The Multidimensional Poverty Index (MPI) is a measure of poverty that identifies people experiencing overlapping disadvantages and summarizes both how many people are poor and how many deprivations they face. Its best-known version, the global MPI, covers health, education, and living standards. It was introduced in 2010 by the United Nations Development Programme (UNDP) and the Oxford Poverty and Human Development Initiative (OPHI). National MPIs use the same general approach but may select different dimensions, indicators, and thresholds to reflect national circumstances.
Conceptual basis
Monetary poverty measures classify people according to whether income or consumption falls below a poverty line. These measures capture command over economic resources, but they do not directly establish whether people have adequate schooling, nutrition, housing, or basic services. Public provision, local infrastructure, household circumstances, and discrimination can affect the relationship between money and actual living conditions.
The MPI measures selected disadvantages directly and examines whether they occur together for the same people. It is influenced by the capability approach, which understands disadvantage in terms of restrictions on what people can do and be, rather than only the resources they possess.
A central distinction is between a dashboard of separate indicators and a multidimensional poverty measure. A dashboard might report the proportions of a population lacking electricity, schooling, and adequate sanitation. It does not necessarily show whether the same households experience all three disadvantages. The MPI uses joint information to identify overlapping deprivation.
The index complements monetary poverty measures rather than replacing them. A person can be monetarily poor without meeting the MPI poverty threshold, or multidimensionally poor without falling below an income poverty line.
Historical development
The global MPI first appeared in UNDP’s Human Development Report in 2010, replacing the Human Poverty Index in that reporting framework. Its construction drew on the Alkire–Foster method, developed by Sabina Alkire and James Foster.
Unlike the Human Development Index, which summarizes average national achievements in broad dimensions of human development, the MPI is constructed from information on overlapping deprivations among individuals or households. Two countries with similar average achievements can therefore have different multidimensional poverty profiles.
The global MPI methodology was revised in 2018, including changes to several indicator definitions. Consequently, estimates produced under different specifications cannot automatically be treated as a consistent time series. Comparisons over time require harmonized definitions and sufficiently comparable survey data.
Alongside the global measure, governments have developed national MPIs for domestic monitoring and policy analysis. These are related measures, not necessarily interchangeable versions of a single international index.
Dimensions and indicators of the global MPI
The global MPI contains three equally weighted dimensions and ten indicators. Each dimension receives a weight of one-third.
| Dimension | Indicator | Weight |
|---|---|---|
| Health | Nutrition | |
| Health | Child mortality | |
| Education | Years of schooling | |
| Education | School attendance | |
| Living standards | Cooking fuel | |
| Living standards | Sanitation | |
| Living standards | Drinking water | |
| Living standards | Electricity | |
| Living standards | Housing | |
| Living standards | Assets |
These indicators cover different aspects of deprivation:
- Nutrition identifies households containing an eligible person who is undernourished, using age-appropriate measures.
- Child mortality records the death of a child under 18 in the household during the five years preceding the survey.
- Years of schooling assesses whether any eligible household member has completed at least six years of schooling.
- School attendance assesses whether school-aged children attend school through the age at which they would complete class eight.
- Cooking fuel identifies reliance on specified solid fuels, such as wood, dung, charcoal, or coal.
- Sanitation identifies unimproved facilities or improved facilities shared with other households.
- Drinking water considers both whether the source is improved and the time needed to collect water.
- Electricity records whether the household has access to electricity.
- Housing assesses the materials used for floors, roofs, and walls.
- Assets assesses ownership of a specified set of goods and vehicles.
These are abbreviated descriptions. Operational rules specify age eligibility, household composition, missing-data treatment, and exact deprivation cutoffs. An “improved” water source is a technical classification and does not by itself establish that water is free from contamination or continuously available.
The global MPI is generally calculated using household surveys containing the necessary indicators together, notably Demographic and Health Surveys and Multiple Indicator Cluster Surveys. Combining separate national averages would not reveal which deprivations overlap within households.
Calculation
Identifying deprivation
For each person and indicator , define a binary deprivation status:
Let be the weight assigned to indicator , with:
The person’s weighted deprivation score is:
In the global MPI, indicator deprivations are generally assigned at household level: household members receive the same deprivation profile even when an indicator is based on information about particular members. The resulting poverty estimates are expressed in terms of people, not simply numbers of households.
Identifying multidimensionally poor people
A second threshold, the poverty cutoff , determines how much weighted deprivation constitutes multidimensional poverty:
For the global MPI:
Thus, a person is classified as multidimensionally poor when their household is deprived in indicators accounting for at least one-third of the total weight.
This is a dual-cutoff procedure: indicator-specific cutoffs first establish individual deprivations, and the cross-indicator poverty cutoff then identifies multidimensionally poor people.
