Cafe Hayek - Latest Comments in Black Swans

Re: Black Swans

Mesa EconoGuy — Fri, 04 May 2007 17:21:13 -0000

Having built several (pseudo) predictive statistical models myself, it probably means that these global warmers have misidentified several inputs and/or their relative weightings, interactions, and other anomalies, and have most likely failed to account for numerous other variables and their relative climatological importance.

In short, they’re looking at a very narrow set of conditions/interactions, and fudging the input numbers (especially for CO2). Don’t believe most of what they say.

I haven’t read Taleb’s books, yet. They’re next up.

Re: Black Swans

Speedmaster — Tue, 01 May 2007 08:59:03 -0000

More thoughts on this. Mr. Taleb points out the problems associated with building models. What does this mean for all we read about global warming predictions that warn us of doom 100 years out?

Re: Black Swans

Russ Roberts — Tue, 01 May 2007 08:47:34 -0000

Speedmaster,

Read the books. They're very accessible. Try Fooled by Randomness first. If you like it, read The Black Swan. They're both listed in the right-hand margin.

Re: Black Swans

Speedmaster — Tue, 01 May 2007 07:42:53 -0000

Fascinating discussion! I loved the concrete examples that the two of you discussed. But I must admit that much of the discussion was over my head. Mr. Taleb is clearly brilliant.

Re: Black Swans

Ray G — Mon, 30 Apr 2007 17:04:40 -0000

I've only skimmed his newest book, but I like it so far. He uses Umberto Eco's large private library as an example from which he pulls the little gem that an unread book is far more valuable than the book one has already read.

Re: Black Swans

Pasha — Mon, 30 Apr 2007 11:32:38 -0000

Good interview - he's much more palatable in person than in his books, which came off a bit heavy-handed for me.

Here's my question: why do so few talk about lognormal distributions with regard to issues like income or wealth? I believe that he (and/or mandelbrot) also implicate the power law distribution, which is similar except at the low end.

I'm not one for mathematical reductionism, but I'm not ready to throw out the practice of statistics either, just because we humans fall prone to cognitive errors.

Any of these log-based models would easily encompass high-tailed incomes, while recognizing the great preponderance of middle and lower incomes. Lognormal is commonly used in finance, for example, to represent asset prices - you can't go below zero because of bankruptcy. People are the same.