Monday, August 23, 2010

Nassim's Selective Returns


Nassim 'the Dream' Taleb's Universa is supposedly advising China Investment Corporation on buying some insurance on all their US exposure. Here's a pro tip for the CIC: if you want to hedge your exposure to US Treasuries, don't buy US Treasuries plus insurance on US Treasuries & the US Dollar. That's churning. Instead, invest more in a different country, like Canada or Iceland. You don't get the insurance for free, you pay the expected loss plus fees--otherwise no one would sell insurance. It's almost as if the Chinese exist to give their money away to Americans.

Nassim is fond of highlighting that public prognosticators, especially financial analysts, do not keep an accurate, testable track record, and so are deceitful, uncalibrated, and most importantly wrong. So it's fun to see his record bandied about in a misleading, tendentious way. Today's Wall Street Journal states the Universa Fund he is affiliated with (but run by Mark Spitznagel), had "client portfolios [in 2009] down an average of about 4%, one person close to the matter said". "This year, Universa clients on average have lost about 2% of their notional account value", but in 2008 "extreme market losses helped many Universa clients profit more than 100%."

There are a lot of qualifiers in those off-the-record statements. As implied vols went from 40 to 20 in 2009, I'd be surprised if his average client only lost 4% that year. Further, as vols peaked around early November 2008 around 80, and then fell dramatically to 40 by December 31 2008, I'd be surprised if that 100% number is representative for 2008, because that was the number bandied about by the credulous press during the top tic in November when vols were 80. Of course, without audited returns, who knows.

Janet Tavakoli busted a GQ article that stated Taleb made $20B for his clients, and Taleb angrily responded that he never said that number, and Tavakoli was on a smear campaign. It's very manipulative when you let an interviewer quote unidentified sources for returns in your piece, and then act shocked, shocked, when people point out they are incorrect: 'I never said that'. Technically true, but knowingly misleading, pathetic for a man who makes a big deal out of the fact that financial professionals are miscalibrated and hide their track records. I know three people he threatened to sue for defamation; he's not forthright when it comes to his own record.

NNT's returns from 1999-2003 were probably only 4% annualized (I don't have audited returns, so I'm going on hear say), and as it closed a year later, probably lower than that for its lifetime, pretty lame, but his 60% return in 2000 is all he needed for marketing. It's called the 'representativeness heuristic', a conspicuous focal point one then irrationally generalizes. Or it could be the 'anchoring heuristic', a number that once set, biases the final estimate towards this number. In any case, he's selling lottery tickets, an unfalsifiable strategy that generates really large returns in big crashes like 2008. Great, but as volatility, especially at the tails, is priced if anything too high, I don't think this strategy has a positive expected value to anyone but the fund owners (and their advisers!).

5 comments:

Dave said...

Eric,

Do you think volatility is priced too high right now?

Anonymous said...

Every time you use the word tendentious - this must be the 4th time at least - I have to look up what it means, but I never remember

Anonymous said...

why is it that in spite of your pathological vendeta on taleb he is still rich and you are still a lozer?

Eric Falkenstein said...

Michael Moore, Howard Zinn, Uri Geller, Cornell West, Thom Friedman, etc., all are or were very successful. I also think they are fundamentally wrong on what they think they know. As finance and uncertainty is my bailiwick, however, I don't criticize them much. You should be thankful, because you learn more from your critics than your fans.

Dave said...

Uri Geller? The spoon-bender? One of those names is not like the other.