## July 21, 2009

### SSE Composite Index Bubble?

#### Posted by John Baez

Physicists like to complain that economists don’t make enough testable predictions. There’s a geophysicist at UCLA named Didier Sornette who has tried to remedy this situation. Here is his latest paper on the arXiv:

I was pointed to this paper by Robert Schlesinger. The paper claims that an equity index called the Shanghai Stock Exchange (SSE) Composite Index is exhibiting faster-than-exponential growth, that it’s in a “bubble” — and that this bubble will burst between July 17th and 27th with 60% probability. It hasn’t burst yet:

You can watch how it does on Yahoo. It’s interesting to see how different the chart looks on different time scales.

For more, try:

Sornette has a model of bubbles, and he’s been using it to make predictions. For example:

predicted the turning point of the real estate bubble in mid-2006. (Does that count as a success?)

For a sketchy but readable explanation of Sornette’s model, leading up to his “log-periodic power law”, try this:

This has references to more technical stuff involving some ideas from condensed matter physics.

Needless to say, don’t make any financial decisions based on what you read at the $n$-Category Café.

Posted at July 21, 2009 11:15 AM UTC

TrackBack URL for this Entry:   http://golem.ph.utexas.edu/cgi-bin/MT-3.0/dxy-tb.fcgi/2019

### Re: SSE Composite Index Bubble?

I’ll make a snarky comment: a more common complaint is that economists and physicists-turned-economists don’t make enough potentially-falsifying predictions. (Ie, they tend not to make predictions that, if they disagree with the observations, can’t be explained away by invoking reasons why the model, whilst valid in general, wasn’t appropriate this time around. Lots of economists and physicists seem to make lots of “testable predictions” that aren’t seen as “binding”.) The paper is the distillation of what I’m sure is very good theoretical work but doesn’t really address the issue of what conclusions about the practical usability of the model should be drawn if SSE doesn’t crash within that period.

(Of course that’s a little unfair in that in my field of machine learning – where there’s no intent to make “physically true” models, only effective predictive models – there’s the opposite problem: most people fuzz the line about whether a given performance level on given test data is really suggestive of effective performance on eventual application data which always contains many more strange and confounding factors than encountered in the test set.)

Posted by: bane on July 21, 2009 12:42 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

It becomes more amusing if you know what “sornette” means in french…

Posted by: NC on July 23, 2009 5:38 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Hi John,

Set the following graph for 5 years:

You will see that the bubble bursted on october 2007, fell ~70% to a minimum on october last year and since then it’s been recovering.

So, it is not a bubble, but a quick recovery curve.

Posted by: Daniel de França MTd2 on July 21, 2009 1:02 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

I am inclined to agree with Daniel, but I have not yet read any of Sornette’s work.

One of the main complaints of traditional quantitative finance models is that they fail to predict the emergence of a herd mentality which then causes asset prices to overshoot (either to the upside or downside), and so perhaps it is useful that Sornette is trying to address this problem.

I would like to tell any non-experts that there is no doubt whatsoever that the greatest quant ever is James Simons (of Chern-Simons fame). However, AFAIK, Jim never talks about finance (which I would argue is one sign of his genius).

Posted by: Charlie Stromeyer on July 22, 2009 2:13 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

So, we were correct and Sornette was wrong. Now, Morgan Stanley’s chief Asian economist Andy Xie is saying that the Chinese stock market bubble won’t start to pop until this October.

Posted by: Charlie Stromeyer on August 1, 2009 8:08 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Actually, this is why it would be interesting/useful if the paper had included some indication of what it would mean for the model if the prediction did not match reality. (And kind of why I quibbled a bit with John’s original point that “economists don’t make testable predictions” whereas they seem to me to make testable predictions all the time, it’s just never made clear what the interpretation of a failed prediction is: does it suggest the model is wrong, is there a strong possiblity of confounding events, is the model inherently non-predictive, is it descriptive but with very little practical predictive power, etc?) Unfortunately it’s easy to give the impression that one thinks economists are all completely wrong, whereas my issue is that economists are very non-transparent (and fond of invoking “special circumstances”) in their evaluation/defense of theories.

