The Chinese math formula that killed Wall Street.

G

Grandluxe

Recipe for Disaster: The Formula That Killed Wall Street
By Felix Salmon

<img src='http://www.wired.com/images/article/magazine/1703/wp_quant4_f.jpg'>

Enter Li, a star mathematician who grew up in rural China in the 1960s. He later moved to Barclays Capital and by 2004 was charged with rebuilding its quantitative analytics team.

In 2000, while working at JPMorgan Chase, Li published a paper in The Journal of Fixed Income titled "On Default Correlation: A Copula Function Approach."

For five years, Li's formula, known as a Gaussian copula function, looked like an unambiguously positive breakthrough, a piece of financial technology that allowed hugely complex risks to be modeled with more ease and accuracy than ever before. With his brilliant spark of mathematical legerdemain, Li made it possible for traders to sell vast quantities of new securities, expanding financial markets to unimaginable levels.

His method was adopted by everybody from bond investors and Wall Street banks to ratings agencies and regulators. And it became so deeply entrenched—and was making people so much money—that warnings about its limitations were largely ignored.

Read more: http://www.wired.com/techbiz/it/magazine/17-03/wp_quant?currentPage=all
 
I learned, had it forcibly instilled in me actually, in electrical engineering studies that 1) yes we use simplifying assumptions in order to make something tractable 2) we must keep track of said assumptions because they change the very nature of what we are doing sometimes. Bankers didn't learn that.

With regard to correlations, an intern with Excel can give you a lot of correlated issues to trade but knowing what will cause them to stop being correlated is a whole 'nuther thingie..

Putting all of the problems in that formula on top of the ghetto loans being given high ratings and packaged into the real world's loans was an exquisite screwup if ever there was one...

I could imagine screwups for say, writing a science fiction script, for a long time and I might not think of things that were screwed up and then matched together that badly.. it would go something like this: "hmmm we know that these idealistic leftists can't do anything but screw up economies... and what if we have a character that is a "genius" providing a black box for trading and the politicians .... and the bankers...." Honestly, I like using my imagination to make up scenarios, and I couldn't come up with such an exquisitely compiled scenario as what really happened...
 
Quote from Fractals 'R Us:

I learned, had it forcibly instilled in me actually, in electrical engineering studies that 1) yes we use simplifying assumptions in order to make something tractable 2) we must keep track of said assumptions because they change the very nature of what we are doing sometimes. Bankers didn't learn that.

With regard to correlations, an intern with Excel can give you a lot of correlated issues to trade but knowing what will cause them to stop being correlated is a whole 'nuther thingie..

Putting all of the problems in that formula on top of the ghetto loans being given high ratings and packaged into the real world's loans was an exquisite screwup if ever there was one...

I could imagine screwups for say, writing a science fiction script, for a long time and I might not think of things that were screwed up and then matched together that badly.. it would go something like this: "hmmm we know that these idealistic leftists can't do anything but screw up economies... and what if we have a character that is a "genius" providing a black box for trading and the politicians .... and the bankers...." Honestly, I like using my imagination to make up scenarios, and I couldn't come up with such an exquisitely compiled scenario as what really happened...
More...

According to Nassim Nicholas Taleb, "People got very excited about the Gaussian copula because of its mathematical elegance, but the thing never worked. Co-association between securities is not measurable using correlation"; in other words, because history is not predictive of the future, "anything that relies on correlation is charlatanism."

If history is not predictive of the future then what is backtesting for?
 
It's all in the details. Which is why general statements/observations are worthless in this business.

Quote from Grandluxe:

According to Nassim Nicholas Taleb, "People got very excited about the Gaussian copula because of its mathematical elegance, but the thing never worked. Co-association between securities is not measurable using correlation"; in other words, because history is not predictive of the future, "anything that relies on correlation is charlatanism."

If history is not predictive of the future then what is backtesting for?
More...
 
Quote from Fractals 'R Us:
I learned, had it forcibly instilled in me actually, in electrical engineering studies that 1) yes we use simplifying assumptions in order to make something tractable 2) we must keep track of said assumptions because they change the very nature of what we are doing sometimes. Bankers didn't learn that.
More...
Of course, it's very nice to think that people involved were idiots (some were indeed), but it's only a part of it. There is a lot of politics to running a structured product business and where the book gets marked is a big part of it. There is like a million ways to estimate where various non-observable correlations/spreads/offsets should be marked. The sad reality is that if your assumptions are too conservatively vs. your peers you'll never win any trades; if your assumptions are too aggressive, you will blow up when the known substance hits the fan. Some people felt that correlations in the credit world were too low before "the crisis", yet it was nearly impossible to convince your manager to mark the book down.

A few large firms resorted to "global hedges", many of which resulted in disasters (some trades, like the large RMBS equity/mezz ratio that lost MS about nine billion, are pretty funny in retrospect) and some made money (e.g. Goldman trade). A few firms decided not to get into the structured credit business at all (wise move) and some just loaded up on a variety of cheap risk-premia. I remember buying some 2y 2s/10s CMS curve straddles (free carry, as the curve was steeper in the forward space) and various CDS spreads on AA companies (hedging my funding risk at 12bpa) in the summer of 2007 and my boss telling me "it's a waste of money".

PS. I have never ran a structured credit business but I was responsible for a large hybrid derivatives book that had similar features in terms of correlation assumptions and a bunch of Finagle's Constants. It's about as much fun as a rectal exam and as confusing as the Fathers Day in a small village.
 
That old saying holds as true as ever.

" If you don't understand it then leave it alone "

Some people get hoodlywinked when it comes to " black box " methodologies. Usually get blown away when it all goes phut.
The sharp boys selling this sort of stuff are miles away when it blows up.
 
I have not posted in a while, but that article is nothing more than misdirection and misinformation of the apparently not so obvious. The world of finance, beyond the basics, is not some wild mathematical model, it's human psychology, emotion & manipulation.
The bottom line is that this whole CDO/MBS mess was just another big scam, as the whole rating system was mostly an illusion. The ones relying on this very silly formula that does not, at all, account for why such a massive debt market even exists, let alone how it truly functions, are just puppets or acting the part.
 
You must log in or register to reply here.
Back
Top