Call for papers: What to do when you cannot forecast
Uncategorized April 2nd. 2008, 9:59amNassim Nicholas Taleb is co-editing a special issue of the International Journal of Forecasting with focus on “decision making and planning under low levels of predictability”. The official invite can be found here and while I will not be submitting a paper, I will surely obtain a copy of the issue when it is published.
April 2nd, 2008 at 11:22 pm
That’s a really interesting topic!
Have you read “Handbook of Heavy Tailed Distributions in Finance” or “Fat-Tailed and Skewed Asset Return Distributions : Implications for Risk Management, Portfolio Selection, and Option Pricing”?
http://www.amazon.com/Handbook-Tailed-Distributions-Finance-Handbooks/dp/0444508961
http://www.amazon.com/Fat-Tailed-Skewed-Asset-Return-Distributions/dp/0471718866
April 2nd, 2008 at 11:41 pm
Seems to me you have just picked random books out of thin air! Have you read them yourself? Why read them?
April 3rd, 2008 at 6:56 am
Random? lol, both books talk about the same topic, the properties of the statistical distribution on financial data, particularly fat tailed distributions (which are a common distribution in financial data). Most models in finance assume a gausian distribution, which in many cases is not true. For example gaussian distributions cannot explain the big movements in the markets (like news releases), while heavy tailed distributions can. In other way, under the gaussian distribution assumption it’s imposible to even consider the existance of a Black Swan as Taleb would say, meanwhile fat tailed distributions can. As you are working with financial data, understanding this is quite important, and can help you develop better ideas. Forecasting in finance is a quite complicated subject, and it can be a easier if you know your data. I’ve read almost everything on the “Handbook of Heavy Tailed Distributions in Finance”, and when i have some time i’ll order the other book (a friend recommended the other one).
April 3rd, 2008 at 9:11 am
Indeed, you explain very well an important issue over which models like Black-scholes and the like often get slated by people like Taleb. But a gaussain distribution is not too bad an assumption for certain time series particularly high frequency tick data. Under these resolutions you find that the returns distribution takes on a different structure, sometimes positively skewed and with almost non-existent fat tails, and in the event you get extreme outliers you usually have plausible explanations for their existence. But again knowing your data doesn’t guarantee it will continue to behave the same way in the future, and assuming that outliers will not show up for data that has never had outliers in the past (or vice-versa) is probably taking things way too general. The meaning of risk/return take on a different meaning at the tick level and as always a model is only as good as the assumptions supporting it, doesn’t necesarily have to take into account fat tails particularly if it is not a function of such a variable.
The handbook looks good and I shall obtain a copy to have a look. Thanks for the links!