Posted by: bane on August 3, 2009 10:18 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Bane wrote:

And kind of why I quibbled a bit with John’s original point that “economists don’t make testable predictions” whereas they seem to me to make testable predictions all the time…

Not that it matters much, but just to be clear: I never said “economists don’t make testable predictions”. I said “Physicists like to complain that economists don’t make enough testable predictions.” And that, I know, is true.

(Since I don’t consider myself a physicist, this wasn’t a sneaky way of saying I like to make this complaint.)

Posted by: John Baez on August 3, 2009 10:31 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

This is actually a perfect example of what I find fascinating about economics predictions: the way that the prediction itself affects the system in question.

So let’s say a whole bunch of people read about this prediction. And let’s say that a bunch of them decide that, hey, 60% is pretty good, and this guy is as good a source as any for tips. So they decide to sell at the first sign of weakness in this period.

But then if enough people have this idea, then a minor quiver in the market becomes the signal for all of them to put this prediction into play. Everybody decides to sell at once, and all of a sudden the market actually does fall precipitously, when it wouldn’t have if this prediction had never been made.

The question, then, is “did the authors factor in the effects of their very own prediction into the prediction itself?”

I think that if we’re going to have testable predictions and actually test them in the financial markets, we’re going to need some sort of escrow service. An economist makes a prediction about something that’s going to happen (hopefully soon) and seals it with a definite end-date after the period about which he’s making the prediction. The escrow trustee opens the prediction on the end-date and publicizes it. The economist is not allowed to make predictions and then only publicize the successes of his model, and the economist’s predictions are insulated from the system he studies.

Of course, the trustee must somehow be trusted not to peek at the predictions under her care. And, knowing traders, someone would probably find a way to make some sort of derivative out of this system and trade on it…

Posted by: John Armstrong on July 21, 2009 1:19 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Of course it does depend why you want to validate the prediction. If you want to know the model was “physically true” (although physical probably is more mental here), then it’s an issue. If all you want to know is that it’s a robust predictor (say, in order to use it), there’s no a priori problem with having people know providing the market view of which people are viewed as “the smart predictions guys” doesn’t undergo wild swings.

Posted by: bane on July 21, 2009 2:11 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Last year a lot of “experts” suddenly appeared in the media, including public radio, declaring that the housing bubble was about to burst. It almost seemed to be a coordinated manipulation. Sure enough, the slide began immediately and precipitously.

Posted by: Richard on July 22, 2009 3:15 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Even worse, suppose that the economist proves her theory makes good predictions and publishes it. Then everyone start behaving in a way that takes the theory into account. This may change the dynamics, impairing the predictive ability of the theory. It cannot destroy it completely, since once the theory becomes less predictive it becomes less used, until this process reaches an equilibrium. After that the cycle begins again with a new theory. And so on, until it reaches a fixed point :-)

Posted by: Squark on July 22, 2009 8:00 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

.. or cycles infinitely.. or goes chaotic !!!

Posted by: Tim Porter on July 22, 2009 9:47 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Hi John and Squark,

You just described one of the things I love about finance. The “laws of nature” are ever changing and, unlike the physical universe, models can and do have a significant feedback effect on the financial universe they are intended to describe.

When I worked in computational electromagnetics, we would perform highly accurate simulations such as radar striking aircraft. We understood the physics so well that with an accurate CAD model of the aircraft, we could simulate radar scattering to amazing accuracy. One of my favorite stories is when my friend was comparing his simulation results to measurements. He determined that in order to get the simulation to match the measurement, he had to rotate the model aircraft .5 degrees relative to what the Air Force had specified. He called up the range to have them check on the experimental set up. Sure enough, they had measured the position of the aircraft incorrectly.

Finance is a different story and that is what is exciting about it.

