Yes/No Only Please: When the Bell Curve Applies
Taleb contrasts bounded yes/no observations with outcomes whose magnitudes can be dominated by extremes. A binary outcome is not itself Gaussian; whether a normal approximation works for aggregated results depends on the sampling conditions.
THE IDEA
“Yes/No” Only Please
Let me discuss here the extent of the damage. If you’re dealing with qualitative inference, such as in psychology or medicine, looking for yes/no answers to which magnitudes don’t apply, then you can assume you’re in Mediocristan without serious problems. The impact of the improbable cannot be too large. You have cancer or you don’t, you are pregnant or you are not, et cetera. Degrees of deadness or pregnancy are not relevant (unless you are dealing with epidemics). But if you are dealing with aggregates, where magnitudes do matter, such as income, your wealth, return on a portfolio, or book sales, then you will have a problem and get the wrong distribution if you use the Gaussian, as it does not belong there. One single number can disrupt all your averages; one single loss can eradicate a century of profits. You can no longer say “this is an exception.” The statement “Well, I can lose money” is not informational unless you can attach a quantity to that loss. You can lose all your net worth or you can lose a fraction of your daily income; there is a difference.
This explains why empirical psychology and its insights on human nature, which I presented in the earlier parts of this book, are robust to the mistake of using the bell curve; they are also lucky, since most of their variables allow for the application of conventional Gaussian statistics. When measuring how many people in a sample have a bias, or make a mistake, these studies generally elicit a yes/no type of result. No single observation, by itself, can disrupt their overall findings.
When to use it
Evaluating medical test results where the outcome is binary (e.g., infection present/absent).
Analyzing income distributions to set tax policies, recognizing that extreme wealth skews averages.
Risk management in finance: modeling portfolio returns with heavy-tailed distributions instead of the bell curve.
Interpreting psychological survey data that measures the prevalence of a bias (yes/no response).
FROM THEORY TO EVERYDAY LIFE
In practice
Medical Diagnosis (Yes/No)
Coding a diagnosis as present or absent bounds each observation’s contribution to the sample proportion. This does not make the binary outcome Gaussian or remove problems such as dependence, selection bias, or inaccurate classification.
Income Distribution (Aggregate)
A person with exceptionally high income in a sample of 1,000 people can raise its mean substantially. A Gaussian model may understate that possibility if the income distribution has a heavy upper tail.
A MOMENT TO REFLECT
Put it to work.
What yes/no variable could you study, and what sampling conditions would you check before using a normal approximation?
Identify a real-world situation where a single outlier could completely change the average—how would you handle that?
How might you protect against tail risks in financial portfolios, given that a single loss can wipe out years of profits?
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)