Multidimensional Deprivation Analysis
- Introduced by
- Anthony B. Atkinson; Sabina Alkire & James Foster
- Year
- 2003
- Type
- Family of multidimensional deprivation measurement approaches
- Family
- Multidimensional poverty measurement
What does it do?
Multidimensional deprivation analysis is the broad family of methods for measuring and describing disadvantage across several dimensions at once — health, education, living standards, work, and more — rather than through income alone.
It spans the counting approach championed by Anthony Atkinson and formalized by Sabina Alkire and James Foster, the dashboard tradition of reporting deprivation indicators side by side, fuzzy-set treatments that soften sharp thresholds, and overlap analysis that asks whether the same people are deprived in many dimensions. The unifying questions are how to decide who is deprived in each dimension, how to identify the multiply deprived, and whether to summarize deprivation in one index or display it as a panel of indicators.
The intuition behind the method
Start by deciding, dimension by dimension, who falls short of an acceptable threshold — no schooling, poor sanitation, bad health, insecure work. Now you face two choices that define the whole field. The first is identification: do you call someone deprived overall if they are short in any one dimension (the union view), only if short in all of them (the intersection view), or somewhere in between (a dual cutoff that requires a minimum count)? The second is presentation: do you aggregate the deprivations into a single headline number, which is easy to communicate but buries detail and forces you to weight incommensurable things, or do you keep a dashboard of separate indicators, which is honest about plurality but gives no overall ranking? Different methods sit at different points on these two axes.
History
Anthony Atkinson's 2003 Journal of Economic Inequality paper drew the sharp contrast between the social-welfare approach to multidimensional deprivation, which aggregates dimensions through a welfare function, and the counting approach, which simply counts the dimensions in which people are deprived — and showed how the latter underlay European social-inclusion indicators. The intellectual roots run back to Townsend's relative-deprivation work, Sen's capability approach, and the European tradition of social-exclusion dashboards. Alkire and Foster's 2011 paper formalized the counting approach with their dual-cutoff identification and the adjusted-headcount family, supplying the axiomatic backbone now used in the global and national multidimensional poverty indices, while fuzzy-set and overlap methods developed in parallel within the broader literature.
When to use it
Use multidimensional deprivation analysis when poverty or disadvantage is genuinely plural and an income line would misrepresent it, and you need to decide deliberately how to identify the deprived and whether to summarize or display the results. The framework is the right lens for designing a national or thematic deprivation measure, for diagnosing which dimensions and which overlaps drive disadvantage, and for choosing between an index (when a single comparable number is needed) and a dashboard (when transparency about distinct dimensions matters more). It demands explicit, contestable choices of dimensions, thresholds, weights, and identification rule, so it suits settings where those normative choices can be justified and subjected to robustness checks rather than hidden.
Limitations and considerations
- Every result hinges on contestable normative choices — dimensions, thresholds, weights, identification rule — and can shift substantially under reasonable alternatives.
- The union and intersection identification rules are usually too extreme to be useful, forcing an arbitrary intermediate cutoff whose level is hard to justify.
- Aggregating incommensurable dimensions into one index requires weights that have no objective basis, while dashboards give no overall ranking — neither resolves the underlying tension.
- Counting approaches based on binary deprivation discard information on the depth of shortfall within each dimension.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Captures disadvantage across multiple dimensions and, through overlap analysis, whether deprivations pile on the same people — information income measures and single indices miss.
- Makes the two key choices — identification rule and aggregation versus dashboard — explicit, so methodological decisions are transparent and debatable.
- Flexible across data types and contexts: works with ordinal indicators, accommodates fuzzy thresholds, and supports both index and dashboard presentation.
- Grounds measurement in the capability approach, giving a principled rationale for which dimensions of deprivation to include.
Applications
- Designing national multidimensional deprivation or social-exclusion measures and the dashboards that accompany them.
- Diagnosing which dimensions and which overlapping combinations of deprivation drive disadvantage in a population.
- European Union social-inclusion monitoring through the Laeken indicators and material-deprivation dashboards.
- Comparing counting-based, fuzzy-set, and dashboard treatments of the same data to test how robust conclusions are to method.
- Targeting and sequencing of policy by revealing whether tackling one dimension would also relieve co-occurring deprivations.
