Why Wall Street Flubs

Here is an hour lecture at MIT by Eric Rosenfeld, an MIT graduate who in the early 1980s helped create the new quantitative strategies for Wall Street risk management and hedging strategies. He is fairly unapologetic for the failings of Wall Street despite the collapse of LTCM and the recent credit crisis/market collapse.

He begins the lecture by claiming that the critics of Wall Street got it all wrong and that LTCM was a sound company built on a sound quantitative strategy build into their risk modeling and hedged trading strategy that had "proven" itself for ten years at Salomon Brothers. He runs down a list of "myths" about LTCM that he wants to knock down:

• Greed and hubris brought LTCM down. LTCM believed in its own "invincibility".


• LTCM used 100-1 leverage.


• It used "black box" models.


• They demonstrate the perils of applying academic theory in the hard reality of the market place.

But at the end of his talk, he does admit that LTCM was bound to fail. He admits to these weaknesses:

• The models were not always sound. He gives a story about how some of the quantitative models were built by "rocket scientists" who didn't have good intuitions about market dynamics.


• He admits that leveraging is bound to fail. It may work fine for 9 years, but it will fail in the 10th year when "the unexpected" happens. (Actually LTCM lasted just 4.5 years, so even here Rosenberg is overly optimistic.)

As I watched and listened carefully, I came away convinced that the "myths" were not myths. They were truths. But somehow Eric Rosenfeld thinks they are myths. That astounded me.

The very facts presented in his talk can be marshalled against his claim of four "myths" about LTCM. Here is what I gleaned from the talk:

• It wasn't "greed" that killed LTCM, but there was a fair amount of hubris which led them to take positions that their "risk analysis" showed to be safe. In fact Rosenfeld walks you through their daily valuation and claims that this shows that their approach was sound. This from a guy whose company failed. My definition of sound doesn't include "works most of the time but fails under 'exceptional' circumstances". And, the 'exceptional' circumstances of 1998 don't sound like a 'once in a hundred year' he claims. He admits they would have failed in 2008. I'm 100% sure they would have failed in 1929. I'm pretty sure they would have failed in 1914 when the markets closed for months. I would bet they would have failed in 1987 when programmed trading caused a crash of the stock market. (He didn't discuss this, but I'm convinced that having Scholes and Merton on the board of directors gave the company these sense that they were "untouchable". They were building strategies using the best models out there such as the Black-Scholes which won their inventors a Nobel Prize.)


• In his talk he wants to claim they were "only" leveraged 30:1 in normal times but that in 1998 it ballooned to 300:1. So the critics talking about 100:1 doesn't sound like a "myth" to me.


• Near the end of the talk he complains about a guy hired from CalTech as a real "rocket scientist" who started at $50K/year to build models who couldn't build models and who went on after the collapse of LTCM to make $500K/year duplicating the risk models of LTCM/Salomon at other Wall Street firms. Rosenberg says he gave this "rocket scientist" a problem and the models produced were "garbage" while the same problem given to an MIT student (his alma mater and the crowd he is talking to) was partially solved in a day with a "good" model. But the fact is, Rosenberg has -- by his own anecdote -- convicted the LTCM group of using black box models built by "rocket scientists" who didn't understand what they were modelling. Moreover the "good intuitions" of the MIT kid couldn't have been that good because the good models still let LTCM blow up in 4 years!


• Rosenberg is quite angry at the claim that academic theory doesn't apply to the market. But this in fact is an issue hotly debated in academia today. The models of "rational economics" have proven to fail to fit reality. The very risk models that Rosenberg talks about failed LTCM. He constantly refers to using "uncorrelated" bets to bring down the level of risk, but the LTCM crisis, the current crisis, and all major crises demonstrate that when things go really bad, what is "normally" uncorrelated becomes correlated. He admits this fact, but he still clings to his "risk management" calculations based on uncorrelated bets in the marketplace. He puts down the "black swan" claims of Nassim Nicholas Taleb. The behavioural economists have undercut the idea of rational economic agents by producing empirical proof and developed theories of irrational human behaviours.

Finally, I was amazed at how passive this MIT audience was. They are living through an historical market meltdown. They are being lectured by a guy who participated in a mini-meltdown in 1998. Yet none of them challenge him in his claims of expertise. This is supposed to be the "cream of the crop", bright students, self-actualized, and critically rational. But I see no evidence of this in their interaction with Rosenberg. Amazing!