Since no one seemed to read the book I posted yesterday, I wonder if anybody has ever looked into the application of chaos and complexity in the area of financial markets. Do you know anybody doing this? General concepts, idea, etc.
I have a list of books that are great I’ll check out when I get home. The first ones that jump to mind are: Fooled By Randomness, Taleb and A Demon of Our Own Design: Markets, Hedge Funds and the Perils of Financial Innovation, Bookstaber Two of my favorites of all time.
Vampire Diaries?
Well…I’ve read some of Peters’ books and he’s interesting, but I hadn’t read the book you were talking about (or if I did it was only 20 or so pages in a book store). I read Mandelbrot’s Misbehavior of Markets and he has some others that are supposed to be good. I probably should re-read it now that I think about it, would probably get more out of it now. I think the most important insight of some of this work is that variances and correlations are time-varying and that returns can exhibit fat tails, though not exactly unique to the chaos/complexity theorists. There are techniques that handle some of this stuff. Garch is pretty quick, but in my experience regime-switching models take a bit longer to estimate. You might try reading a book about econophysics. I see that word thrown around, but I haven’t read any books to see how it is different. My gut tells me it applies more to finance than economics proper. Some people incorporate jump diffusion into option modeling, but I’m not really sure how much people do that or how well it works. There’s also work in high frequency trading that seems to do a better job at tackling the fractal nature of markets, or really the fact that trades don’t happen at the same time. http://en.wikipedia.org/wiki/Autoregressive_conditional_duration
The real problem with this stuff (and I’ll admit that maybe some people have made progress since I last looked at it) is that it seems to be correct as a description of reality, but it’s not at all clear what one can do about it. If everything depends sensitively on initial conditions but you haven’t the foggiest idea what those initial conditions actually are, or even much idea about what the underlying causal rules are, then all you really can get out of it is to “expect the unexpected.” I actually do think that chaos theory provides an explanation for why technical analysis might work… technical chart patterns may well be projections of strange attractors in phase space. But again, how to take advantage of that in any money-making way is still a big mystery. About the only way seems to be to try “robust” statistical methods, or perhaps fuzzy logic approaches. All the math that goes into that can be fun for the math geeks among us, but ultimately what it boils down to the following: underperforming everyone else during “normal” times because you aren’t taking as much risk when risk appears to pay well, and getting whacked less than others (possibly) during regime shifts and black swan events. You can make the argument that in the very long term, this ultimately preserves capital better and generates a higher geometric return, and that might (MIGHT) be true (depending on what the black swan events actually are, and whether you’ve included them in your risk modeling, which is almost by definition impossible). But what we’ve seen is when there is a crisis like that, the political economy changes so much and the wild risk takers get bailed out and pay themselves bonuses anyway, so what is the point of all that prudence, really?
I understand the practical short comings of chaos theory and tightly coupled systems in a world given to fat tails. The problem isn’t use of normal statistical methods per say, it’s the over reliance on them, the common ignorance to the magnitude of their short coming and the high leverage and tightly wound financial system that results. In short, because people are chronically ignorant to the reality of a chaotic system (admitting it in academic arguments but failing to reflect it in practice), we build these intricate systems as if we did live in a normally distributed world. You can use stat methods all day, but just build your foundation strong enough, if by no other means than reduced systemic leverage. It’s astounding the number of idiots I meet in NYC that work in ALM and don’t understand the implications of a non-normal system. A simple fix would be to widely acknowledge the disparity, then de-lever and increase qualitative analysis with the proper oversight.
Bchad has a good point about robust analysis. Skewness and kurtosis tend to get estimated with a log of uncertainty.
The first and second papers listed at the link below seem pertinent to this discussion. I’d be interested in thoughts on them. https://www.joim.com/2010_winners_summaries.pdf
Agreed, chaos surrounds models and they should be extremely robusty.
