Why stock markets crash: A scientific perspective

Zach Alexopoulos, The University of Sydney

Didier Sornette Why Stock Markets Crash: Critical Events in Complex Financial Systems, Princeton, Princeton University Press, 2003 (421 pp). ISBN 0-69109-630-9 (hardback) RRP $62.90.

Why Stock Markets Crash is a difficult book to categorise. It is not a get rich quick guide, nor is it a standard financial economics text or a scientific monograph. It is aimed at the general reader, but reads like a deep scientific text while tackling a classic economic problem. This diversity of influences arises because the author is working at the rapidly expanding interface between economics and the natural sciences.

The book is immediately distinctive because of the background of its author. Didier Sornette is not an economist, but a geophysicist. His previous work has concentrated on the behaviour of complex natural systems under stress. Sornette’s perspective as an outsider is part of what makes the book such interesting reading—he uses the cutting edge tools of the natural sciences, most of which have not found their way into economics. In particular, he applies some advanced mathematical methods used to analyse catastrophic natural events such as earthquakes and volcanic activity, to examining the great catastrophe of modern economies: the stock market crash.

On methodological grounds alone, Sornette’s book is a valuable addition to the financial literature, because most economic studies use old-fashioned linear models mostly long abandoned by the natural sciences. The few nonlinear models actually used in finance do not fit the data particularly well, as Sornette points out. They especially underestimate the risk of crashes.

Sornette’s main argument is that crashes are not—as the media propounds after they happen—the result of immediate news-related ‘causes’ alone. Instead, crashes are the result of growing instability in the market system, the outcome of a long build-up of herd behaviour that forces the market into an increasingly precarious position. Once the system has reached a ‘critical point’ where instability is at its peak, even a small stimulus can cause a crash. However, the real blame should lie not with any news-related stimulus, but with the unstable nature of the market system under stress.

The first part of Sornette’s thesis is that financial crashes are ‘outliers’, which have their own statistical signatures. This departs from the standard finance view that big price drops are just ‘small drops that did not stop’ (p. 26), with no other distinguishing or predictable features. The specific statistical character of crashes means that they must be analysed separately from other market movements, but it also means that they can be tracked, and to some extent predicted. To demonstrate this, Sornette establishes the distinctive character of large market movements, using examples from markets across the world.

The second plank of his argument is that before crashes, market movements can be approximately described by a process known as log-periodicity. Roughly speaking, log-periodicity is a series of oscillations that become more and more rapid before a crash. Log-periodicity is significant because it seems to emerge only in the lead-up to a crash and not at other times; hence its presence is an important signal that a crash may be imminent. Sornette specifies and tests gradually more refined models of the process on crashes in different markets throughout the world, making and testing predictions about real markets movements. This is the most mathematically demanding part of the book, but Sornette tries very hard to explain the results intuitively, and keeps the equations to a minimum in technical inserts.

Sornette’s findings provide fascinating insight into dynamic market behaviour under stress.

Interestingly, the crash predictions Sornette makes using log-periodicity models actually test out reasonably well, but his framework is still a long way from forecasting crashes. The market’s direction cannot be predicted with these models at any time frame more than a year out, and even with detailed price data, they can only predict a crash to within a few weeks, if it happens at all. Only three out of six of Sornette and his past coauthors’ public predictions of a crash have come to pass. As Sornette points out, that track record is not as statistically unfavourable as it looks, given how few months crashes actually occur in, but the framework is still far from completely robust. Even so, Sornette’s work is likely to improve economists’ ability to track crashes in advance.

As well as being highly original, Sornette’s findings provide fascinating insight into dynamic market behaviour under stress. The behaviour of stock markets is not smooth, rational and self-adjusting, as standard equilibrium-based market models would have it. Instead, they are subject to positive feedbacks (price changes feeding on themselves), they often move in the absence of any changes in information, and they behave chaotically under stress. This research is a clear advance on how to understand market behaviour. Indirectly, this book underscores why it is valuable to study financial markets. If the most liquid markets in the world—perhaps the closest thing we have to a truly free market—can behave in such a volatile, out-of-equilibrium fashion, and generate such unstable outcomes, what does that imply about the quality of markets as a method of economic organisation more generally? All economists, indeed, anyone with an interest in economic matters, should be concerned about this question.

Sornette concludes with an interesting prediction for the next century, making a bold call for the end of the ‘growth era’ around 2050. His prognosis is based on the observation that crashes tend to take place after the market has spent some time increasing at increasing rate—following a log-periodic trend. Crashes aside, during the last two centuries all major financial indices have advanced at a faster-than-exponential trend rate, as have population and GDP growth. The log-periodic model implies that this trajectory is unsustainable and that the world economy, as well as the market, may cease to grow or even slump heavily in the middle of the next century. Such predictions are worth keeping an eye out for in 50 years time. However, nonlinear dynamic models like the log-periodic one Sornette uses tend to lose predictive power over long horizons fairly quickly, so we should view his prediction somewhat sceptically.

