Kakwani Index¶
A rank-based progressivity measure comparing an intervention's concentration coefficient with pre-intervention income inequality under a declared ranking and sign convention.
Core Idea¶
The Kakwani index is a rank-based measure of the progressivity of a tax, health payment, or other social intervention. Under a common tax convention, it is calculated as K = C − G, where C is the concentration coefficient of payments ranked by pre-intervention income and G is the Gini coefficient of that income. A positive value then indicates progressivity and a negative value regressivity.
The formula is inseparable from its convention. Some applications treat benefits or use a reversed subtraction, changing the sign interpretation. A valid report must therefore name the ranking variable, intervention, coefficients, weighting, and subtraction order rather than present a bare number.
Scope of Application¶
The index is used in public finance, health-financing equity, and distributional analysis to compare how an intervention is distributed relative to ability to pay. It can compare systems or periods when populations, definitions, and sign conventions are aligned.
Because both coefficients use the same pre-intervention ranking, reranking after the intervention belongs to a broader redistributive analysis and cannot be read directly from this index.
It is not the Gini coefficient itself, a complete measure of redistribution, or a direct measure of welfare. A highly progressive intervention may have little redistributive effect if its overall magnitude is small.
Clarity¶
The abstraction separates two questions: how unequal pre-intervention income is, and how concentrated payments or benefits are along that same ranking. Their difference states whether the intervention departs from proportionality. Writing the formula and sign convention makes otherwise contradictory-looking results comparable and prevents a progressive tax from being mislabeled because the subtraction order changed.
Manages Complexity¶
An entire ranked distribution is compressed into two coefficients and their difference. That compactness enables comparison and decomposition, but it loses information about where departures occur, how large the intervention is, and which households change rank. Concentration curves, incidence tables, uncertainty estimates, and redistributive measures are therefore useful companions rather than substitutes.
Abstract Reasoning¶
Define the population and pre-intervention ranking, calculate income shares and G, calculate intervention shares and C over the identical ranking, and apply the declared convention. Check survey weights, zero and negative incomes, ties, sampling uncertainty, and comparability across datasets. Interpret the sign only after these choices are fixed. Then inspect curves or subgroup results to learn which parts of the distribution generate the summary.
Knowledge Transfer¶
The procedure transfers across taxes, insurance payments, fees, and benefits when the ranked population and intervention meaning are explicit. Numerical values do not transfer across different rankings or conventions. As a kind of measurement, the index shows how an equity judgment is operationalized through a defined scale and procedure. It does not eliminate normative questions about the appropriate income concept, unit of analysis, or desirable distribution.
Relationships to Other Abstractions¶
Current abstraction Kakwani Index Domain-specific
Parents (1) — more general patterns this builds on
-
Kakwani Index is a kind of Measurement Prime
Kakwani Index is a strict kind of Measurement: A rank-based progressivity measure comparing an intervention's concentration coefficient with pre-intervention income inequality under a declared ranking and sign convention.
Hierarchy path (1) — routes to 1 parentless root
- Kakwani Index → Measurement
Neighborhood in Abstraction Space¶
Kakwani Index sits in a sparse region of the domain-specific corpus (99th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Lorenz Curve — 0.78
- Theil Index — 0.77
- Outbreak Underascertainment — 0.75
- Clinical-Trial Stratification — 0.74
- Systematic sampling — 0.74
Computed from structural-signature embeddings · 2026-10-08