The obvious difficulty with making falsifiable statements in finance and economics is, as John von Neumann enunciated in Theory of Games and Economic Behavior, that there is no way to “repeat an experiment”. The initial conditions for any observation can never be duplicated in finance and economics. One thing we can do and what we do do is make statistical statements, such as

I am 99% certain that in the next 10 days, we will not experience a loss more the x%.

After 10 days, we (and regulators) do look to see what we said 10 days ago and “count” how many times we violated that 99% threshold. If we said we were 99% certain, but we are wrong 75% of the time, then something is obviously wrong. The FED and the OCC will not tolerate that, not to mention internal risk management.

So over the past 10-15 years, as “quants” invaded Wall Street, these risk models became more and more sophisticated. We became able to adjust behavior so the we just barely stayed within that 99% confidence interval. Of course, we’re greedy and the closer we get to the threshold without violating it, the more money we can make (no risk, no reward).

One thing the statistical statement above does not tell you about is what happens if one of those 1% “tail events” happen. How much could I lose then? All risk models did not even attempt to answer this question (but you can be sure they will going forward).

So if you were greedy (and who on Wall Street isn’t?) and if you were smart (and there are a LOT of smart people on Wall Street), then you’d try to push the limits as much as possible. If ONLY you could create some security where ALL of the risk was in the tails, i.e. under most circumstances, it doesn’t lose anything, but once in a while, say after 2-3 years (after you’ve received your bonus and moved to another firm), the security loses everything, that would be PERFECT. The risk would remain completely invisible to all risk management systems (“We fooled you Mister regulator!”) and you’d be rich.

That is what a CDO is.

Anyway, reading this paper on an SSE bubble is low on my priority list, but I give it very little credence, i.e. my “BS” detector is going off. I’m guessing they built a model that is telling them,

We are 60% confident that the SSE Composite will lose more than 20% of its value over the next 10 days.

The very same model would probably also say

We are 10% confident that the SSE Composite will gain more than 20% over the next 10 days.

Basically, they are saying the direction of the SSE Composite is currently highly unpredictable. Well, gee, thanks!

China is a powerful country full of hard working, extremely bright, and entrepreneurial people that take higher education seriously. There is no doubt in my mind that in 20 years, they will be the dominant economic (and technological) world power. So am I surprised, that they are seeing greater than exponential growth? No, exponential growth is “normal”. Since China (and India) are both developing resources at amazing rates, it should not be surprising they are growing faster than normal. Would I be surprised if the SSE Composite crashed? No. There will be ups and downs along the way for sure, but the long term direction for China is definitely up. The Hype Cycle comes to mind. Are we coming out of the “Trough of Disillusionment”?

Here is the difficult question:

Assume you thought China was experiencing a bubble, would you invest in China? Bubbles can go on for years and years before popping. Should you “miss out” on years and years of large returns?

Posted by: Eric Forgy on July 22, 2009 7:36 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

PS: When I say “radar striking aircraft”, I mean the radar “signal”. I don’t mean the radar equipment is hitting the aircraft :)

Posted by: Eric Forgy on July 22, 2009 7:45 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

I mostly agree with what Eric says, but I just want to clarify for non-experts that a CDO is a form of credit derivative and in 2007 about 47% of CDOs were backed by structured products.

I fully agree with the quant Paul Wilmott that there is currently no mathematically reliable way to correctly price a portfolio of structured credits. This is one reason why some have raised doubts about the potential effectiveness of the Treasury Department’s PPIP program.

Also, John Maynard Keynes famously said that sometimes a market can remain irrational longer than one can remain solvent, so Eric is correct about the nature of bubbles.

Posted by: Charlie Stromeyer on July 22, 2009 8:41 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

This is going off-topic, but it’s interesting that the economics terminology still presupposes a worldview: bubbles are “irrational”, implying both that participation in a bubble is more “irrational” than other behaviours (whereas as Eric points out that from an individual trading entity’s viewpoint participation may be a “quite sensible” behvaiour [I point-blank refuse to use the term “rational”]) and indeed that there are other markets behaviours that actually occur in the real-world (rather than in economic textbooks) that could be called significantly more “rational” than other forms of “irrational” behaviour. (As a mathematics analogy, I remember a lecturer claiming that there was a significant change in world view at the time that general mathematical usage changed from “non-linear” to “nonlinear” phenomena.)

