What Does the Gini Index Measure? A Worked Example
What does the Gini index measure?
The Gini index summarizes how unequally income or consumption is distributed within a population. Read the scale and the measured resource before interpreting the number. A useful comparison names the dataset, survey year and definition, then shows what the summary leaves unresolved. The fictional calculation below illustrates distribution; it does not establish a country's living standards or the effects of a policy.
The question is about a distribution: who has what share of a specified total. Our Human Development Index explainer addresses a different kind of composite measure. Neither label should be substituted for the other.
Why do some tables use 0–1 and others 0–100?
The World Bank's Gini metadata describes an index from 0 for equality to 100 for inequality. The OECD indicator presents the coefficient on a 0–1 scale.
On corresponding scales, 0.25 becomes 25 after multiplication by 100. That conversion changes the display, not the underlying distribution. Do not interpret 25 as the percentage of people who are poor, or as the share of income held by one group.
Check the original table's unit rather than inferring it from a screenshot. In your notes, keep both the number and its scale. Otherwise, a copied decimal can appear to describe a much smaller value when it merely uses different notation.
How can you calculate a small fictional example?
The World Bank metadata describes the coefficient through average absolute differences between pairs, divided by twice the mean. For this demonstration, use four equally weighted people with positive fictional incomes measured over the same period. This is an arithmetic illustration, not a household-survey estimator or a real country's data.
| Person | Income in fictional units | Share of the total |
|---|---|---|
| A | 5 | 12.5% |
| B | 5 | 12.5% |
| C | 15 | 37.5% |
| D | 15 | 37.5% |
| Total | 40 | 100% |
The mean is 40 ÷ 4 = 10. Compare each person's income with every person's income, including their own. “Absolute difference” means take the size of the gap without a minus sign.
| Starting person | Difference from A | From B | From C | From D | Row total |
|---|---|---|---|---|---|
| A | 0 | 0 | 10 | 10 | 20 |
| B | 0 | 0 | 10 | 10 | 20 |
| C | 10 | 10 | 0 | 0 | 20 |
| D | 10 | 10 | 0 | 0 | 20 |
There are 4 × 4 = 16 ordered comparisons. Their differences total 80, giving an average difference of 80 ÷ 16 = 5. Divide that by twice the mean: 5 ÷ (2 × 10) = 0.25. Expressed on the 0–100 scale, the index is 25.
Including both A-to-C and C-to-A is intentional. This calculation averages all ordered pairs; it does not count each distinct pair once while retaining the same denominator. A common bookkeeping mistake is to change the pair list but leave the divisor unchanged.
What happens if every income doubles?
Replace the four amounts with 10, 10, 30 and 30. Total income becomes 80 and the mean becomes 20. Each nonzero gap doubles to 20, so the pairwise total becomes 160. The average difference is 160 ÷ 16 = 10; dividing by twice the new mean gives 10 ÷ 40 = 0.25 again.
The same index now accompanies different income amounts. That result follows from this calculation: both its numerator and denominator doubled. It does not establish what those units buy or whether anyone meets a defined poverty threshold.
As another check, give all four people 10 units each. Every pairwise difference is zero, so the coefficient is zero. The total remains 40, just as in the first example, but its distribution has changed. Keep the total and its distribution as separate observations.
Which income definition is being compared?
The OECD's linked indicator uses household disposable income. Its definition includes public cash transfers and deducts household income taxes and social security contributions. Check that definition before comparing its result with another income series.
The World Bank warns that differences in household-survey methods and welfare measures limit comparability across countries and years. A matching index label does not resolve those differences.
For an analysis note, our suggested record is: source table, download date, observation year, measured resource, population, scale and comparability note. Copy the publisher's definitions beside the values. Leave an unanswered field visibly unresolved rather than supplying an assumption.
What conclusion can the number support?
In a World Bank methodological discussion, Francisco Ferreira argues for using multiple inequality measures rather than relying on one summary. He also distinguishes measuring inequality from explaining complex causes. Read his discussion as an argument about interpretation, not a universal policy verdict.
Our editorial rule is to separate the calculation from the claim. “This defined series has a lower index” needs compatible observations. “This programme caused an improvement” needs additional evidence. The development indicators collection keeps those questions distinct; our ODA guide covers the separate measurement of development assistance.