Masaccio,
Is it too simplistic to say that the Wall St types look at their statistical analysis and models and see that 99 out of 100 times something good will most likely occur (good for them anyway) and go for it even though the 100th time things will blow up the previous 99 good.

I think they go for it and hope they can be gone before the 100th time smacks everyone in the a**.

But as always, I might be an id10t.

IMO pension plans and university endowments have no business investing in anything but listed and clearer securities, period. that would reduce their risk and coincidentally deny greedy wall street bastards getting their hands on all that cash and gearing it up.

From Naked Capitalism

Andrew Foland said…

Thanks for the word “Gadarene”!

Your exposition of the math was quite clear.

Personally, I’ve always suspected that most of the model makers understood that their various gaussian models were inadequate, but that most of them were thinking, “how wrong can it be?”

I also bet the “same models on the same data” aspects weren’t given the attention they deserved in the answering of that question.

Also, to be honest, I’ve trained physicists (in physics) for years, and even at the best schools, only the truly best of the best understand what it looks like, in detail, when you sample from a distribution of infinite variance. (The merely good ones would have a vague sense of unease about sampling from such a distribution; the bad ones probably think such a distribution can’t exist.) I know that Wall Street hired many more of my students than the number who understood it, and feel confident I’m not alone in that.”
July 5, 2009 8:24 AM

If the people creating the financial models don’t understand the mathematical theories behind them,were the models doomed to fail?

Why not just go to Las Vegas? 99 out of 100 times the house wins. But on that hundredth try . . .

probability and statistics are not rocket science. Basic probability is taught to college freshmen, and can be taught to high school students.

The problem came from misuse of statistics, and some of the quants advising Wall Street were negligent in how they explained their models, not demonstrating the potential of “low” risk events to have catastrophic consequences.

More likely, the events at the tail are mild. So, when one happens, it is likely that the losses will not be huge, even though they are unexpected. That isn’t always true. Some events are serious downers or uppers. Like last September or October, 1987, which wipe out years of gains. Others are really bad, and wipe out months or a year of gains.

The problem is that these events have not been properly taken into account in the pricing. Take a look at the paper on the Kiwi dollar. Ignore the math, let’s assume the rocket scientists get that right. Look at the last part, which shows the pricing differences. They are surprisingly big.

One of the regulatory changes that people ignore is that these funds used to follow rules about allocation of assets that closely mirror what you say. Those rules have fallen by the wayside, so that a number of endowments allocate large amounts to investments with large downside possibilities.

Taleb discusses this very issue in The Black Swan. The house can price the risk of a big payoff, because it the normal distribution actually works in casinos. The events are unconnected, so the law of large numbers and the central limit theorem work, as long as the bets stay small. Casinos do have to watch out for the giant bettors, called “whales”, who theoretically could hurt them. But the big risks are external to the gambling businesses, such as the $100mn one casino lost when an animal ate a showman. See page 129.

“The problem is that top management felt justified in ignoring the events at the tails of the bell curve, the very thin parts at each end of the bell curve.”

No. NO. NO. NO. Low probability events might follow a Poisson distribution. Not a Normal distribution. And then, only in a system with no or linear feedback.

And, there is an assumption that large external events are rare. A specific external event is rare (Mt St Helen’s, Flying Airplanes into Towers, etc). External events, assassinations, disasters man made or natural, deaths, scandals, etc. are not rare at all.

For example Ito’s Lemma, used to predict perturbations in the stock market (an insiders game, and better method would be to analyze insider communication, the NSA could help here), clearly states it is invalid for large external events. So Wall St Quants, with enthusiastic support from their management, went ahead and used it.

The real Governing Math is chaos or catastrophe theory, which is the behavior of systems with non-linear feedback.

A non-linear system? Us humans talking. As anyone who repeats gossip already knows. They never repeat it verbatim – that’s a non-linearity. (Womaen are non-linear, ask any man).

Chaotic systems always exhibit “tipping point” behaviour – sudden changes in the state of the system, to a new stability. The real fallacy is that non-linear system are manageable – they are not (Which questions that management is paid so much for what? Managing the unmanageable?). The best one can achive is to slow the rate of change.

My heroes are Orville and Wilbur Wright. They conceived and engineered discrete controls in an unstable, non-linear medium. No Nobel Prize going to them! Yet their model is still being used with the exception of flying
by wire(electronics)

“(Womaen are non-linear, ask any man).”

(Men are non-linear, ask any woman)

Fixed it.

My hero nomination is for the birds!

Casinos use linear systems with no feedback. (Feedback, counting cards, is called cheating).

Casinos based statistics have no relevance to everyday life, because we all live in a sea (pun intended) of interconnected system with non-linear feedback.

Non-linear system look like linear system over some part of their “range”. At the extremes of their range they change abruptly. An example of abrupt change is a person “reaching their limit” or “Going Postal”.

Both are non-linear – their behavioral non-linearities do not coincide. :-) There are other places to discuss physical non-linearities.

When in doubt, renormalize!

Indeed! We are all non-linear now!

By the way is there a model for the protection of the hoards of cash of us the unwashed?

People I like to trust, the liberal economists and financial bloggers occasionally drop what appears to be a throwaway line, ‘the eventual collapse of the dollar’. I don’t know how seriously I should take this. What should I do?

Can you recommend any reading? Blogs?

My point being, the Models used in Wall St have not failed. The Models Never Worked. A small application of Applied Math (Ask any engineer), would determine that Wall St was negligent, fraudulent and a criminal conspiracy.

Starting with Ignorance and Arrogance, which becomes a circular discussion.

