sponsored links
TEDGlobal 2013

Didier Sornette: How we can predict the next financial crisis

June 11, 2013

The 2007-2008 financial crisis, you might think, was an unpredictable one-time crash. But Didier Sornette and his Financial Crisis Observatory have plotted a set of early warning signs for unstable, growing systems, tracking the moment when any bubble is about to pop. (And he's seeing it happen again, right now.)

Didier Sornette - Risk economist
Didier Sornette studies whether it is possible to anticipate big changes or predict crises in complex systems. Full bio

sponsored links
Double-click the English subtitles below to play the video.
Once upon a time
00:12
we lived in an economy of financial growth and prosperity.
00:14
This was called the Great Moderation,
00:20
the misguided belief by most economists,
00:24
policymakers and central banks
00:27
that we have transformed into a new world
00:31
of never-ending growth and prosperity.
00:34
This was seen by robust and steady GDP growth,
00:38
by low and controlled inflation,
00:43
by low unemployment,
00:46
and controlled and low financial volatility.
00:48
But the Great Recession in 2007 and 2008,
00:52
the great crash, broke this illusion.
00:58
A few hundred billion dollars of losses in the financial sector
01:02
cascaded into five trillion dollars
01:07
of losses in world GDP
01:11
and almost $30 trillion losses
01:13
in the global stock market.
01:17
So the understanding of this Great Recession
01:21
was that this was completely surprising,
01:26
this came out of the blue,
01:31
this was like the wrath of the gods.
01:33
There was no responsibility.
01:35
So, as a reflection of this,
01:37
we started the Financial Crisis Observatory.
01:39
We had the goal to diagnose in real time
01:43
financial bubbles
01:47
and identify in advance their critical time.
01:49
What is the underpinning, scientifically, of this financial observatory?
01:55
We developed a theory called "dragon-kings."
01:58
Dragon-kings represent extreme events
02:03
which are of a class of their own.
02:06
They are special. They are outliers.
02:09
They are generated by specific mechanisms
02:12
that may make them predictable,
02:16
perhaps controllable.
02:19
Consider the financial price time series,
02:21
a given stock, your perfect stock,
02:25
or a global index.
02:27
You have these up-and-downs.
02:30
A very good measure of the risk of this financial market
02:32
is the peaks-to-valleys that represent
02:35
a worst case scenario
02:38
when you bought at the top and sold at the bottom.
02:40
You can look at the statistics, the frequency of the occurrence
02:44
of peak-to-valleys of different sizes,
02:48
which is represented in this graph.
02:50
Now, interestingly, 99 percent
02:52
of the peak-to-valleys of different amplitudes
02:56
can be represented by a universal power law
02:59
represented by this red line here.
03:03
More interestingly, there are outliers, there are exceptions
03:06
which are above this red line,
03:10
occur 100 times more frequently, at least,
03:12
than the extrapolation would predict them to occur
03:16
based on the calibration of the 99 percent remaining
03:20
peak-to-valleys.
03:24
They are due to trenchant dependancies
03:26
such that a loss is followed by a loss
03:31
which is followed by a loss which is followed by a loss.
03:34
These kinds of dependencies
03:37
are largely missed by standard risk management tools,
03:41
which ignore them and see lizards
03:46
when they should see dragon-kings.
03:49
The root mechanism of a dragon-king
03:53
is a slow maturation towards instability,
03:57
which is the bubble,
04:00
and the climax of the bubble is often the crash.
04:02
This is similar to the slow heating of water
04:04
in this test tube reaching the boiling point,
04:08
where the instability of the water occurs
04:11
and you have the phase transition to vapor.
04:13
And this process, which is absolutely non-linear --
04:17
cannot be predicted by standard techniques --
04:20
is the reflection of a collective emergent behavior
04:23
which is fundamentally endogenous.
04:27
So the cause of the crash, the cause of the crisis
04:29
has to be found in an inner instability of the system,