How it works
How it works
The first step selects the dimensions of well-being to consider and sets a deprivation threshold z_j in each. A person is flagged as deprived in dimension j when their achievement x_ij falls below that threshold, producing a binary deprivation indicator d_ij. The choice of dimensions and thresholds is normative and grounded in theory (often the capability approach), data availability, and public deliberation; it determines what 'falling short' means before any aggregation.
Formula
d_{ij} = \mathbb{1}\!\left(x_{ij} < z_j\right)The individual deprivation indicators are combined into a weighted deprivation count or score, c_i = sum of w_j times d_ij, where the weights w_j (often summing to one) express the relative importance of each dimension. This count summarizes in how many — and, with weights, in which — dimensions a person is simultaneously deprived, and it is the raw material for deciding who counts as multidimensionally deprived.
Formula
c_i = \sum_{j=1}^{d} w_j\, d_{ij}, \qquad \sum_{j} w_j = 1An identification rule turns the deprivation count into a poor/non-poor distinction. The union rule counts anyone deprived in at least one dimension; the intersection rule counts only those deprived in every dimension; the dual-cutoff rule of Alkire and Foster counts those whose weighted count reaches a threshold k, between the two extremes. The union rule tends to label almost everyone deprived, the intersection rule almost no one, so an intermediate cutoff is usually adopted to capture genuinely joint deprivation.
Formula
\text{union: } c_i > 0 \;\;\Big|\;\; \text{intersection: } c_i = 1 \;\;\Big|\;\; \text{dual cutoff: } c_i \ge kFinally the analyst chooses how to present deprivation. Aggregation produces a single index — for example the adjusted headcount, or fuzzy-set scores that replace 0/1 deprivation with degrees of membership — which is communicable but requires weighting incommensurable dimensions and obscures the pattern. The dashboard approach instead reports each deprivation indicator separately and supplements it with overlap analysis showing how deprivations co-occur in the same people, preserving detail at the cost of a single ranking. The choice between a welfare-style aggregate and a counting/dashboard display is the central methodological fork Atkinson highlighted.
Common pitfalls
- Defaulting to a union or intersection rule without realizing each is an extreme that labels almost everyone, or almost no one, deprived.
- Presenting a single aggregate index while hiding the weighting and identification choices that drive it, giving a false sense of objectivity.
- Reporting a dashboard of separate indicators but never analyzing overlap, so joint deprivation concentrated in the same people goes unseen.
- Treating the dimensions, thresholds, and weights as objective givens rather than running robustness checks over reasonable alternatives.
Frequently asked questions
What is the difference between the union, intersection, and dual-cutoff identification rules?
They are three answers to 'who counts as multidimensionally deprived?'. The union rule flags anyone deprived in at least one dimension, which tends to label almost the whole population as deprived. The intersection rule flags only those deprived in every dimension, which usually identifies almost no one. The dual-cutoff rule, central to the Alkire-Foster method, sits between them: a person is identified as deprived if their weighted count of deprivations reaches a chosen threshold k. The intermediate cutoff captures genuinely joint deprivation while avoiding the extremes.
When should I use a single index rather than a dashboard?
Use a single aggregate index when you need one comparable, communicable number for ranking places or tracking change over time, and you can defend the weights and thresholds it requires. Use a dashboard when transparency about distinct, possibly incommensurable dimensions matters more than a single ranking, or when stakeholders disagree about weights. The two are complementary: many analyses report a headline index for communication and a dashboard plus overlap analysis for detail, which is the honest way to present multidimensional deprivation.
How does this broader analysis relate to the Multidimensional Poverty Index?
The Alkire-Foster Multidimensional Poverty Index is one specific, fully aggregated instance of multidimensional deprivation analysis: it fixes a dual-cutoff identification rule and the adjusted-headcount aggregation. The broader analysis here situates that index within a wider menu of choices — union and intersection rules, fuzzy-set deprivation, and dashboard-plus-overlap presentation — and foregrounds the methodological forks (identify how, aggregate or display) that the MPI resolves in one particular way. Understanding the family clarifies what the MPI assumes and what alternatives exist.
Related methods
Sources
Atkinson, A. B. (2003). Multidimensional Deprivation: Contrasting Social Welfare and Counting Approaches. Journal of Economic Inequality, 1(1), 51-65.
Open DOI referenceCopied.Alkire, S., & Foster, J. (2011). Counting and multidimensional poverty measurement. Journal of Public Economics, 95(7-8), 476-487.
Open DOI referenceCopied.