So my biases are clear and fully disclosed, I should point out that I am a big fan of James Montier and his work, including his books and his “white papers” both when he was at SocGen and now that he is at GMO. Since I also try to be a deep value investor whose work is informed by behavioral finance, we share the same general approach. And since I met him this afternoon for the first and found him both friendly and engaging, I am obviously inclined to be an appreciative audience. James said that he was a bit nervous about his speech — The Flaws of Finance — because it is a new one, but he needn’t have been. My rough notes follow (with no guaranty of accuracy or comprehensiveness).
The focus: bad models. bad policies, bad incentives and bad behavior.
- If you give CAPM and VAR to monkeys, they’re going to create a financial crisis. Indeed, I think they just did.
- Remember that models are abstractions and don’t represent reality — they have clear weaknesses (low beta outperforms high beta).
- CAPM assumes risk is volatility and ignores liquidity and leverage (and that’s nuts).
- VAR — like a vest that is 95% bulletproof (fails when you need it).
- Bad models and bad assumptions tend to replace common sense.
- Graham — the more abstruse the maths, the more uncertain the results (“complexity to impress”).
- Derman and Wilmott (2008): The Modelers’ Hippocratic Oath.
- VAR — like asking children to grade their own schoolwork.
- Bad policies generate bad incentives.
- Experts tend to have the tendency of giving us “permission” to turn our brains off.
- Anchoring — give someone a number — any number — and s/he thinks it means something (even when it doesn’t).
- Narrow framing — risk means much more than volatility.
- Bad policies encourage bad behavior.
- Asymmetric policy to problems — cut rates; environment of low rates encourages investors to reach for yield (a cardinal sin of investing — following Keynes’s law, that demand creates its own supply); Buffett — “Never ask a barber if you need a haircut.”
- Commission-based loan originators’ loans failed 30% more often than salary-based loan originators; incentives create asymmetric responses.
- With a simple game, the higher the incentive, the worse people played.
- We tend to neglect risk.
- Leverage can’t make a bad investment good but can make a good investment bad.
Problems: Five impediments to recognizing predictable surprises:
- Over-optimism (hope is not a strategy).
- The illusion of control (Kahneman’s “planning fallacy”).
- Self-serving bias (confirmation bias).
- Myopia (lunch is not a long-range plan).
- Inattentional blindness (we don’t see what we don’t expect to see).
A Manifesto for Change
- The Modelers’ Oath (no points for elegance).
- More realistic models.
- More practical experience.
- Less complex math.
- Aim for robustness, not optimality.
- Define risk as the permanent impairment of capital.
- Be skeptical of alleged innovation (it’s usually simply out for leverage).
- Know the limits of your models (and don’t exploit them).
- Focus on the long-term.
- Study history, especially financial history.
- Don’t get bogged down in details (complexity).
- Look for predictable surprises (start with Minsky’s models).
- Central banks should lean against the wind.
- The markets are *not* efficient (science advances one funeral at a time).
- Capital adequacy should be contra-cyclical/
- Guard against regulatory capture and industry-based self-serving bias.
- Lessons from the 2008-09 crisis (not equity-based risk, but expensive equity-based risk — that has sown the seeds of our next crisis as we are now preaching “buy more government bonds”).
Investing is simple but not easy.
- James (and GMO) uses simple models, but it is hard to stick to them.
- The accounts they have the most confidence in are the accounts they are most likely to be fired on.
- Buy when the consensus is selling and vice versa (be contrarian).
- Chasing what is popular is a great business model but an unethical one.
- Sadly, regulators are no less susceptible to behavioral bias than anyone else (the bias blind-spot).