Krugman:

http://web.mit.edu/krugman/www/
http://www.wws.princeton.edu/~pkrugman/

Roubini:

http://www.rgemonitor.com/blog/roubini/

You might try reading Bloomberg, the Financial Times and the Economist.

http://www.Bloomberg.com
http://www.ft.com (Buy a subscription, it’s worth it)
http://www.theeconomist.com (Always has snark at the end of the article, that’s the British way)

All the Gold bugs, however they have a not-so-hidden agenda, buy gold, which is a poor investment, and a safe haven.

When the Treasury did the stress tests it used those models.

The banks used their own models and data imputs for a couple of economic scenarios that Treasury had. What the economy has been doing was more in line with the worse case scenario. The banks then reported the results and because they still didn’t look sufficiently good, they were allowed to negotiate them further. Now if the economy gets worse which it likely will when this suckers market sputters out then reality may be worse than either of the scenarios used in the stress tests.

The giant banks have so much cash they can’t rely on individual bankers to estimate the risk of individual loans.

It’s not a case of a temporary oversupply of cash. The big banks aren’t making loans. They are holding some of the cash they have from the Treasury and Fed to make their balance sheets look better and they are gambling with the rest. And for years they haven’t been hiring loan managers who could actually assess the risk of individual loans. Models were cheaper and they weren’t intending to hold on their toxic sludge anyway.

Of course that’s true, even though top brass on Wall Street seems totally unaware of this obvious fact. Taleb says that Mandelbrot was talking about the problems with their models as long ago as the early 60’s. Page 260.

I read somewhere that David Li, who showed how to use the Gaussian Copula to value collateralized debt obligations, also tried to explain their limits, but was ignored. Can’t find it now.

I blame top management anyway. They should have known those limits. Instead, they all used it with sheep-like intensity, and went down together, until they were buoyed up by taxpayers.

Thanks for those refs. I look at the first two regularly, and Galbraith, and Dean Baker.

You mention the gold bugs. I have invested a bit from time to time. I have never found one that didn’t sound like a carnival barker.

I get the feeling that those you referenced are afraid of the word gold, afraid of sounding alarmist, or being tainted by the hustlers. Are there legit goldbugs?

The oil embargo
The S&L debacle
The Asian banking crisis
The Russian default
Japan and Japanification
The LTCM collapse
Junk bonds
The dot com bubble
The housing bubble
The oil and commodities bubble I
The oil bubble II
The financial meltdown
The current suckers’ rally in the stock market
The explosion in the Fed’s balance sheet
The building crisis in commercial real estate
The loan refinancing crisis that should hit in 2012-2014

These are just the ones that I can think of off the top of my head. The fact is that markets and the financial system are being perturbed every couple of years. The idea that there is some golden equilibrium or Gaussian distribution is hogwash.

The best source on competition to Dollar hegemony can be found by reading Brad Setser’s tireless work.
http://blogs.cfr.org/setser/

He used to be over at RGE Monitor.

Thanks masaccio.

Beware of geeks with formulas

– Warren Buffet

Even if you win 99 times in a row and keep putting it all back on the line that 100th time can take it all away.

I think what happened is the world changed (law and ideas) and they kept playing the same game with the same stupid ideas (ratings agencies paid for by debt salesmen) and the idea that if they just insured everything SOMEBODY would pay it off…meaning the taxpayers (who are already stretched to the limit) through the federal government.

They gambled, took all the winnings (for over a decade) and handed the losses to the public!

Two things. One, their models are more complex than you are allowing. Two, it’s easy to mistake where you are on a curve when someone else has lied to you about the data you’ve built your curve on.

Aren’t you confusing chaotic systems with systems that have catastrophes?

Also, in response to your statement at 13, there are non-linear systems that do not look like linear systems over a part of their range.

Also, to no one in particular, Axiom A attractors (a type of chaotic dynamical system) have gaussian statistics.

Finally, a question: Am I correct in believing that what this ‘multifractal analysis’ consists of is putting together simulated data using generators and multifractals, and using that, rather than a random number generator, to run Monte Carlo simulations to determine what a derivative instrument is going to do? If so, one can both be in favor of dumping the gaussian models and in favor of using derivatives, no?

And are people aware that computer random number generators are just chaotic dynamical systems that are sampled to produce different kinds of distributions?

Sorry for all the questions and comments, I get very muddled sometimes reading these things, about the fat tails especially since it seems to me that a distribution with fat tails (kurtosis) isn’t gaussian in the first place and the events in the tails are more probable than they would be in a gaussian distribution, check — but I can’t understand how anything in the distribution itself predicts the effect assigned to the event occurring (as in masaccio at 6).

I’m not that good at statistics, though.

The problem is that the financial elites don’t have even a layman’s grasp of the rocket science that underlies the models they use to gamble billions and to encourage their even more ignorant customers to throw money away.

Last week I described Nicholas Nassim Taleb’s explanation for the failure of the models used on Wall Street. In The Black Swan, he explains that all Wall Street models depend on the normal distribution, the bell curve, the Gaussian distribution. The problem is that top management felt justified in ignoring the events at the tails of the bell curve, the very thin parts at each end of the bell curve.

I have to disagree. IMHO, these are very smart guys who knew damn well what they were doing. Per Upton Sinclair: “It is difficult to get a man to understand something when his salary depends on his not understanding it.”

IMHO, these guys knew exactly what they were doing. The models assume that defaults on mortgages are not statistically correlated, which they would be if the housing bubble were to burst, and everyone knew that it had to burst eventually — what can’t go on forever won’t.