(I’d also refer to Wall Street as “trading Chinese stocks” rather than as “investing in China”, but then I’m an old-fashioned lefty :-) )

Posted by: bane on July 23, 2009 12:49 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

bane, there is truth to what you say, and I strongly recommend reading the book “Predictably Irrational (Revised and Expanded Edition)” by Dan Ariely.

This book is based on actual scientific experiments and analysis, and sheds light on understanding financial bubbles among other human behaviors that are irrational in ways that are not without some kind of precedent.

Posted by: Charlie Stromeyer on July 23, 2009 1:36 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

I might have a look at that book (although I’ve read some of the popular writings of Shiller et al on the subject). But the point I was raising (hopefully in a “this is an interesting phenomenon” rather than confrontational way) was the way economics sticks with the terms rational/bounded-rationality/irrational rather than moving to (admittedly kludgy) terms like “classical, perfect-information behaviour”/”normal human”, rather than saying such things aren’t considered by economists. (Purely by chance, I happen to know a couple of academic economists working in such areas.)

It just struck me as an interesting way that a choice of language (that hasn’t been updated from Bachelier’d time) discourages certain viewpoints.

Posted by: bane on July 23, 2009 2:37 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

For the record, the economic profession has been talking about “rational bubbles” for a long time - at least since Olivier Blanchard’s “Speculative Bubbles, Crashes, and Rational Expectations” Economics Letters, vol. 3 (1979), pp. 387-89 (ScienceDirect link: http://dx.doi.org/10.1016/0165-1765(79)90017-X).

If anything, the profession could have been criticized for talking almost ONLY about RATIONAL bubbles - until the recent explosion of behavioral economics research.

Btw, Didier Sornette has a paper (with Thomas Lux) whose title is in fact “On Rational Bubbles and Fat Tails” (Journal of Money, Credit and Banking, vol. 34, no. 3, Part 1 (August 2002), pp. 589-610 (available on JStor at http://www.jstor.org/stable/pdfplus/3270733.pdf)

Posted by: Valter Sorana on July 27, 2009 6:22 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

One quick comment: the point wasn’t about the substance of economic research but rather the terminology. Referring to the behaviour that is currently accepted in modeling as “rational” and that is not as “irrational”, it creates an viewpoint that what’s not modeled is “wrong behaviour”. (It’s a analogous to the use of the term “structured programming” in computer science to mean “does not use goto’s”: some felt that because the term for code which, whatever it’s understandability in other respects, did use goto’s was “unstructured code”, and no-one wants to be accused of writing that…)

Posted by: bane on July 28, 2009 9:59 AM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Oh, sorry - I thought you were saying that economists call “bubble” behaviour irrational.

Your point instead (if I now understand it correctly) is that economics call “irrational” all behavior in a given model that would not be optimal for the agent if the model were otherwise a perfect representation of reality - even though that behavior may well be optimal in the real world because of something that is not correctly considered in the model (be it the environment or the “rationality” of other agents).

Fair point, then.

Posted by: Valter Sorana on July 28, 2009 4:36 PM | Permalink | Reply to this

### Personal note

Quand je n’avais que seize ans et demi, I lived with a French family for the summer -1952. It was meant to be an exchange of children between families, but the daughter Nicole visited my family only the following summer and only for a week or two. Nicole later married Christian Sornette. His first born son is indeed this same Didier Sornette.

Posted by: jim stasheff on July 21, 2009 4:04 PM | Permalink | Reply to this

### Re: SSE Composite Index Bubble?

Finally! A topic I might actually be able to say something about since I AM one of those physicists turned to finance, but alas…

Real life calls (for now!)…

Posted by: Eric Forgy on July 22, 2009 3:15 PM | Permalink | Reply to this

Post a New Comment