Gini Coefficient Health

Gini Coefficient Health: Concept Architecture

Orientation and learning roadmap

This page uses a specified health measure to show how inequality across people can be summarised. It builds the calculation from individual values, compares health achievement with health shortfall, and explains which equity questions this statistic leaves unanswered.

Specify the health distribution first

Identify the population, observation period, and nonnegative cardinal measure before computing the statistic. Health-adjusted life-years over a common horizon, for example, can be ordered and compared numerically; an ordinal self-rated-health category cannot be treated as though gaps between categories have a natural size without a defensible transformation.

Each observation should refer to a comparable unit, such as an individual, or use explicit population weights if observations represent different numbers of people. State whether larger values mean better health or greater ill health, because the interpretation of concentration at the low and high ends reverses. The World Bank guide to analysing health equity distinguishes the overall distribution of health from inequality associated with socioeconomic rank.

Calculate the relative spread

For nonnegative individual health values, order people from the smallest to largest value and plot the cumulative share of the measured health against the cumulative share of people. The Gini coefficient is twice the area between that Lorenz curve and the diagonal representing equal measured values.

For $n$ equally weighted individuals with values $h_i$ and positive mean $\bar h$, the same quantity can be written:

$$G_h=\frac{\sum_{i=1}^{n}\sum_{j=1}^{n}|h_i-h_j|} {2n^2\bar h}.$$

Zero means everyone has the same measured value; larger values indicate greater relative dispersion, with the finite-sample maximum below one for a fixed number of equally weighted people. The formula is undefined when the mean is zero, and the usual Lorenz interpretation requires nonnegative values. Multiplying every health value by the same positive constant leaves the coefficient unchanged, but adding a constant generally changes it.

Worked four-person distribution

Suppose four people accrue 2, 2, 4, and 8 health-adjusted life-years over a common ten-year horizon. Their mean is 4 years, and the sum of the six unordered pairwise absolute differences is $0+2+6+2+6+4=20$ years.

Counting each pair in both directions gives 40, so the coefficient is $40/(2\times4^2\times4)=0.3125$. The cumulative health shares after the four sorted people are $2/16$, $4/16$, $8/16$, and $16/16$, which trace the associated Lorenz curve. The value describes dispersion in this small distribution, not sampling precision or the moral importance of any particular gap.

Why achievement and shortfall differ

A health shortfall can be defined against a stated maximum or target, but it is a different variable from health achievement. Even when the absolute pairwise differences are unchanged, changing the mean changes the denominator and therefore changes the Gini coefficient.

Against a ten-year maximum, the example's health shortfalls are 8, 8, 6, and 2 years, with mean 6 years. Their pairwise absolute differences still total 40 when ordered pairs are counted, but their Gini coefficient is $40/(2\times4^2\times6)\approx0.2083$. Do not report the achievement coefficient as an ill-health coefficient, or compare them without stating the horizon, transformation, and direction of measurement.

What an equity comparison can show

The coefficient describes overall relative inequality in the measured health outcome without identifying which socioeconomic group has more or less health. A concentration index instead ranks people by a separate variable, commonly socioeconomic position, and answers a different distributional question.

The WHO health inequality monitoring manual emphasises choosing and reporting indicators appropriate to the inequality being assessed. A change in a Gini coefficient alone does not establish whether everyone improved, whether disadvantaged groups benefited, or whether a policy caused the change.

Compare results responsibly

Report the underlying health measure and mean alongside the coefficient, and check whether population composition, age structure, missing outcomes, sampling weights, and observation horizons differ across settings. Bootstrap or other design-appropriate methods can quantify sampling uncertainty when results come from a survey rather than a full population.

Health measures bounded above, such as years of full health during a fixed period, can restrict the attainable coefficient in ways that depend on the mean and scale. Present distribution plots or group breakdowns when possible so that a single number does not conceal where disparities occur. Keep descriptive inequality separate from value judgements about fairness and from causal claims about health policy.