1: \begin{abstract}

2: By combining (i) the economic theory of rational expectation bubbles,

3: (ii) behavioral finance on imitation and herding of investors and

4: traders and (iii) the mathematical and statistical physics of

5: bifurcations and phase transitions, the log-periodic power law (LPPL)

6: model has been developed as a flexible tool to detect bubbles. The LPPL

7: model considers the faster-than-exponential (power law with finite-time

8: singularity) increase in asset prices decorated by accelerating

9: oscillations as the main diagnostic of bubbles. It embodies a positive

10: feedback loop of higher return anticipations competing with negative

11: feedback spirals of crash expectations. 

12: The power of the LPPL model is illustrated by two recent real-life

13: predictions performed recently by our group: the peak of the Oil

14: price bubble in early July 2008 and the burst of a bubble on the Shanghai

15: stock market in early August 2009. We then present the concept

16: of ``negative bubbles'', which are the mirror images of positive bubbles.

17: We argue that similar positive feedbacks are at work to fuel

18: these accelerated downward price spirals. We adapt

19: the LPPL model to these negative bubbles and implement 

20: a pattern recognition method to predict the end times of 

21: the negative bubbles, which are characterized by rebounds

22: (the mirror images of crashes associated with the standard

23: positive bubbles). The out-of-sample tests quantified by

24: error diagrams demonstrate the high significance of the

25: prediction performance.

26: \end{abstract}

27: