It's not often you find prophecies of stock market slumps on the website of a respected scientist. But that’s exactly what you will find on the homepage of Didier Sornette, a professor of geophysics at the University of California at Los Angeles.
What’s more, he’s making the prediction with similar models to those he uses to predict earthquakes. Sornette claims it’s likely that, by the end of 2003, the S&P 500 will begin to slowly and progressively plunge. During this acceleration into a bearish regime, the index will drop by around 30%, he claims; a recovery is likely to begin in early 2005.
External shocks – such as September 11 – can prompt extreme market moves. But Sornette also believes that crashes can be endogenous – emerging as an inherent property of markets. Sornette – who has collaborated with his colleague Wei-Xing Zhou in recent research – has attracted his share of the brickbats for some of his work in recent years. According to Imre Lakatos, a philosopher of science who died in 1974, scientists need to move outside existing orthodoxies to progress understanding significantly. Sornette not only follows this principle; he’s prepared to go out on a limb to prove it. “The best way to assess the validity of our endeavours is to announce our forecasts publicly,” he says.
Sornette claims that around two thirds of his previous predictions have come true. There are essentially two types of predictive error for crashes: failures to predict a true crash (missing events) and false alarms (prediction of a crash that does not materialise). Sornette says failures to predict crashes occur when there are exogenous factors at play, such as political coups, terrorist attacks or the outbreak of war.
His current prediction of a slump is in fact an update of a prediction made in December 2002. He says the earlier prediction was essentially correct, but that the beginning of the bearish regime in the original prediction was incorrect. In the interim, he has further developed his methodology, which is where the most recent prediction comes in. “Our model doesn’t necessarily predict crashes, it’s more about predicting a change in regime,” he says.
Sornette has had some intriguing past successes. For example, a public prediction he made in January 1999 that the Nikkei index would rebound by 50% over the remainder of that year was uncannily accurate – it rose by 49%. “We then predicted the market should turn down around February 2000 and depreciate significantly – just as it did,” Sornette says.
His work on complex systems – of which stock markets are an example – began in the 1980s, when he studied how matter ruptures under stress. He realised that before rupture, small defects – known as micro cracks – multiply and combine. Initially, the micro cracks are uncorrelated. But there’s positive feedback: cracks weaken the matter, causing more cracks, which then combine, until the matter weakens sufficiently to rupture. Sornette reasoned that this signature could be present in other complex systems. His application of the technique to earthquake prediction was as controversial as his recent research into financial markets.
When looking at stock markets, the analogue of the acoustic signature is that index prices move in a well-defined mathematical way – following a ‘log-periodic power law’. Similarly, the analogue to the multiplication and merging of cracks is the herding of traders.
Analysing market data for signals that predict future behaviour such as crashes remains highly controversial, and econophycisists such as Sornette are often criticisised by traditional economists. It’s not all inter-discplinary snobbery: some have questioned the statistical significance of the results presented in Sornette’s work. Also, his previous collaborator, Danish physicist Anders Johansen – who was involved in the earlier analysis of the Nikkei index – has publicly questioned some of Sornette’s recent work.
While the specifics of Sornette’s models are controversial, econophysics certainly has followers among leading financial firms. “The idea that there are instabilities in a market before a crash, similar to those seen when studying phase transitions in physics, is a good one,” says Michel Dacorogna, a former research physicist and now manager of financial analysis and risk modelling at Converium, a Zurich reinsurer. “However,” he cautions, “these models can suffer from many problems – for example, calibration is often very difficult”.
But while Dacorogna agrees that herding is a plausible explanation of the phenomena of crashes, others are unconvinced. Like Sornette, Xavier Gabaix, an assistant professor of economics at the Massachusetts Institute of Technology (MIT), is building a theory motivated by the observation that various stock market variables – including returns – exhibit a mathematical distribution defined by a power law. “It’s surprising when you first see the robustness with which stock market returns follow power law patterns,” says Gabaix, whose recent research was co-authored with MIT colleague Eugene Stanley, Boston University’s Vasiliki Plerou and Parameswaran Gopikrishan at Goldman Sachs.
Poring over a vast amount of trading data, the researchers found that power laws were present in the data of all the 11 major exchanges analysed. The ratio of the number of days a 1% price move is recorded, compared with the number for which there is a 2% move is eight: this is the cube of the ratio of the percentages. Volume data exhibits similar patterns. The inverse cube relationship has its root in the size distribution of large traders.
Unlike Sornette, Gabaix does not subscribe to the herding hypothesis. Instead, he and his co-authors claim that optimal trading by large institutions – such as mutual funds and hedge funds, for example – explain the form of empirical data. The team’s research also suggests stock exchange circuit-breakers – such as those operated by the New York Stock Exchange, where trading is suspended for specified periods, dependent on the magnitude by which an index plunges – may be ineffective.
Meanwhile, other research by Sornette is making the transition from whiteboard to dealing desk. He co-founded Insight Finance with Didier Darcet in 1999 to create systematic trading models. Barep Asset Management – a subsidiary of Société Générale – has coughed up $12 million of seed capital this year to begin trading the models in a new joint venture. Risk understands that the internal funds are breaking even, despite the recent decline in European stock market volatility. Maybe those sceptical about econophysics should think again. Then again, we’ll know soon enough whether Sornette’s latest forecast is correct. Just wait for the end of the year.
The week on Risk.net, July 7-13, 2018Receive this by email