Incidence, intensity, and the index
The headcount ratio, or incidence of poverty, is:
where is the number of multidimensionally poor people and is the total population.
The intensity of poverty is the average weighted deprivation score among poor people:
The MPI is their product:
These expressions assume equally weighted population observations. Survey estimates incorporate sampling weights and the number of people represented by each household. If no one is identified as poor, the MPI is zero.
In the Alkire–Foster framework, this measure is called the adjusted headcount ratio, often denoted . Deprivations experienced by people below the poverty cutoff do not enter its value; this procedure is known as censoring.
Numerical example
Suppose 30% of a population is multidimensionally poor and poor people experience, on average, 50% of the total weighted deprivations. Then:
An MPI of 0.15 does not mean that 15% of people are poor. The poverty rate is 30%; the index combines that rate with deprivation intensity.
Interpretation and decomposition
The MPI ranges from zero to one. Zero means that no one meets the selected multidimensional poverty cutoff, not necessarily that every deprivation has disappeared. One means that everyone is deprived in every indicator.
Population subgroups
The index can be decomposed by mutually exclusive population groups:
where is group ’s population share. This allows national poverty to be examined by region, rural or urban residence, or other characteristics, subject to adequate sample sizes.
A small region can have a high MPI but contribute less to the national total than a much larger region with a lower MPI.
Indicator contributions
A censored headcount ratio for an indicator is the proportion of the total population who are both multidimensionally poor and deprived in that indicator. The MPI equals the sum of these ratios multiplied by their indicator weights.
This decomposition shows which indicators account for the measured poverty burden. It does not establish that those indicators are the underlying causes of poverty or that changing them would produce the largest policy effect.
Related classifications
Global MPI reporting also distinguishes:
- Vulnerability to multidimensional poverty: a weighted deprivation score of at least one-fifth but below one-third.
- Severe multidimensional poverty: a score of at least one-half.
Here, “vulnerability” describes a deprivation band; it is not an estimated probability of becoming poor. Likewise, intensity measures the breadth of weighted deprivation, not how far someone falls below each indicator’s adequacy threshold.
National MPIs and policy uses
National MPIs adapt the framework to local priorities and data. They may include employment, housing security, access to services, or other dimensions absent from the global MPI. Their weights and poverty cutoffs may also differ.
Applications include:
- monitoring poverty beyond income or consumption;
- identifying geographic concentrations of overlapping disadvantage;
- coordinating policies across education, housing, infrastructure, and other sectors;
- examining differences between population groups;
- informing social protection planning and evaluating changes in deprivation profiles.
The Sustainable Development Goals distinguish international monetary poverty from poverty in all its dimensions according to national definitions. National MPIs can contribute to monitoring the latter, but there is no single mandatory national MPI specification.
An aggregate MPI does not automatically constitute a household eligibility system. Survey estimates describe populations, whereas selecting beneficiaries requires suitable administrative information and separate operational rules.
Limitations and methodological debates
Normative choices
Dimensions, thresholds, weights, and the poverty cutoff embody judgments about what counts as deprivation. Equal dimension weights are transparent, but they are not a uniquely correct valuation of human needs.
Alternative defensible choices can change poverty levels, rankings, and subgroup comparisons. Sensitivity analysis examines whether conclusions remain stable across such choices.
Household-level identification
Assigning a common deprivation profile to all household members captures shared conditions but can conceal inequalities within households. Differences in nutrition, schooling, control over resources, disability-related needs, or exposure to violence may not be fully represented.
Data coverage and comparability
Countries’ survey years differ, sometimes substantially. An international release therefore need not describe every country at the same point in time. Conflict, displacement, inaccessible areas, and non-response can create additional coverage problems.
Indicator definitions and survey questions may also change. Trend analysis requires harmonized data, while subgroup comparisons require attention to sampling error and measurement uncertainty.
Breadth rather than depth
Binary indicators record whether a cutoff is crossed, not the magnitude of the shortfall. Two people with very different levels of schooling can receive the same schooling deprivation status.
The index also excludes deprivations among people below the multidimensional poverty cutoff. Improvements or deteriorations within this group may leave the MPI unchanged.
Aggregation and causal interpretation
Different combinations of incidence and intensity can produce the same MPI. Different deprivation profiles can also yield identical scores. Reporting , , indicator contributions, and subgroup results therefore provides information that the headline index alone cannot convey.
Finally, the MPI is descriptive, not a model of causation. A reduction in the index documents a change in measured deprivation; it does not, by itself, establish which intervention caused that change. Comparisons with monetary poverty, income inequality, and other measures of well-being answer related but distinct questions.