Didier Sornette is not your average stock market forecaster. As a professor of geophysics at UCLA, Sornette spends a lot of his time analyzing natural occurrences, such as earthquakes. The author of "Why Stock Markets Crash" believes there are similarities between "complex systems" such as the structure of the earth's crust and financial markets.

The following is an excerpt from a conversation last week in which the professor seeks to explain those commonalities and why he believes they're applicable in forecasting both earthquakes and the financial markets.

Q: What does geophysics have to do with financial markets?

A: At UCLA we're doing earthquake modeling, studying them as complex systems. We're studying and predicting the stock market from the same point view. There are universals in complex systems.

Earthquakes are the results of stress relief and transmission between

tectonic plates that interact at a distance. In a similar way, the stock market is the result of the collective behavior of investors who are not isolated, not unconnected, but connected through the media and professionals talking to each other. This can be summarized from complex systems point of view as 'existence of interaction'.

Stock market dynamics result from an interplay between the idiosyncratic behavior of investors, which is a disordering force, but also the collective influence which tends to align investors in big groups -- bearish vs. bullish, trend followers, etc. -- which is an ordering force.

From that you see the emergence of relatively organized trading patterns.

Q: And from that 'interplay' you've deduced that investors act in similar, predictable ways before, during and after bubbles, which is why we should be concerned about the Japanese experience?

A: By looking at stock market prices of many markets, from different countries over several centuries, we found the market is not always efficient as academics would like it to be. Investor psychology, overconfidence, and interaction play a very important role in shaping the dynamics of stock markets. Herding probably plays a role which is increasing in last decade.

Q: Is your work related to Elliot Wave theory, which suggests human behavior can be charted in predictable ways using Fibonacci levels?

A: I don't fully endorse Elliot Wave theory. To me, Elliot Wave is more pattern recognition method

and not grounded in collective behavior. What is lacking in Elliot Wave theory and such charting is that I've not seen rigorous statistical tests

proving its efficacy.That said, I admire the work of people trying to detect patterns and respect the efforts and ability of the human brain to detect patterns; We started to work on applying complex theories to stock markets in 1996. Our first paper we referred to a superficial relationship between patterns we find and those elaborated in Elliot Wave theory. But maybe analysts/chartists following the Fibonacci numbers and trading on basis of their existence enhance their occurrence. It's a 'chicken and egg' question.

One problem many people have with Elliot Wave is this notion that chart patterns are somehow preordained, that we don't have free will. What does your work say about the inevitability of patterns you discuss?

TheStreet Recommends

A: Our theory outlines probabilities. It's not 100% deterministic and cast in stone. We try to identify a pattern, which expresses the collective behavior of investors interacting constantly. We try to calibrate a bubble and predict a crash but a crash is absolutely not a certain event. If we recognize a bubble is ripening we try to predict within a range of one month when a crash is most probable, but it's not a certain event. There's a 30% probability

that where when we predict a crash, it doesn't happen.

Our theory also accounts for the fact that many investors are able to realize the probability is rising and a crash may occur. It's rational to remain invested while a crash is looming, if the crash is not certain. But it's not a free lunch on a risk-adjusted return

basis. If people underestimate the risks, they will accrue big gains but suffer increasing risk.

Q: And are you predicting a "crash" in the U.S. today?

It looks like the market is approximately corresponding to

our models and should not move much

higher, then start a slow acceleration toward negative trends within the next quarter. Not a crash but the opposite of a crash, a change of trend

from rally to decline that will bottom in the summer of 2004.

I'm not in the business of giving advice, my interest is academic. But I'd certainly advise

investors to take into account the risk, and hedge against it in some way.

And what about beyond mid-2004? What is your longer-term forecast?

A: "I expect the market to start a new super bubble after 2005,

although the exact timing is very difficult: it could be 2006. This was chapter 10 of my book.

Aaron L. Task writes daily for TheStreet.com. In keeping with TSC's editorial policy, he doesn't own or short individual stocks, although he owns stock in TheStreet.com. He also doesn't invest in hedge funds or other private investment partnerships. He invites you to send your feedback to

Aaron L. Task.