Evidence-based investing research
Value Investing Strategy (Strategy Overview)
Allocations for November 2026 (Final)
Cash TLT LQD SPY
Momentum Investing Strategy (Strategy Overview)
Allocations for November 2026 (Final)
1st ETF 2nd ETF 3rd ETF

You Should Look at Didier Sornette’s Work Again

Steve LeCompte | | Posted in: Big Ideas

Key Insight

Sornette's log-periodic power law model reportedly predicted Chinese stock bubble peaks in 2007 and 2009, but lacks demonstrated economic value for investors and raises data snooping concerns (ex-post analysis; no strategy returns tested).

A reader suggested: “I know you’ve looked at Didier Sornette’s work in the past, but I think it would be worthwhile to look at his work again. His latest is ‘Bubble Diagnosis and Prediction of the 2005-2007 and 2008-2009 Chinese Stock Market Bubbles’, with abstract as follows:”

“By combining (i) the economic theory of rational expectation bubbles, (ii) behavioral finance on imitation and herding of investors and traders and (iii) the mathematical and statistical physics of bifurcations and phase transitions, the log-periodic power law model has been developed as a flexible tool to detect bubbles. The LPPL model considers the faster-than-exponential (power law with finite-time singularity) increase in asset prices decorated by accelerating oscillations as the main diagnostic of bubbles. It embodies a positive feedback loop of higher return anticipations competing with negative feedback spirals of crash expectations. We use the LPPL model in one of its incarnations to analyze two bubbles and subsequent market crashes in two important indexes in the Chinese stock markets between May 2005 and July 2009. Both the Shanghai Stock Exchange Composite and Shenzhen Stock Exchange Component indexes exhibited such behavior in two distinct time periods: 1) from mid-2005, bursting in Oct. 2007 and 2) from Nov. 2008, bursting in the beginning of Aug. 2009. We successfully predicted time windows for both crashes in advance with the same methods used to successfully predict the peak in mid-2006 of the US housing bubble and the peak in July 2008 of the global oil bubble. The more recent bubble in the Chinese indexes was detected and its end or change of regime was predicted independently by two groups with similar results, showing that the model has been well-documented and can be replicated by industrial practitioners. Here we present more detailed analysis of the individual Chinese index predictions and of the methods used to make and test them.”


Some key points from the paper you cite are:

  1. “In a nutshell, bubbles are identified as ‘super-exponential’ price processes, punctuated by bursts of negative feedback spirals of crash expectations. These works have been translated into an operational methodology to calibrate price time series and diagnose bubbles as they develop.”
  2. “Here, we present an ex-post analysis of what we identified earlier in their respective epochs as being two significant bubbles…”
  3. “Our main method for detecting bubbles and predicting the critical time…when the bubble will end either in a crash or change of regime is by fitting observed price time series to a log periodic power law model. This is a stochastic fitting procedure that we complement with other techniques… This philosophy of using multiple measures aids in filtering predictions, in that a candidate prediction must pass all tests to be considered worthy.”
  4. “…we stress that, notwithstanding the common use of the term ‘crash’ to refer to the aftermath of a bubble, a real crash does not always occur. Rather, the end of a bubble may be the most probable time for a crash to occur, but the bubble may end without a splash and, instead, transition to a plateau or a slower decay. This point is actually crucial in rational expectation models of bubbles in that, even in the presence of investors fully informed of the presence of the bubble and with the knowledge of its end date, it remains rational to stay invested in the market to garner very large returns since the risk of a crash remains finite.”
  5. “…we predicted both crashes with our techniques before the actual dates of the observed peaks…”

Reactions are:

The methods of the authors appear to be highly empirical and involve considerable experimentation through (1) fitting of data to model through parameter selection and (2) application of other filtering tests. Do the methods described fully mitigate data snooping bias?

Skepticism is in order when forecasters report their own past successes. Confirmation bias (and purposeful deception) can result in overemphasis of past successes and underemphasis of (or silence about) past failures. The iterative failures of related methods described in the blog entry you cite (see also “Testing the Stability of the 2000-2003 US Stock Market ‘Antibubble'”) support skepticism about the accuracy rate for a meaningful sample of forecasts.

The analyses do not apply bubble-related forecasts to returns from any investing strategies. In other words, the authors do not demonstrate any economic value of their forecasts for investors (and imply in the fourth point above that there may be none).

Measurement of investing strategy outcomes for a reasonably large sample of ex ante forecasts of bubble bursts from the authors would be interesting.

Frequently Asked Questions

What is the log-periodic power law (LPPL) model for detecting stock market bubbles?

The log-periodic power law (LPPL) model is a bubble-detection tool that treats a faster-than-exponential (power law with finite-time singularity) rise in asset prices, decorated by accelerating oscillations, as the main sign of a bubble. It combines rational expectation bubble theory, behavioral finance ideas about herding, and statistical physics of phase transitions. It reflects a positive feedback loop of rising return expectations competing with negative feedback spirals of crash expectations.

Can the LPPL model predict when a stock market bubble will end?

The LPPL model's authors report that it identified time windows for two Chinese stock market bubble peaks, one bursting in October 2007 and another in early August 2009, in the Shanghai Composite and Shenzhen Component indexes. However, these are self-reported, ex-post analyses, and the authors do not test any investing strategy returns, so the model's economic value for investors is not demonstrated. The evidence for reliable predictive accuracy across a meaningful sample of forecasts remains unestablished.

How is the LPPL model used to diagnose a bubble and forecast its critical end time?

The LPPL model is applied by fitting observed price time series to the log-periodic power law form through a stochastic fitting procedure, which estimates the critical time when the bubble will end in a crash or change of regime. The authors complement this fit with other techniques, and a candidate prediction must pass all of these tests to be considered credible. The end of a bubble may bring a crash, a plateau, or a slower decay rather than always a crash.

Does a bubble ending always mean a stock market crash, and is it irrational to stay invested in a bubble?

A bubble ending does not always mean a crash, according to the LPPL authors: the end may be the most probable time for a crash, but prices can instead move to a plateau or decline slowly. Under rational expectation bubble models, it can remain rational to stay invested even when investors know a bubble exists and when it will end, because the crash risk stays finite and the returns before the end can be very large.

What are the limitations and criticisms of using the LPPL model to predict bubbles?

The LPPL approach is highly empirical, relying on parameter selection and multiple filtering tests, which raises data snooping bias concerns. Forecasters reporting their own past successes may suffer confirmation bias and underemphasize failures, and related methods have shown iterative failures, such as the 2000-2003 US stock market antibubble. The analyses also apply no bubble forecasts to investing strategy returns, so any economic value for investors is unshown.

Guru Grades

1,206 predictions by 42 market gurus since 2022: 41% right. Always saying stocks will go higher: 62%.

As of the Nov 18, 2026 close

  1. 1 Ed Yardeni A 71%
  2. Always bullish B 62%
  3. 2 Jim Paulsen B 61%
  4. 3 Sam Stovall B 58%
  5. 4 Adam Turnquist B 58%
  6. 5 Ryan Detrick B 56%

Full leaderboard ›