Global MPI 2021 Frequently Asked Questions
Global MPI – an introduction
What is the global MPI?
The global MPI is an internationally comparable measure of acute multidimensional poverty in over 100 countries. It complements traditional monetary poverty measures, such as the World Bank’s $1.90 a day poverty line, by capturing the acute deprivations in health, education, and living standards that a person experiences simultaneously.
In 2021, the global MPI covers 5.9 billion people living in 109 countries across the developing regions of the world. This represents over 92 percent of the population in lower- and middle-income countries and over three-quarters of the world’s population.
On the OPHI website, the global MPI results are accessible through Excel data tables. These data tables include results at the country level, by age group, rural/urban area, subnational regions, and gender of the household head. As a measure of robustness, we also publish results for multiple poverty-cutoffs. In addition, trends over two or three time periods are published for countries with harmonised data. Finally, auxiliary statistics such as standard errors and sample sizes are part of these data tables.
The global MPI also includes country briefings (that present country-specific results for the countries covered); an interactive databank where users can view the data through different visualisations; and, a Methodological Note which provides the methodology and technical decisions for calculating the global MPI results. This year, through joint research work between OPHI and HDRO, the results by the ethnicity and caste of household head are also included in the global MPI report.
The global MPI is produced annually by OPHI in collaboration with United Nations Development Programme’s Human Development Report Office (UNDP HDRO). The global MPI was originally co-designed and launched in 2010 by both institutions and was jointly revised in 2018 to align insofar as is possible with the Sustainable Development Goals as well as with recommendations of the World Bank’s Atkinson Commission on Monitoring Global Poverty.
How often is the global MPI updated?
The global MPI is updated once a year to include the newly available datasets.
How was the global MPI created?
The global MPI was developed by Alkire and Santos in 2010 in collaboration with the United Nations Development Program’s Human Development Report Office (HDRO) – see Alkire and Santos 2010 and Alkire and Santos 2014. It was first published in HDRO’s 20th anniversary flagship report in 2010 to replace the Human Poverty Index for the Human Development Report of UNDP. In 2018, the first major revision of the global MPI since its inception was undertaken, in order to take into account improvements in survey microdata, and to better align with the 2030 development agenda – see Alkire and Jahan (2018); Alkire, Kanagaratnam and Suppa (2018) and Alkire and Kanagaratnam (2021). The global MPI is a leading practical application of the Alkire–Foster method pioneered by Sabina Alkire and James Foster.
How is the global MPI aligned with the Sustainable Development Goals (SDGs)?
The global MPI is aligned with the SDGs in several important respects:
- SDG 1: The preamble to the 2030 Agenda for Sustainable Development, which defined the SDGs, states that ‘eradicating poverty in all its forms and dimensions… is the greatest global challenge and an indispensable requirement for sustainable development.’ SDG Target 1.2. calls for countries to halve the proportion of men, women, and children living in poverty in all its dimensions according to national definitions. Poverty is understood to be both multidimensional and measurable. National MPIs developed by countries to reflect their particular context and the global MPI both assess progress in poverty reduction: one with respect to national priorities and the other from a comparative perspective.
- Leave no one behind: The 2030 Agenda pledges that ‘no one will be left behind’. The global MPI considers the depth or intensity of an individual’s poverty, going beyond the overall number of poor people (headcount ratio) and providing measurement incentives to reduce the deprivations of the poorest – even if they do not yet exit poverty. Disaggregation of the MPI by age group, rural/urban area, subnational region and female/male headship identifies specific pockets of poverty. The 2021 global MPI report include results that disaggregated by the ethnicity and caste of household head. This enables more targeted policies and actions and helps ensure that particular areas and groups are not left behind. In analysing MPI trends, it is important to note if the poorest groups made the fastest progress – because if they progressed slowly, they are being left behind.
- Interlinkages: Rather than viewing challenges in silos, the MPI shows how deprivations related to the following SDGs are interlinked in the lives of poor people: No Poverty (SDG 1), Zero Hunger (SDG 2), Health & Well-being (SDG 3), Quality Education (SDG 4), Clean Water & Sanitation (SDG 6), Affordable & Clean Energy (SDG 7), Sustainable Cities & Communities (SDG 11). The MPI brings many concerns together into one headline measure. Given that people are MPI poor if they are deprived in one third of the weighted indicators, the MPI focuses on people who are being left behind in multiple SDGs at the same time.
What can the global MPI be used for?
