Skip to content
Back to search results

Concentration Curve and Index

Introduced by
Adam Wagstaff, Pierella Paci & Eddy van Doorslaer
Year
1991
Type
Bivariate inequality measure
Family
Socioeconomic health inequality measurement
Font size

What does it do?

The concentration curve and concentration index, established as the standard tools for measuring socioeconomic inequality in health by Wagstaff, Paci, and van Doorslaer in 1991, capture how a health variable is distributed across the population ranked by socioeconomic status.

The concentration curve plots the cumulative share of health (or ill-health) against the cumulative share of people ordered from poorest to richest; the concentration index is twice the area between this curve and the line of equality. Unlike the Gini coefficient, which measures pure dispersion, the concentration index is bivariate — it measures inequality in one variable that is systematically related to a second, socioeconomic ranking.

The intuition behind the method

Ask not just whether health is unequally distributed, but whether ill-health is concentrated among the poor. Line everyone up from poorest to richest and trace the running total of, say, illness. If illness were spread evenly the cumulative curve would track the 45-degree line. If the poor bear more than their population share of illness, the curve bulges above the line; if the rich do, it sags below. The concentration index condenses this bulge into one number between minus one and plus one: negative when ill-health concentrates among the poor (the usual finding), positive when it concentrates among the rich, and zero when there is no socioeconomic gradient. It is the Gini's cousin, but the ranking is by income, not by health itself.

History

Adam Wagstaff, Pierella Paci, and Eddy van Doorslaer set out the case for the concentration index in their 1991 Social Science and Medicine paper 'On the Measurement of Inequalities in Health', arguing that measures of pure dispersion such as the Gini or the range fail to reflect whether ill-health is borne disproportionately by the poor, and that an SES-ranked concentration index does. The tool drew directly on the Lorenz-curve and concentration-curve apparatus of income distribution analysis. Subsequent work — notably by Wagstaff and by Erreygers — addressed the normalization of the index for bounded and binary variables, and the regression-based decomposition (Wagstaff, van Doorslaer, and Watanabe) made it a workhorse of the health-equity literature.

When to use it

Use the concentration curve and index when you want to measure and test inequality in a health variable that is systematically related to socioeconomic status — for example, whether child mortality, undernutrition, unmet care needs, or health spending falls disproportionately on the poor. They are the standard tools in health economics and global-health equity analysis precisely because they incorporate the socioeconomic ranking that a Gini ignores. Use the curve for a visual, dominance-style comparison and the index for a single summary that can be tested and decomposed into the contributions of determinants. Be careful with bounded or binary health variables, where the raw index has restricted range and a normalized version (Wagstaff or Erreygers) is appropriate, and report whether you measure health or ill-health since that flips the sign.

Limitations and considerations

  • The raw index has a restricted range and changes with the mean when the health variable is binary or bounded, requiring Wagstaff or Erreygers normalization.
  • It measures association with the SES ranking, not causation; a nonzero index does not establish why health and SES are related.
  • Results depend on the choice and measurement of the socioeconomic ranking variable, which is often a noisy wealth index.
  • The sign convention is easy to misread, since whether a negative index is 'bad' depends on whether the variable measures health or ill-health.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Explicitly measures the socioeconomic gradient in health by ranking on SES, capturing inequality the univariate Gini cannot.
  • The signed index reveals both the direction (pro-poor or pro-rich) and the magnitude of inequality in a single interpretable number.
  • The covariance form supports a regression-based decomposition into the contributions of individual health determinants for explanation and policy.
  • The concentration curve enables dominance comparisons that are robust to the exact functional form of the inequality summary.

Applications

  • Measuring socioeconomic inequality in child mortality, malnutrition, and immunization coverage across and within countries.
  • Assessing the distribution of health-care utilization and out-of-pocket health spending by socioeconomic status.
  • Evaluating whether health programs and subsidies reach the poor (benefit-incidence and equity analysis).
  • Decomposing health inequality into the contributions of income, education, location, and other determinants.
  • Monitoring progress toward universal health coverage and health-equity targets over time.

