Not So Perfect Warren Buffett

I love this Colby Cosh analysis of Warren Buffett in the National Post: (read)

This is champagne-flavoured optimism at its finest. But more than in most years, one is aware that Buffett’s letter is an exercise, not only in teaching and accounting, but in selling the Warren Buffett brand. After a 12-month period in which shares in his company lost 19% of their value, even the Oracle can start to sound a little like that old hypocrite Polonius.

Even before the Dow Jones began its shocking collapse, market-watchers had noticed that Buffett’s rhetoric about derivative contracts as “weapons of mass destruction” didn’t square well with his actual portfolio. Now the chickens have come home to roost: “Derivatives are dangerous,” Buffett says in 2008, shortly before going on to explain why Berkshire is a party to no fewer than 251 derivatives contracts. “I believe each contract we own was mispriced at inception, sometimes dramatically so,” he says by way of excusing himself. In a time of economic grouchiness, one notices that as an explanation, this is little better than putting one’s hands together to make a whoopee-cushion noise. One party in a contract always thinks the other has “mispriced” the good or service being exchanged. I had understood Buffett’s endless attacks on derivatives to be founded on the premise that pricing them correctly is impossible in principle.

One notices other slightly hucksterish tendencies that might have gone overlooked in the good times. Buffett makes a big deal of endorsing mark-to-market accounting, but uses “book value” in stating the worth of many of his investments, including some of those derivatives. His big leap into insuring tax-exempt municipal and state bonds is outlined in the manner of an old-time mesmerist hypnotizing a maiden lady. Buffett loves the insurance business, in all its forms, because it’s characterized by “float” — the temporary costless use of free money from premiums — but he admits that if those bond-issuing institutions ever start defaulting, a cascade is likely to follow, and the quantitative models are pure shinola in such circumstances.