The global MPI measures acute multidimensional data using a comparable methodology for over 100 countries. For global poverty data across developing regions, it is the multidimensional complement of the World Bank’s $1.90-a-day measure (which will be updated to $2.15-a-day later this year).
The global MPI dataset has many potential applications particularly for researchers; INGOs and international agencies; regional, national and local governments; and journalists. These include:
- Seeing where the poorest people live – by country, by rural/urban areas, by subnational regions, by age cohorts, and by gender of the household head – from a multidimensional perspective to cross-check whether programmes are reaching the poorest.
- Using the global MPI to report multidimensional poverty levels for SDG Indicator 1.2.2. in the Global SDG Indicators Database in cases where national governments do not yet have their own tailored national MPI, or where they have officially adopted the global MPI as their national MPI (e.g. Nepal).
- Analysing the composition of poverty by indicator, both at the national level and disaggregated by age, rural/urban, subnational, female/male headship, and occasionally by other variables such as ethnicity or those living with disabilities.
- Comparing absolute and relative trends (and their significance) over time of the MPI, incidence, intensity, and indicators of MPI for countries, and comparing absolute and relative trends for all of these variables disaggregated by subnational regions, rural-urban areas, and age groups.
What does the global MPI measure?
The global MPI is composed of three dimensions (health, education, and living standards) and ten indicators. Each dimension is equally weighted, and each indicator within a dimension is also equally weighted. A person is identified as multidimensionally poor if they are deprived in at least one third of the weighted indicators. The MPI identifies overlapping deprivations for each person. It shows the incidence of poor people in a population and the intensity of deprivations faced within poor households.
What makes a person ‘multidimensionally’ poor?
One deprivation alone may not represent poverty. The MPI requires a person to be deprived in multiple indicators at the same time. A person is multidimensionally poor if they are deprived in at least one third of the weighted indicators. This is termed acute multidimensional poverty.
A feature of the MPI is that by changing the cut-offs or thresholds of the measure we can analyse different poverty levels and disaggregate from there. The global MPI can therefore also tell us about ‘severe’ poverty (those deprived in half or more of the weighted indicators) and people who are not yet poor according to the standard cut-off, but ‘vulnerable’ to poverty (deprived in 20–33% of the weighted indicators).
For examples of how a person is categorised as multidimensionally poor in practice, please see OPHI’s case studies.
The global MPI is described as a measure of acute poverty. How does this differ from extreme poverty?
We have described the MPI as a measure of ‘acute’ poverty to avoid confusion with the World Bank’s measure of ‘extreme’ monetary poverty that captures those living on less than $1.90 a day.
The MPI reflects the acute deprivations that people face at the same time. Because it was designed to compare acute poverty across developing nations, it is most relevant to low- and middle-income countries.
Why is income not included?
Income is not included due to data constraints. Monetary poverty data derive from different surveys, and these surveys often do not have information on health and nutrition. For most countries we are not able to identify whether the same people are monetarily poor and also deprived in any of the MPI indicators. There may be additional technical considerations: to be valid in an MPI, the consumption aggregate or household must be a reliable estimate of that particular household’s monetary poverty over the same time period as the MPI indicators.
Nevertheless, monetary poverty is a fundamental perspective on poverty. We consider that the global MPI and the international measure of extreme monetary poverty – $1.90 / day – complement each other by bringing different aspects into view.
What do your figures for ‘population vulnerable to poverty’ and ‘population in severe poverty’ mean?
Since 2011, two additional categories of multidimensional poverty have been reported in the UNDP’s Human Development Report Tables. These are called the ‘population vulnerable to poverty’ and the ‘population in severe poverty’. The population vulnerable to poverty is defined as the percentage of the population at risk of suffering multiple deprivations – that is, those people with a deprivation score of 20–33 percent of the weighted indicators. The incidence of severe poverty, measures the percentage of the population with a deprivation score of 50 percent or more.
How do I interpret the various values presented with the global MPI results?
The MPI is always reported with a set of linked poverty measures that together make up a powerful information platform. These measures can be unpacked to show the composition of poverty both across countries, regions and the world and within countries by ethnic group, urban/rural location, age group, as well as other key household and community characteristics. This is why OPHI describes the MPI as a high-resolution lens on poverty: it can be used as an analytical tool to identify where and in what form poverty is greatest.