04:32
and any tiny perturbation will make this instability occur.
04:35
Now, some of you may have come to the mind
04:42
that is this not related to the black swan concept
04:45
you have heard about frequently?
04:48
Remember, black swan is this rare bird
04:51
that you see once and suddenly shattered your belief
04:53
that all swans should be white,
04:57
so it has captured the idea of unpredictability,
04:59
unknowability, that the extreme events
05:03
are fundamentally unknowable.
05:04
Nothing can be further
05:07
from the dragon-king concept I propose,
05:09
which is exactly the opposite, that most extreme events
05:11
are actually knowable and predictable.
05:15
So we can be empowered and take responsibility
05:18
and make predictions about them.
05:22
So let's have my dragon-king burn this black swan concept.
05:24
(Laughter)
05:27
There are many early warning signals
05:29
that are predicted by this theory.
05:32
Let me just focus on one of them:
05:34
the super-exponential growth with positive feedback.
05:36
What does it mean?
05:39
Imagine you have an investment
05:40
that returns the first year five percent,
05:42
the second year 10 percent, the third year 20 percent,
05:46
the next year 40 percent. Is that not marvelous?
05:49
This is a super-exponential growth.
05:51
A standard exponential growth corresponds
05:55
to a constant growth rate, let's say, of 10 percent
05:57
The point is that, many times during bubbles,
06:00
there are positive feedbacks which can be of many times,
06:04
such that previous growths enhance,
06:07
push forward, increase the next growth
06:11
through this kind of super-exponential growth,
06:15
which is very trenchant, not sustainable.
06:17
And the key idea is that the mathematical solution
06:20
of this class of models exhibit finite-time singularities,
06:23
which means that there is a critical time
06:27
where the system will break, will change regime.
06:30
It may be a crash. It may be just a plateau, something else.
06:34
And the key idea is that the critical time,
06:37
the information about the critical time is contained
06:39
in the early development of this super-exponential growth.
06:42
We have applied this theory early on, that was our first success,
06:47
to the diagnostic of the rupture of key elements
06:51
on the iron rocket.
06:55
Using acoustic emission, you know, this little noise
06:58
that you hear a structure emit, sing to you
07:00
when they are stressed, and reveal the damage going on,
07:03
there's a collective phenomenon of positive feedback,
07:06
the more damage gives the more damage,
07:09
so you can actually predict,
07:10
within, of course, a probability band,
07:13
when the rupture will occur.
07:15
So this is now so successful that it is used
07:17
in the initial phase of [unclear] the flight.
07:19
Perhaps more surprisingly, the same type of theory
07:24
applies to biology and medicine,
07:27
parturition, the act of giving birth, epileptic seizures.
07:29
From seven months of pregnancy, a mother
07:33
starts to feel episodic precursory contractions of the uterus
07:37
that are the sign of these maturations
07:42
toward the instability, giving birth to the baby,
07:46
the dragon-king.
07:50
So if you measure the precursor signal,
07:52
you can actually identify pre- and post-maturity problems
07:55
in advance.
08:01
Epileptic seizures also come in a large variety of size,
08:03
and when the brain goes to a super-critical state,
08:06
you have dragon-kings which have a degree of predictability
08:10
and this can help the patient to deal with this illness.
08:13
We have applied this theory to many systems,
08:19
landslides, glacier collapse,
08:21
even to the dynamics of prediction of success:
08:24
blockbusters, YouTube videos, movies, and so on.
08:27
But perhaps the most important application
08:32
is for finance, and this theory
08:35
illuminates, I believe, the deep reason
08:37
for the financial crisis that we have gone through.
08:41
This is rooted in 30 years of history of bubbles,
08:43
starting in 1980, with the global bubble
08:47
crashing in 1987,
08:51
followed by many other bubbles.
08:53
The biggest one was the "new economy" Internet bubble
08:56
in 2000, crashing in 2000,