Black Swan Wrote: ------------------------------------------------------- > I understand the practical short comings of chaos > theory and tightly coupled systems in a world > given to fat tails. The problem isn’t use of > normal statistical methods per say, it’s the over > reliance on them, the common ignorance to the > magnitude of their short coming and the high > leverage and tightly wound financial system that > results. In short, because people are chronically > ignorant to the reality of a chaotic system > (admitting it in academic arguments but failing to > reflect it in practice), we build these intricate > systems as if we did live in a normally > distributed world. You can use stat methods all > day, but just build your foundation strong enough, > if by no other means than reduced systemic > leverage. It’s astounding the number of idiots I > meet in NYC that work in ALM and don’t understand > the implications of a non-normal system. > > A simple fix would be to widely acknowledge the > disparity, then de-lever and increase qualitative > analysis with the proper oversight. By the way, this is a neat conversation. Reminds me of AF in its heyday. I wonder what JDV would have to contribute to this (actually, I think he did say something along these lines at some point). I think the problem with your solution (widely acknowledging the disparity, then de-levering) is that in a competitive economy, the temptation is to keep upping the risk levels in good times. Since the fat tails are relatively rare (more frequent than they “ought” to be, but still relatively infrequent), people will be making money and claiming that they’re brilliant, when in fact they are taking unwarranted risks and just being lucky. By the time a fat tailed event does arise, everyone’s abandoned the prudent people because they’re only making 10% and the riskier competition has been making 20% for the past 4 years in a row. So it sounds like a sensible response, but I don’t see how you can stop this without regulatory rules, which have their own problems. I do agree that tight coupling with chaotic systems is a problem (isn’t marriage a pain?) and that it would be good to have better tools to solve it, but I have yet to see any (though I’m open to learning).
This is good stuff.
bchadwick Wrote: ------------------------------------------------------- > > By the way, this is a neat conversation. Reminds > me of AF in its heyday. I wonder what JDV would > have to contribute to this (actually, I think he > did say something along these lines at some > point). ^Agreed. We had quite the dream team back in the day. Kinda weird that a group of anonymous internet finance professionals had such an impact on my development. I opened my first L1 CFAi book in October 2007 for the December exam. Before that, I’d only ever taken one stat class that I literally failed “F”, same with my accounting and calc 1 classes. I’d graduated Dec 06 with a 2.5 GPA in business admin from an unheard of school (partied to hard, thought the haze could last forever). Only job experience was landscaping. Been a long road to recovery. Definitely owe a lot to AF and the CFA. > I think the problem with your solution (widely > acknowledging the disparity, then de-levering) is > that in a competitive economy, the temptation is > to keep upping the risk levels in good times. > Since the fat tails are relatively rare (more > frequent than they “ought” to be, but still > relatively infrequent), people will be making > money and claiming that they’re brilliant, when in > fact they are taking unwarranted risks and just > being lucky. By the time a fat tailed event does > arise, everyone’s abandoned the prudent people > because they’re only making 10% and the riskier > competition has been making 20% for the past 4 > years in a row. > > So it sounds like a sensible response, but I don’t > see how you can stop this without regulatory > rules, which have their own problems. > > I do agree that tight coupling with chaotic > systems is a problem (isn’t marriage a pain?) and > that it would be good to have better tools to > solve it, but I have yet to see any (though I’m > open to learning). ^Also agreed. I realize the proper changes will almost definitely never happen for all those reasons. Still frustrating none the less. Even worse when you hear people utside this forum referring to “black swans” without any comprehension of the true principals behind these theories. I’d almost prefer a world of willfull ignorance to the fact that there are so manny dummies out there with top tier mba’s and no understanding of their craft. Also, kinda weird that this convo kinda took place across three threads simultaneously.
So I flipped through Mandelbrot’s book last night to sort of jog my memory on this chaos/fractal stuff. He seems to have modeled some financial prices as a levy stable distribution. Turns out that these distributions have infinite variance. They seem comparable to a skew-t distribution, except that skew-t has finite variance and you can add up stable distributions. Has anyone worked with these?