The forum won't allow a direct link to the site you offer for reasons I don't know. say "XYZ dot com" in addition so it can be followed.
I apologize for my mistake. Hopefully this link (
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1367278) will work instead. The pagination is different from the article to which I linked earlier so I would say the most relevant material is probably located from 16-29 and 36-55.
What bankers called innovation was really something else entirely- technical rationalization.
I think there are two points here: 1) it is most definitely true that you can't rely solely on numbers, statistics algorithms or the like -- there must be a human hand in control. As someone who possess an avid interest in Sabrmetrics, this is an argument with which I am quite familiar and one whose validity I recognize; 2) Part of the problem with the Gaussian Copula risk model was that almost every market participant in the field of securtization used it to price and risk any particular product. This unwittingly increases coordination amongst market participants which, in turn, bellies one of central tenants of efficient market theory (market participants act independently of one another). The implication of this fact is that there is the likelihood of the market experiencing volatility is increased rather significantly. Although this series of coordination was brought about by market convention, government often introduces this sort of problem -- such as the widespread requirement for market participants to use VaR --as well (see, e.g.,:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1656075).
What you describe as the market turning irrational was exactly the opposite- it was an epiphany, a sudden realization that things weren't as they had seemed to be.
I disagree with this statement because, as I mentioned previously, many of the securitized assets held by the banks weren't, in the full course of time, that bad. To be sure, the lower tranches or equity sold to outside investors were wiped out but that was what was supposed to happen if the entire market went into a free fall. The super-senior tranches retained by the banks a massive downturn in value when the market was behaving irrationally just as it had behaved irrationally in inflating the bubble in the first place. Indeed, as I have mentioned before, AIG, which had entered into a huge number of CDS on such notes didn't have to pay out because they never actually defaulted. Rather, the panic in F08 temporarily depressed their value below its 'fundamental value'. This is not to say that all super-senior CDOs were good or that none defaulted -- just not nearly to the extent one would expect if you thought they were all garbage in the first place.
As to the article to which you linked, I have two thoughts:
a) The Notion that "Our technical capacity is considerable, but there is still much that is beyond us.":
In the first instance, the problem with residential CDOs is that the banks simply did not have the data necessary to formulate accurate models as to the correlative effects of defaults on underlying mortgages. This meant that, at best, banks could only guess as to how initial defaults would affect future rates of defaults. There guesses turned out to be wrong. As mentioned above, even with better data we should not become enslaved to models or the like. Indeed, those banks at which the 'humans' stepped in and overrode the models and algorithms -- such as JP Morgan and Goldman -- were precisely those banks that did the best during the crisis. That does change the fundamental point, that even the shortcomings of the bigger banks would have been avoided with better data.
If you turn to CDO based on corporate debt, you'll see that they fared much better during the downturn because there was so much more information on which banks could base their assumption as to correlative effect of corporate defaults.
So, in effect, I would suggest that the crisis does not demonstrate that there is some human element for which mathematics and models cannot account. On the contrary, it simply suggests that we must first get better data. As I mentioned earlier, this is a common argument amongst the baseball community regarding the use and value of Sabrmetrics.
b) "No one, it seems, imagined a financial disruption of a scale sufficient to degrade the value of all the underlying securities attached to the swaps"
I think this partially speaks to the bubble mentality associated with housing prices in the late 90's and 2000's. Equally, I think many of the people involved in AIG thought that if the stuff for which they were writing CDS ever actually did default then the entire financial sector would be in collapse, viz. AIG would have bigger problems than having to pay out on those contracts. And this frame of thought was necessarily wrong as many of the items for which they wrote CDS never actually defaulted. It was simply they never imagined that AIG would lose its 'AAA' rating and, when it did, AIG was not adequately capitalized to meet the collateral calls being made against it. The guy who was principally responsible for developing AIGFP's risk models, who was an MIT graduate, simply never assumed that AIG would lose that rating. This oversight does not signal a fatal flaw to the use of models rather highlights the need to use the right data and models.