Concentration Curve and Concentration Index

Description

Shows whether an outcome is concentrated among people who rank low or high on another variable, and summarises the gradient in one signed index from -1 to +1.

The concentration curve ranks people by one variable (in health economics, living standard from poorest to richest) and plots the cumulative share of people against the cumulative share of an outcome such as illness or subsidy received. If the curve lies on the diagonal, the outcome is spread evenly across the ranking; above the diagonal it is concentrated among those ranked low, below it among those ranked high. The concentration index is twice the area between the curve and the diagonal, ranging from -1 to +1, and is zero when there is no rank-related inequality. Unlike the Gini coefficient, which measures how unequally one quantity is spread by ranking on that quantity itself, the concentration index is bivariate: it asks whether a good or bad lands disproportionately on one end of another ranking. It says nothing about the average level of the outcome, so it is read alongside means.

Attributes

Loss type
Unmigrated
Effort
Days

Steps

  1. Choose the ranking variable (for example device age, plan tier, income band or digital-skills score) and the outcome whose distribution matters (errors, support contacts, feature benefit).
  2. Rank all cases by the ranking variable from lowest to highest, breaking ties consistently.
  3. Plot the cumulative share of cases against the cumulative share of the outcome, with the line of equality.
  4. Compute the index as twice the area between the curve and the diagonal, signed negative when the curve lies above the diagonal.
  5. Attach bootstrap confidence intervals and compare indices across products, releases or time points.
  6. Interpret the sign according to whether the outcome is a good or a bad, and report it with the outcome's mean.

Sources

  • Quantitative Social Research Methods (Singh, 2007)

Also known as

  • Concentration index
  • CI (health inequality)
  • Concentration curve

Categories

  • Analytics & Behaviour
  • Statistics & Modelling