The book focuses on the observed aggregate behaviour of markets as systems; it does not really address the question of what makes individual traders act like they do. Unfortunately, in this way, it resembles standard finance arguments—expressed with dazzling mathematical beauty and impressive quantitative firepower, but with people almost entirely left out of the analysis. A commonly-heard quip is that most finance theories would not change much if markets were populated solely by robots. Sornette puts the question of underlying causes into a black box of blaming the sometimes unstable nature of the market system. But if volatility is the fault of the system, then what factor(s) are causing instability? Sornette surveys the economics literature on ‘rational bubbles’ and some of the newer ‘behavioural finance’, in an attempt to cover the established theories of trader behaviour, but he doesn’t reach any firm conclusions. I happen to be quite partial to the behavioural finance literature, which uses insights from cognitive psychology to help explain investor behaviour and thereby brings people back into the analysis, and it may provide some answers as to how and why bubbles begin to inflate and deflate. For interested readers, Shleifer (2000) is a good overview of the field.

Within the behavioural literature, recent research into why bubbles arise and, more pertinently for our purposes, why they burst, complements Sornette’s argument well. Abreu & Brunnermeier (2003), for example, reject the orthodox finance argument that bubbles burst when rational traders finally obtain the upper hand in the market and move prices back to their appropriate level. Rather, they argue, bubbles burst simply because a herd of uninformed traders all happen to react at the same time to the same piece of information, or to some other stimulus such a price fall below a ‘psychological barrier’. Trader behaviour may be the immediate ‘cause’ of the crash, after the system has reached a critical point. Further research along these lines is an obvious way forward from Sornette’s findings.

Why Stock Markets Crash is not for those whose eyes glaze over at the sight of a pronumeral.

However, interesting as its empirical work and mathematical setting are, the book’s argument suffers from a few theoretical problems irksome to one familiar with recent advances in financial economics. The most frustrating is that Sornette places excessive faith in the concept of stock market efficiency. This is the notion that a stock’s price accurately reflects all available information about the stock. Efficiency is said to be enforced by arbitrage, that is, by smart investors trading against the market to correct inaccurate prices. While formerly considered almost a gospel truth within the finance profession, market efficiency has come under sustained empirical and theoretical attack over the last two decades (see, for example, Lamont & Thaler (2003) and Rashes (2001) for some spectacular examples of the failure of arbitrage to correct unambiguous instances of mispricing). Nor does Sornette consider the implications for market efficiency of the very existence of crashes, which often occur in the absence of any news events. In fact, the crash of 1987 was responsible for a new wave of literature casting doubt on the theory of efficient markets. For a book that is so refreshingly new in other ways, its subscription to an old theory on the empirical back foot is disappointing.

Further, Sornette seems unaware of some not-so-recent research demonstrating emphatically that stock price movements are not random, but are somewhat predictable over short and medium term horizons (the classic references are Lo & MacKinlay (1988) for the short term, and De Bondt & Thaler (1985) for the medium term). He uses the standard ‘random walk’ setting constantly, even after demonstrating that large price changes take place far too often for market movements to be truly random, given the market’s average return and standard deviation. This is a problem with Sornette’s theory, given that random prices are supposed to be a sign of market efficiency, and randomness is the major plank of the evidence for efficiency he uses. Admittedly, the random walk assumption does not seem to affect the results of the empirical work, but given the otherwise boldly original nature of the book, again, the use of an empirically questionable framework is jarring.

Another shortcoming of the book is that it does not much address what might be done to contain crashes. Sornette seems to subscribe to a mild laissez-faire approach, leaning towards the view that it is better not to interfere in the operations of a complex system, lest the intervention precipitate an even more catastrophic outcome later. At one point he draws an analogy with forest fires, noting that when fires are allowed to burn without hazard control, they paradoxically tend to be smaller and cause less damage than fires that have been deprived of fuel through backburning. Similarly, the use of ‘circuit breakers’ and enforced trading halts in financial markets have generally been found not to reduce market fluctuations and are actually associated with an increase in volatility. While this half-answer is unsatisfying, it is the only one Sornette provides. This is clearly an avenue for future research in light of the book’s findings on market dynamics.

Why Stock Markets Crash is not a book for those who want to learn how to cash in by bailing out of the market the minute before a crash. Nor is it suitable for those whose eyes glaze over at the sight of a pronumeral. But it is an interesting application of new scientific tools to an old economic problem, and it provides clear lessons about how financial markets behave under stress. Despite its theoretical limitations, it is to be hoped that members of the finance and economics professions will attend both to the empirical findings of this book, and to its distinctive method of undertaking financial research.

REFERENCES

Abreu, D. & Brunnermeier, M. K. 2003, ‘Bubbles and crashes’, Econometrica, vol. 71, no. 1, pp. 173–204.

De Bondt, W. F. M. & Thaler, R. H. 1985, ‘Does the stock market overreact?’, Journal of Finance, vol. 40, no. 3, pp. 793–805.

Lamont, O. & Thaler, R. H. 2003, ‘Can the market add and subtract? Mispricing in tech stock carve-outs’, Journal of Political Economy, vol. 111, no. 2, pp. 227–268.

Lo, A. W. & MacKinlay, A. C. 1988, ‘Stock prices do not follow random walks: Evidence from a simple specification test’, Review of Financial Studies, vol. 1, no. 1, pp. 41–66.

Rashes, M. S. 2001, ‘Massively confused investors making conspicuously ignorant choices (MCI-MCIC)’, Journal of Finance, vol. 56, no. 5, pp. 1911–1927.

Shleifer, A. 2000, Inefficient Markets: An Introduction to Behavioural Finance, Oxford University Press, New York.

Zach Alexopoulos is an honours student in political economy at The University of Sydney. His thesis deals with the causes of stock market volatility and market-level policies for reducing it. Thanks to Rita Bhattacharya and Roni Demirbag for helpful comments.

View other articles by Zach Alexopoulos: