Avoid Conventional Risk Metrics in Fat-Tailed Domains
Taleb warns that familiar summary measures can conceal extreme-loss exposure in domains dominated by rare events. His critique concerns whether those measures adequately represent the risks being taken.
THE IDEA
6. Avoid Some Risk Metrics
Conventional metrics, based on Mediocristan, adjusted for large deviations, don’t work. This is where suckers fall in the trap—one far more extensive than just assuming something other than the Gaussian bell curve. Words like “standard deviation” are not stable and do not measure anything in the Fourth Quadrant. Neither do “linear regression” (the errors are in the Fourth Quadrant), “Sharpe ratio,” Markowitz optimal portfolio, ANOVA shmanova, Least square, and literally anything mechanistically pulled out of a statistics textbook. My problem has been that people can accept the role of rare events, agree with me, and still use these metrics, which leads me to wonder if this is a psychological disorder.
When to use it
Assessing the risk of a portfolio containing high-yield bonds and emerging market equities during a financial crisis.
Evaluating a fund that earns premiums by selling protection against rare losses.
Modeling insurance losses whose totals may be dominated by rare catastrophes.
Analyzing the risk of algorithmic trading strategies during volatile market periods.
FROM THEORY TO EVERYDAY LIFE
In practice
Portfolio Risk Using Standard Deviation
A manager estimates a portfolio’s risk from recent standard deviation while holding leveraged ETFs and low-liquidity bonds. A crash produces losses far outside that recent record. Standard deviation can summarize variation without capturing the full loss exposure; the measure itself does not require a normal distribution.
Sharpe Ratio Misleading Hedge Fund Performance
A hedge fund sells protection against rare losses and receives small, regular premiums. Its recent Sharpe ratio looks attractive until a severe event triggers a large payout. The ratio alone does not describe the strategy’s asymmetric loss exposure.
A MOMENT TO REFLECT
Put it to work.
Why is standard deviation unreliable as a risk measure in fat-tailed distributions?
How might using the Sharpe ratio mislead investors in extreme market conditions?
Can you think of a real-world situation where linear regression assumptions are violated due to heavy tails?
CONTINUE THE CONVERSATION
Source texts
- The Black Swan The Impact of the Highly Improbable With a new section On Robustness and Fragility (Nassim Nicholas Taleb)