Thursday, September 25, 2008

Got Black Swan?

I picked up Nicholas Taleb's book the Black Swan a few weeks ago, and was intrigued... his reservations about things somewhat mirrored my own, but as a amateur, I just figured that someone else knew better. Here are some of the ideas that I gleaned from his book. Specifically what bothered me when I was learning Stochastic Calculus as applied to finance... was the basic argument that the underlying device operated under something that could be modeled by geometric brownian motion or a fixed step random walk. What bugged me about this was the idea that there was a fixed amount of badness that could happen in a given time frame. I think the models are effective and good most of the time, but it seems that you expose yourself to events that you thought were seriously not possible... for instance under a standard normal distribution, the common attribution to standard deviation lets you take a poke at the probability of events happening that are away from the mean. So 1 standard deviation implies that ~68% of events should occur within this range, and for 2 standard deviations about ~95% of events occur within this range etc... So usually you have some level of confidence that you do not have a say... 10 sigma event happen... say like the stock market crash of 1987 where a 20-22 sigma event definitely punched us in the face.... One of the reasons this is important is that, when they price these options and swaps... which are 'insurance policies if you will', the academic argument is that you take all future possible payouts, and then discount them back to what they in aggregate will be worth... so if you have a huge range of possibility, you should be discounting huge swings in possible payouts... if your guess at the range or 'volatility' is way off... the premium you will charge will be off as well. This is just what I gleaned from a few classes and I think the implication of using the wrong assessment for risk is correct.

I would recommend checking out the book The Black Swan by Nicholas Taleb. I don't think he has it perfect either, but I think a contrarian view is healthy in these times of madness.

Btw I started reading the book the BELL CURVE... and after checking out the appendix decided to stop reading the book. The R2 values of a lot of those 'graphs' were below 0.1, which to me... should have been posted on each graph... vs being hidden in the appendix... also R2 value implies how much of the actual data is 'explained' by the regression model... I don't mind low R2... but the way the arguments are presented in my opinion at least, you would hope they have R2 of 0.8 or higher...

No comments: