I have been re-reading the books in Nassim Taleb’s “Incerto’ series (Fooled by Randomness, 2001, The Black Swan,2010, The Bed of Procrustes, 2010, Antifragile, 2012, and Skin in the Game, 2018). I am finding it fruitful to read them each several times – there are some profound observations buried in them that one doesn’t easily get on first reading. Here is one:
Beginning in The Black Swan, Taleb makes the useful distinction between things in what he calls “Mediocristan” and things in “Extremistan”.
Think about the average height of humans. Worldwide the average male height is reported to be about 5’9”. If we lined up 100 random men, we would expect the average of their heights to be around 5’9”. If we found the tallest human who had ever lived and added them to the other 100, it wouldn’t change the average much, because the tallest human who ever lived might be 8’ or even 9’ tall, but wouldn’t be 80’ or 800’ or 8,000’ tall. The same would be true for average human weights, or size of dogs, or variation in outdoor temperature in Chicago, or human lifespans…or….or. These are things in Mediocristan, things for which the most extreme possible outliers don’t deviate all that much from the average.
Now think about wealth. In the US the median net worth these days is reported to be about $120,000. And if we picked 100 random Americans and looked at their median wealth it would probably be around that number. But if we added in the net worth of the wealthiest American living (currently Amazon’s Jeff Bezos at about $177 billion) it would change the average net worth immensely. And in fact there is no theoretical limit to how wealthy the wealthiest American might be. Or think about the size of planetary debris hitting the earth. The average piece of debris is a dust particle, but the largest possible could be the size of a planet larger than earth. These are things in Extremistan, things which have outliers very, very far from the median.
What is the importance of this distinction? It is that Mediocristan doesn’t have black swans, so that our common statistical methods of estimating risk work reasonably well, given enough data points in the history to get a valid feeling for the shape of the underlying distribution. But Extremistan does have black swans, which means that our common statistical methods of estimating risk are meaningless, irrespective of how much historical data we have. The turkey story is illustrative – for 1000 days it gets fed regularly, so from historical data it can reasonably expect to get fed every day, then Thanksgiving arrives…
The difficulty is that too many experts in too many fields think they are working in Mediocristan with their models and theories and risk estimates, when they are really in Extremistan.
Posts