The MPI measures are explained below:
Incidence of poverty: the proportion (%) of the population who are multidimensionally poor (those who are deprived in at least one third of the weighted indicators). This is also sometimes referred to as the ‘headcount ratio’ or the ‘poverty rate’.
Intensity of poverty: the average share (percentage) of deprivations across the ten weighted indicators which people experience simultaneously.
MPI: The MPI ‘value’, which ranges from zero to one, is calculated by multiplying the incidence of poverty by the intensity of poverty. It shows the proportion of deprivations that a country’s poor people experience out of the total possible deprivations that would be experienced if every person in the society were poor and deprived in every indicator. The MPI therefore increases or decreases when either the incidence and/or the intensity of poverty changes.
Number of poor: The number of multidimensionally poor people in a country is important for budgeting and targeting, and reflects both demographic change and population growth. It is computed by multiplying the population of the country by the incidence of MPI.
Where can I find out more about how to apply the MPI approach?
Background materials that provide the technical guidance needed to apply and adapt the MPI approach are available from OPHI’s Online Training Portal. See also ‘How to Apply the Alkire Foster Method’– 12 Steps to a Multidimensional Poverty Measure, and our short online course with UNDP on ‘Designing a Multidimensional Poverty Index’.
For technical professionals and statisticians, our website also advertises our annual OPHI Summer School which provides a two-week technical introduction to the MPI.
OPHI Executive Education provides short courses for policymakers on how to use the MPI in a national context for poverty reduction policies.
How did we calculate the global MPI trends over time?
Trends are estimated using indicators in the global MPI that are adjusted so that harmonised information is used across the time periods. Harmonisation is necessary to ensure that any differences observed are due to changes in the conditions of poverty in the country rather than changes in the questionnaire. A description of the harmonisation principles for each indicator and details of the estimation process are available in MPI Methodological Note 50 and updated policies in 2021 are presented in MPI Methodological Note 51.
To compare the speed of reduction across countries, we refer to ‘annualised changes’, the rate of absolute or relative change divided by the number of years between the survey period. In the case of split survey years (e.g. India 2005/06 and 2015/16), we take the latter years to calculate the annualised changes (in this case between 2006 and 2016).
What is OPHI’s global destitution measure?
The global MPI destitution measure is a destitute measure introduced in 2014 by Alkire, Conconi and Seth. The destitution measure aims to assess the situation of the poorest of the poor within the multidimensional poverty framework.
The destitution measure has precisely the same dimensions, indicators, weights as the global MPI. Only one set of parameters changes: the deprivation cut-offs. Those who are poor according to these deeper deprivation cut-offs are classified as ‘destitute’. In 2020, the measure was revised as described in MPI Methodological Note 49.
Findings and report
What is new in the global MPI 2021 compared to 2020?
The global MPI 2021 covers 109 countries. Of these, 21 have updated surveys: Algeria, Bolivia, Cameroon, Central African Republic, Chad, Cuba, Ethiopia, Ghana, Guinea-Bissau, Guyana, Liberia, Morocco, Nepal, North Macedonia, Palestine, Sao Tome and Principe, Senegal, Serbia, Sierra Leone, Thailand and Turkmenistan. This year two new countries have been added: Costa Rica and Tonga.
For the first time this year, the global MPI is disaggregated by gender of the household head. Out of the 109 countries included in the 2021 global MPI, disaggregation results by female and male headed household were produced for 108 countries – all except China. Information on household head and relationship to head of household were not available in the survey dataset used to produce the MPI estimates for China.
Furthermore, a research collaboration between OPHI and HDRO resulted in the publication of disaggregation results by ethnicity or race for 40 countries and by caste for India in the global MPI report. All countries that have data on ethnicity, race or caste was included in the analysis. In the previous years, specifically for the global MPI 2019, OPHI analysed ethnicity in the 25 countries for which data was available (see OPHI Briefing 55). Smaller case studies were also undertaken in year 2010 (OPHI Briefing 01) and 2014 (OPHI Briefing 21).
In terms of the global MPI trends over time results, the previous round was limited to 80 countries with changes across two time periods. This 2021, we published harmonised estimates for 84 countries; where 56 of the 84 countries have harmonised estimates for two time periods; and 28 countries have harmonised estimates for three time periods. Countries with trends over three time periods include Bolivia, Cameroon, Central African Republic, Chad, Democratic Republic of Congo, Ethiopia, Gambia, Ghana, Guinea, Guyana, Kyrgyzstan, Lesotho, Liberia, Mali, Mongolia, Nepal, North Macedonia, Palestine, Sao Tome and Principe, Senegal, Serbia, Sierra Leone, Suriname, Thailand, Togo, Turkmenistan, Zambia and Zimbabwe.
