Comment Re:You can't blame it all on the qunats. (Score 1) 198
***Use Gaussian if you have an idea of what the parameter probably is but aren't exactly sure, rectangular if you really have no idea. A rectangular distribution says "I have no idea, the parameter could be anywhere within this particular range."
Here you have exactly why the quants got things so wrong. If you have an arbitrary random variable with a finite variance, then a law of large numbers will tell you that it converges to a Guassian under repetition. That's what most educated people know.
The problem is: the odds of an arbitrary distribution with no bounds having a finite variance is zero. In finance and economics, we know this, and we make up so many excuses to use Gaussians instead of more general LLN collection families. Gaussians are so tractable and easy to use, and so consistently used in theory that its second nature for us to use them. And since Gaussians are MLEs with relatively few restrictions, they tend to minimize measures of entropy. And since this is economics, everything ultimately has a bound... somewhere... unless its derivatives.
Not even most economists, financiers, or quants want to have to apply Levy-stable distributions all the time to variables, because working in complex spaces with integrals that never seem to converge when you want them to is a giant mathematical pain in the ass. But that's what real risk management is - your models need to be robust to whether that "variance" number your data indicates is real or bullshit or infinite.
Quants who knew better willfully ignored this, because "risk" (and "volatility" and "variance") makes "return," and they like big paychecks when things are going well. And when things aren't going well, well, who else could understand this stuff? Your average banker with an MBA might fire them, but doesn't have a chance at understanding, so he'll hire the average economist with a PhD who has a small chance, but recommends a mathematician who will say he doesn't understand the interpretation, and will refer you to a quant.