How it works

How it works

1

Every individual is ranked by a continuous socioeconomic status variable — income, consumption, or a wealth index — from poorest to richest, and assigned a fractional rank R_i = (i − 0.5)/N giving their relative position in the SES distribution. This ranking variable, not the health variable, defines the horizontal axis, which is what distinguishes the concentration apparatus from a univariate inequality measure that would rank by the health variable itself.

Formula
R_i = \frac{i - 0.5}{N}, \qquad i = 1, \dots, N \;\; (\text{individuals sorted by SES, poorest first})
2

The concentration curve L_h(p) plots, for each cumulative population proportion p (ordered by SES), the corresponding cumulative share of the health variable h. If health were independent of SES the curve would coincide with the diagonal of perfect equality. A curve lying above the diagonal means the health variable is concentrated among the poor; a curve below means it is concentrated among the rich. For an ill-health variable (where more is worse), a curve above the diagonal signals a pro-poor concentration of bad health.

Formula
L_h(p) = \frac{1}{N\bar{h}} \sum_{i=1}^{\lceil pN \rceil} h_{(i)}, \qquad p \in [0,1]
3

The concentration index C is twice the area between the diagonal and the concentration curve, signed so that C is negative when the curve lies above the diagonal (variable concentrated among the poor) and positive when below. Equivalently it can be computed from the convenient formula C = (2/(N times mean health)) times the sum of h_i R_i, minus one. Its sign reveals the direction of the socioeconomic gradient and its magnitude the strength; C ranges from minus one to plus one.

Formula
C = 2 \int_0^1 \bigl[p - L_h(p)\bigr]\, dp = \frac{2}{N\bar{h}} \sum_{i=1}^{N} h_i R_i - 1
4

The index has an illuminating covariance form: C equals two over mean health times the covariance between the individual health variable and the individual fractional SES rank. This makes explicit that the concentration index measures the degree to which the health variable co-varies with socioeconomic position. It also underpins the regression-based decomposition of the index into the contributions of health determinants, and motivates corrections — such as the Wagstaff and Erreygers normalizations — needed when the health variable is binary or bounded.

Formula
C = \frac{2}{\bar{h}}\,\operatorname{cov}(h_i, R_i), \qquad C \in [-1, 1]

Common pitfalls

  • Applying the standard concentration index to a binary outcome without a bounded-variable correction, so the index is mechanically constrained by the mean.
  • Misinterpreting the sign — forgetting that for an ill-health variable a negative index means worse health among the poor, which is the adverse case.
  • Treating the index as causal and reading determinant contributions from its decomposition as policy effects rather than associations.
  • Comparing indices across populations with different SES measures or different prevalence without harmonizing the ranking variable and applying normalization.

Frequently asked questions

How is the concentration index different from the Gini coefficient?

Both are twice the area between a cumulative curve and the line of equality, but they rank the population differently. The Gini ranks people by the variable being measured (income by income), so it captures pure dispersion and is always nonnegative. The concentration index ranks people by a separate socioeconomic variable and measures how the health variable co-varies with that ranking, so it is signed: negative when the variable concentrates among the poor and positive when among the rich. The concentration index is therefore a bivariate measure of the socioeconomic gradient, not of dispersion alone.

What does the sign of the concentration index mean?

A negative index means the health variable is concentrated among the poor (the concentration curve lies above the diagonal); a positive index means it is concentrated among the rich. The welfare reading depends on the variable. For a good-health variable (e.g., having insurance), a negative index is adverse because the poor have less. For an ill-health variable (e.g., illness or mortality), a negative index is also adverse because the poor bear more bad health. Always state whether the variable measures health or ill-health when interpreting the sign.

Why do binary health variables need a corrected index?

When the health variable is binary (or otherwise bounded), the minimum and maximum attainable values of the standard concentration index depend on the mean prevalence, so the raw index cannot reach plus or minus one and is not comparable across populations with different prevalence. Normalized versions — the Wagstaff index, which rescales by the mean, and the Erreygers index, which rescales by the variable's range — restore comparability and a full range. The choice between them depends on whether relative or absolute inequality is the quantity of interest.

Related methods

Sources

  1. Wagstaff, A., Paci, P., & van Doorslaer, E. (1991). On the measurement of inequalities in health. Social Science & Medicine, 33(5), 545–557.

    Open DOI reference
    Copied.