As a special feature on COVID-19 impacts and risks posed, analysis based on high frequency phone surveys during the pandemic covering 45 countries was presented in the report. Data from these phone surveys were merged with the global MPI results to understand current deprivations experienced during the pandemic.
What are the data sources used in the global MPI 2021?
The global MPI is updated when new data become available from the following sources: Demographic and Health Surveys (DHS), Multiple Indicator Cluster Surveys (MICS), and national surveys. We also explore whether there are new national surveys in the public domain that have indicators comparable to those included in the global MPI. National surveys are considered in the absence of surveys produced by DHS and MICS, or if DHS and MICS data are more than three years older than the national surveys.
The 2021 global MPI covering 109 countries relies on using 45 Demographic and Health Surveys (DHS), 51 Multiple Indicator Cluster Surveys (MICS), 3 Pan Arab Project for Family Health (PAPFAM) surveys and 10 national surveys. The estimates produced for the 109 countries included in the global MPI 2021 are built on the most recent data available for each country. The estimates use the latest survey data available from 2009–2019/2020. For example, the latest available survey for Syria is from 2009; while in Guyana, Liberia and the State of Palestine it is 2019/2020. A total of 79 countries – home to 84 percent of multidimensionally poor people – have data fielded in 2015 or later, and 22 of those countries have data fielded in 2019 or later.
The 84 global MPI countries with trends over time have data that span from 2000–2019/2020. The trends estimated for 32 countries are solely based on data from DHS across all time periods. For 23 countries, the trends were estimated solely from MICS across all time periods. For four countries (China, Jamaica and Mexico), trend data is based on national data, while for Morocco we have used PAPFAM across the years. For 25 countries, namely Afghanistan, Bangladesh, Benin, Burkina Faso, Cameroon, Chad, Congo, Côte d’Ivoire, Democratic Republic of Congo, Dominican Republic, Gambia, Ghana, Guinea, Guyana, Lesotho, Madagascar, Mali, Moldova, Nepal, Sao Tome and Principe, Sierra Leone, Togo, Ukraine, Yemen and Zimbabwe we have used a mix of DHS and MICS across the time periods.
All population aggregates (numbers of people living in multidimensional poverty) which are presented in the data tables and the global MPI report use 2019 population data from the World Population Prospects (UNDESA, 2019), unless otherwise indicated. Data tables available online also provide the results using population data for the year of the survey.
Why were these 109 countries selected for the 2021 global MPI?
To estimate trends over time, countries had to have at least two datasets with comparable sampling frames, which were fielded with a minimum period of three years between surveys. Countries with no suitable datasets, or with data for only one time period were not included. For more detail on the countries, data and time periods selected, see the MPI Methodological Note 50 and MPI Methodological Note 51.
Is it possible to compare the global MPI 2021 to previous years?
We would caution against using the published numbers of the global MPI for strict comparisons. All global MPI estimates published before 2017 refer to the ‘original’ MPI structure, which was updated in 2018 with additional minor revisions in 2019 and 2020. Comparisons of the global MPI estimates should account for these changes. However, the multidimensional poverty trends results, as published in OPHI’s Table 6, harmonises the indicators precisely to allow for comparability over time.
Why does the report feature 80 countries for trends over time while the table on the OPHI website has 84 countries?
84 of the 109 countries included in the global MPI have results on trends over time. Table 6 on OPHI website (under Data Tables 2021) covers all 84 countries. The 2021 global MPI report published in partnership with UNDP’s Human Development Report Office, profiled 80 countries. The countries excluded from the joint publication but included in OPHI table are Afghanistan, Trinidad and Tobago, Viet Nam, and Yemen. These countries are excluded from the joint publication because of data constraints in health indicators over the time periods. The complexity of the harmonised estimates of these four countries is detailed in MPI Methodological Note 50.
Income vs. MPI
Why are there wide discrepancies between MPI poverty estimates and $1.90/day poverty estimates in many countries?