08:58
the real estate bubbles in many countries,
09:00
financial derivative bubbles everywhere,
09:02
stock market bubbles also everywhere,
09:05
commodity and all bubbles, debt and credit bubbles --
09:07
bubbles, bubbles, bubbles.
09:11
We had a global bubble.
09:14
This is a measure of global overvaluation
09:17
of all markets, expressing what I call
09:21
an illusion of a perpetual money machine
09:25
that suddenly broke in 2007.
09:28
The problem is that we see the same process,
09:32
in particular through quantitative easing,
09:36
of a thinking of a perpetual money machine nowadays
09:39
to tackle the crisis since 2008 in the U.S., in Europe,
09:42
in Japan.
09:47
This has very important implications
09:49
to understand the failure of quantitative easing
09:51
as well as austerity measures
09:55
as long as we don't attack the core,
09:57
the structural cause of this perpetual money machine thinking.
10:00
Now, these are big claims.
10:05
Why would you believe me?
10:08
Well, perhaps because, in the last 15 years
10:11
we have come out of our ivory tower,
10:14
and started to publish ex ante --
10:17
and I stress the term ex ante, it means "in advance" —
10:19
before the crash confirmed
10:23
the existence of the bubble or the financial excesses.
10:25
These are a few of the major bubbles
10:28
that we have lived through in recent history.
10:31
Again, many interesting stories for each of them.
10:36
Let me tell you just one or two stories
10:39
that deal with massive bubbles.
10:41
We all know the Chinese miracle.
10:43
This is the expression of the stock market
10:46
of a massive bubble, a factor of three,
10:49
300 percent in just a few years.
10:52
In September 2007,
10:54
I was invited as a keynote speaker of a macro hedge fund
10:57
management conference,
11:01
and I showed to the conference a prediction
11:03
that by the end of 2007, this bubble
11:07
would change regime.
11:11
There might be a crash. Certainly not sustainable.
11:13
Now, how do you believe the very smart,
11:16
very motivated, very informed macro hedge fund managers
11:22
reacted to this prediction?
11:27
You know, they had made billions
11:29
just surfing this bubble until now.
11:31
They told me, "Didier,
11:34
yeah, the market might be overvalued,
11:35
but you forget something.
11:38
There is the Beijing Olympic Games coming
11:40
in August 2008, and it's very clear that
11:43
the Chinese government is controlling the economy
11:45
and doing what it takes
11:48
to also avoid any wave and control the stock market."
11:50
Three weeks after my presentation,
11:54
the markets lost 20 percent
11:56
and went through a phase of volatility,
11:59
upheaval, and a total market loss of
12:01
70 percent until the end of the year.
12:04
So how can we be so collectively wrong
12:06
by misreading or ignoring the science
12:09
of the fact that when an instability has developed,
12:13
and the system is ripe, any perturbation
12:16
makes it essentially impossible to control?
12:18
The Chinese market collapsed, but it rebounded.
12:22
In 2009, we also identified that this new bubble,
12:27
a smaller one, was unsustainable,
12:31
so we published again a prediction, in advance,
12:34
stating that by August 2009, the market will correct,
12:38
will not continue on this track.
12:42
Our critics, reading the prediction,
12:45
said, "No, it's not possible.
12:48
The Chinese government is there.
12:52
They have learned their lesson. They will control.
12:53
They want to benefit from the growth."
12:56
Perhaps these critics have not learned their lesson previously.
12:58
So the crisis did occur. The market corrected.
13:01
The same critics then said, "Ah, yes,
13:05
but you published your prediction.
13:07
You influenced the market.
13:09
It was not a prediction."
13:11
Maybe I am very powerful then.
13:14
Now, this is interesting.
13:17
It shows that it's essentially impossible until now
13:19
to develop a science of economics
13:22
because we are sentient beings who anticipate
13:24
and there is a problem of self-fulfilling prophesies.
13:28
So we invented a new way of doing science.
13:32
We created the Financial Bubble Experiment.
13:35
The idea is the following. We monitor the markets.
