Multidimensional Poverty Index

A dual-cutoff poverty measure that identifies each person or household by a weighted count of simultaneous deprivations, then aggregates the poor population's incidence and average deprivation intensity as MPI = H × A.

Version

v2 · 2026-09-07 · History

Domain-specific #

2325

Origin domain

development economics

Subdomain

multidimensional poverty measurement

Core Idea

A Multidimensional Poverty Index (MPI) is a counting-based measure that identifies people or households experiencing simultaneous deprivations across several dimensions, then summarizes both how many are poor and how broad their deprivation is. In the widely used Alkire–Foster adjusted-headcount form,

\[ M_0 = H\times A, \]

where \(H\) is the proportion of the population identified as multidimensionally poor and \(A\) is the average weighted share of indicators in which poor people are deprived. The product rises if poverty becomes more prevalent, if poor people's deprivation bundles become broader, or both.

Scope of Application

MPI is used in development economics, official statistics, and social policy. The Global MPI supports internationally harmonized measurement of acute multidimensional poverty where suitable household surveys exist. National and subnational measures adapt the method to constitutional, participatory, policy, and data contexts.

The method can be applied to adults, children, households, regions, or other units if indicator ownership and identification are coherent. Household-level indicators often classify every household member alike. That improves data feasibility but can conceal intrahousehold inequalities, so a dossier must say who is observed, who is classified, and whose deprivation an indicator represents.

Clarity

Let \(g^0_{ij}=1\) when unit \(i\) is deprived in indicator \(j\), and zero otherwise. With normalized weights,

\[ c_i=\sum_{j=1}^{d} w_jg^0_{ij},\qquad \rho_k(i)=\mathbf 1[c_i\ge k]. \]

If population or survey weights are equal, the headcount ratio and intensity are

Manages Complexity

Poverty can involve nutrition, mortality, schooling, sanitation, housing, energy, assets, work, security, and other disadvantages that overlap within lives. Separate marginals cannot show whether the same people experience several deprivations. Income alone cannot reveal all direct achievements or service failures.

MPI compresses the joint deprivation matrix while preserving two interpretable components. \(H\) answers how widespread poverty is. \(A\) answers how broad it is among those identified.

Abstract Reasoning

If one poor person's weighted deprivation score rises while nobody else's profile changes, \(A\) and \(M_0\) weakly rise. A plain headcount \(H\) would not respond unless the person's classification crossed \(k\). This is the central advantage of adjustment for intensity.

If a nonpoor person gains an additional deprivation but remains below \(k\), \(M_0\) does not change because nonpoor deprivations are censored. The method is poverty-focused, not a general population deprivation average.

Knowledge Transfer

The Alkire–Foster MPI architecture transfers across countries and themes when simultaneous deprivation is the target and the configuration is explicitly rebuilt. Child poverty, women's poverty, urban poverty, and national constitutional measures can use the same dual-cutoff and aggregation structure with different indicators.

Numbers do not transfer merely because both are called MPI. A Global MPI and national MPI can legitimately classify different people because they answer different normative questions. Cross-setting comparison requires harmonization, not label equality.

Relationships to Other Abstractions

Current abstraction Multidimensional Poverty Index Domain-specific

Parents (1) — more general patterns this builds on

• Multidimensional Poverty Index is a kind of Aggregation Prime

  MPI is a strict domain-specific form of Aggregation: a high-dimensional microdata matrix is deliberately collapsed into one statistic under explicit information-loss choices, while decompositions preserve selected diagnostic routes back.

Hierarchy path (1) — routes to 1 parentless root

• Multidimensional Poverty Index → Aggregation → Micro Macro Linkage

Neighborhood in Abstraction Space

Multidimensional Poverty Index sits in a sparse region of the domain-specific corpus (92^nd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (1565 abstractions)

Nearest neighbors

• Transversal (Combinatorics) — 0.81
• Validity Scale — 0.80
• Statistical Literacy — 0.78
• Family of Origin — 0.78
• Logic Model (Program Evaluation) — 0.77

Computed from structural-signature embeddings · 2026-09-08