Power Law: Definition and What Self-Similarity Means

A power law is a relationship where one quantity scales as a mathematical power of another. If you double your money, your wealth doesn't double—it increases by 2^n for some exponent n. If a city has double the population of another city, it has 2^n times the economic activity, not 2 times. The relationship isn't linear; it's exponential.

The famous example is Pareto's 80/20 rule: 80% of effects come from 20% of causes. But that's just one snapshot. The deeper insight is self-similarity. If you look at the top 20% of the wealthiest people, you'll find 80% of their wealth comes from the top 20% of them. Zoom in on that subgroup, and the same pattern repeats. The distribution of the richest 1% looks like the distribution of the entire population—just more extreme. The structure is fractal.

This is radically different from a normal distribution, where the average tells you something useful. If human heights follow a bell curve, the average height predicts the distribution well. Most people cluster around the mean. But in a power-law distribution—wealth, city sizes, book sales—the average is misleading. You can't say "the average person's wealth is X" because extreme outliers dominate. One billionaire skews the entire measure.

Why the Average Lies in Extremistan

I think about this constantly when analyzing markets or evaluating risk. In a power-law world, the average conceals what's actually happening. The average return on a stock market might be positive, but 90% of the returns come from 10% of the days—days you can't predict. The average book earns its author very little; 1% of books earn 50% of all book revenue.

The Pareto principle itself—80/20—is just the name for the visible part of a power law. But it implies the structure continues. In 1970, Nassim notes, 50% of all income in the United States came from 2% of earners. That's a 50/2 rule, which is even more extreme. Not 80/20, but closer to 99/1 in the tail.

What makes power laws crucial to Black Swan thinking is that extreme events are intrinsic to the distribution, not rare accidents. In a normal distribution, a 10-sigma event is impossible. In a power-law distribution, extreme events happen regularly. The largest city is often 10 times larger than you'd predict from the median. The wealthiest person often has 1,000 times the average wealth. These aren't anomalies; they're the structure of the system.

Go deeper:

Understand how power laws shape the distributions at the heart of Extremistan: /articles/the-black-swan/power-law-distribution/