13:38
We identify excesses, bubbles.
13:41
We do our work. We write a report
13:45
in which we put our prediction of the critical time.
13:48
We don't release the report. It's kept secret.
13:53
But with modern encrypting techniques,
13:55
we have a hash, we publish a public key,
13:58
and six months later, we release the report,
14:02
and there is authentication.
14:06
And all this is done on an international archive
14:08
so that we cannot be accused of just releasing the successes.
14:13
Let me tease you with a very recent analysis.
14:17
17th of May, 2013, just two weeks ago,
14:21
we identified that the U.S. stock market
14:24
was on an unsustainable path
14:26
and we released this on our website on the 21st of May
14:28
that there will be a change of regime.
14:32
The next day, the market started to change regime, course.
14:34
This is not a crash.
14:39
This is just the third or fourth act
14:40
of a massive bubble in the making.
14:43
Scaling up the discussion at the size of the planet,
14:46
we see the same thing.
14:49
Wherever we look, it's observable:
14:50
in the biosphere, in the atmosphere, in the ocean,
14:53
showing these super-exponential trajectories
14:56
characterizing an unsustainable path
15:00
and announcing a phase transition.
15:02
This diagram on the right
15:05
shows a very beautiful compilation of studies
15:07
suggesting indeed that there is a nonlinear -- possibility
15:10
for a nonlinear transition just in the next few decades.
15:13
So there are bubbles everywhere.
15:18
From one side, this is exciting
15:21
for me, as a professor who chases bubbles
15:23
and slays dragons, as the media has sometimes called me.
15:26
But can we really slay the dragons?
15:31
Very recently, with collaborators,
15:34
we studied a dynamical system
15:36
where you see the dragon-king as these big loops
15:39
and we were able to apply tiny perturbations at the right times
15:42
that removed, when control is on, these dragons.
15:45
"Gouverner, c'est prévoir."
15:50
Governing is the art of planning and predicting.
15:53
But is it not the case that this is probably
15:59
one of the biggest gaps of mankind,
16:02
which has the responsibility to steer
16:06
our societies and our planet toward sustainability
16:09
in the face of growing challenges and crises?
16:12
But the dragon-king theory gives hope.
16:17
We learn that most systems have pockets of predictability.
16:21
It is possible to develop advance diagnostics of crises
16:25
so that we can be prepared, we can take measures,
16:30
we can take responsibility,
16:33
and so that never again will
16:36
extremes and crises like the Great Recession
16:39
or the European crisis take us by surprise.
16:42
Thank you.
16:47
(Applause)
16:48
Translator:Joseph Geni
Reviewer:Morton Bast

sponsored links

Didier Sornette - Risk economist
Didier Sornette studies whether it is possible to anticipate big changes or predict crises in complex systems.

Why you should listen

While financial crashes, recessions, earthquakes and other extreme events appear chaotic, Didier Sornette's research is focused on finding out whether they are, in fact, predictable. They may happen often as a surprise, he suggests, but they don't come out of the blue: the most extreme risks (and gains) are what he calls "dragon kings" that almost always result from a visible drift toward a critical instability. In his hypothesis, this instability has measurable technical and/or socio-economical precursors. As he says: "Crises are not external shocks."

An expert on complex systems, Sornette is the chair of entrepreneurial risk at the Swiss Federal Institute of Technology, and director of the Financial Crisis Observatory, a project to test the hypothesis that markets can be predictable, especially during bubbles. He's the author of Why Stock Markets Crash: Critical Events in Complex Financial Systems.

sponsored links

If you need translations, you can install "Google Translate" extension into your Chrome Browser.
Furthermore, you can change playback rate by installing "Video Speed Controller" extension.

Data provided by TED.

This website is owned and operated by Tokyo English Network.
The developer's blog is here.