over again that the "new" debt (which was really old debt reheated) could be re-sliced into AAA-rated debt plus residual lower rated debt.
This quickly turns into a house of cards which inevitably failed when short term history (low failure rates based on long held mortgages and "perpetually" increasing house values) failed to model reality.
Trick is, the ratings agencies knew 1 and 2 were happening; questions about the practice were raised.
That no-one did anything when questions were raised is what brings us back to fraud. The ratings agencies were allowing a known-flawed system to apply known-incorrect ratings because it served their own financial interests.
Excellent explanation. Also, I think it's relevant that the models that undergirded the ratings also had to assume some probability distribution of mortgage market price volatility... which was generally perceived as being far more stable than it turned out to be...
Yes, all (ex) AAA-rated derivatives of subprime mortgages are highly correlated. No, that (in itself) does not make them more risky. You just do not want to invest %100 of your portfolio in them.
E.g. U.S. Government debt is currently AAA rated. But all bonds and treasuries are highly correlated: in five years either 0% or 100% of US Gov debt will still be AAA.
Yeah, it's somewhat similar to weather predictions for nearby areas. If the weather forecast predicts 10% chance of rain for Palo Alto, and the same for East Palo Alto, Mountain View, Atherton, etc., the expected outcome is not that it will rain in exactly 10% of those places. It will probably rain in close to 0% or close to 100% of them.
The bonds would typically be a composite of loans from all over the US. So the argument was made that even if (say) California's housing bubble crashed then only a small fraction of your bond would suffer losses as the loans made in other states would be ok. Thus the banks claimed that these were uncorrelated safe investments and the rating agencies agreed to rate them highly.
There is some argument to be made that an awful lot of sophisticated investors didn't recognise before the crash that there was such a high level of correlation, so to some extent we're talking with the benefit of hindsight at this point. With hindsight it seems obvious that these bonds were doomed, but without the benefit of what we now know, how much should the ratings agencies be blamed for not recognising the risks?
I wonder if anyone from these rating companies ever bothered to talk to the people actually buying these mortgages - i.e. the people the whole crumbly edifice was based upon. Or indeed how they were being sold.
I would have expected someone to do some due diligence and find out what all of this business was actually based on.
Some guys from Lehman Brothers did some actual research (i.e. they went and talked to the sales guys selling these mortgages) and concluded that it was all going to end in tears. At according to:
Akadien has recommended "The Big Short" already, and I'd second that. It goes into a lot of detail about this situation. The problem with trying to price these CDOs was that an individual CDO might be made up of 100 pieces of other CDOs which each might be made up of parts of another 100 CDOs, which each might be made up of parts of a bond containing thousands of individual loans. So you could go and look at the individual loans but the effort to try and price the whole thing accurately would have been phenomenal.
The most honest thing the agencies could have done was refuse to rate them. But that probably wouldn't have gone down very well with their clients.
The way an AAA-rated derivative of a subprime mortgage could exist to begin with was via "tranching", where debt was sliced up into different buckets of repayment priority, which heightened the correlation. In a minor crash, the AAA securities would do okay, because they'd get the payments from the non-defaulting loans, and the lower-rated tranches would take the losses. But once you go over the threshhold where there isn't enough money to even repay the top tranches, then everything tanks all together. So really, near-0% and near-100% default rates were the likely outcomes.