Power Laws

Power laws express scale relationships through an exponent. A power-law distribution can have a heavy tail in which large observations strongly affect totals, but skewed data alone do not establish that form. The existence of a theoretical mean depends on the exponent and boundaries.

A power law relates quantities through an exponent, often written y = Cx raised to a power. In a distribution, a power-law tail declines slowly enough that large observations can matter greatly. Skew alone does not establish this form: lognormal and other distributions can produce similarly uneven observations over a limited range.

The exponent and boundaries determine what summaries mean. Some idealized power-law distributions have finite means and others do not; every finite dataset still has an arithmetic sample mean. Before adopting a strategy built around rare extremes, assess the fitted range, uncertainty, alternative distributions, and consequences of being wrong. A heavy tail creates a risk-management question as well as an opportunity question.

When to use it

When a proposed scale relationship or heavy-tailed distribution materially affects estimates, exposure, or resource allocation.

How it can help

Examine tail behavior, compare alternatives, and assess how rare large outcomes affect a decision. Use sensitivity analysis rather than assuming every concentrated dataset calls for concentrated bets.