The fact that there are discrepancies between MPI poverty estimates and $1.90/ day poverty estimates is one of the reasons why we compute the global MPI. The 2021 report shows how in 43 of the 60 countries with both multidimensional and monetary poverty estimates, the incidence of multidimensional poverty was higher than the incidence of monetary poverty. Although there is a clear overall relationship between MPI and $1.90/day poverty line, particularly in low poverty countries, the estimates differ for many countries. The MPI therefore complements monetary poverty measures, revealing poverty that would otherwise be overlooked. It measures various deprivations directly.
The mismatch between income and other deprivations is well-documented, including in Europe. Possible explanations include infrastructure; public services such as health, education, water, power, and transportation; market access, spending habits, household size and composition; pro-poorest or discriminatory local institutions; the presence of a large employment industry (e.g. a mine); remoteness, and so on.
Why is the global MPI headcount ratio much higher than national monetary poverty estimates in some countries?
The MPI, like the $1.90/day poverty line, is a globally comparable measure of poverty. It measures acute multidimensional poverty, and only includes indicators that are available for many countries. National poverty measures are typically monetary measures, and thus capture something different. The fact that there are differences does not mean that the national monetary poverty number, or the MPI headcount ratio is wrong – these simply measure different conceptions of poverty.
At the same time, just as national poverty measures are designed to reflect the national situation more accurately and often differ from the $1.90 a day measure, some countries may wish to build a national Multidimensional Poverty Index that is tailored to their context, to complement the global MPI (see the MPPN website for more details).
Is the global MPI intended to replace the standard $1.90 a day measure of poverty used for the SDGs and other international purposes?
No, the global MPI is intended to complement monetary measures of poverty, including $1.90 a day estimates. The relationship between these measures, as well as their policy implications and methodological improvement, are priorities for further research.
Policies and international adoption
How can I find out more about a country’s multidimensional poverty?
Country briefings for the 109 countries included in the global MPI 2021 are available which explain the global MPI results for each country.
What are the policy implications of the global MPI results?
The MPI shows where and how people are poor and trends show the rate of reduction or increase in multidimensional poverty. It helps policymakers analyse the interconnections among deprivations and identify priority areas for interventions, enabling policymakers to target resources and design policies more effectively. The global MPI estimates act as a springboard for countries, who go on to develop national MPIs using the same methodology, which are tailored to the national context to identify key policy priorities. Some countries, such as Nepal, use the global MPI directly as their national MPI. For more information on uses please see the OPHI-UNDP book ‘How to Build a National Multidimensional Poverty Index (MPI): Using the MPI to inform the SDGs’.
Can the MPI be adopted for national poverty eradication programmes?
The global MPI was devised as an analytical tool to compare acute poverty across nations. The MPI indicators and dimensions can be modified to generate national Multidimensional Poverty Indices that reflect local cultural, economic, political, climatic and other factors. An MPI will immediately reflect changes in deprivations in any of its indicators, such as school attendance, so can be used to monitor progress.
Countries select data sources and the most relevant indicators and weights that make sense in their context to create tailored national poverty measures. Many governments including Afghanistan, Bhutan, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Honduras, Nepal, Seychelles, Sierra Leone, Pakistan, Thailand and Viet Nam have released official Multidimensional Poverty Indices – see the National Measures page on the MPPN website. Colombia is a powerful example of how the MPI can be used to coordinate national poverty eradication programmes. Costa Rica shows how it is used for budgeting and Bhutan shows how it can be used for targeting.
How does the MPI respond to the effects of shocks?
The survey data used to estimate the global measure are usually collected every three to five years. The MPI will reflect the impacts of shocks as far as these, for example, cause children to leave primary education, or to become malnourished, or have worsened housing conditions and services and so on in the next survey. For example, if a flood occurred between two survey periods and many people lost their homes and are still living in substandard housing, this will be captured by the global MPI in the next survey. If more frequent data are available at the country or local level, this can be used to capture the effects of larger scale economic and other shocks. However, there are other ways the MPI is being used to provide information on shocks, particularly in light of the COVID-19 pandemic.
First, if certain deprivations in the global MPI signal high vulnerability (for example, people who are undernourished, lack clean water, and are at risk of acute respiratory infections due to solid cooking fuel), then additional elementary analysis can identify which people have one, two, or all three of these deprivations (see Policy Briefing 53a and Policy Briefing 54a).
Second, there may be additional indicators in the survey that can be linked to the risk, so disaggregating by this information (data permitting) or adding it in as an additional indicator or dimension to the MPI will clarify risks from the shock. For example, in the case of COVID-19, one might incorporate a dimension for each household that also includes data on handwashing, overcrowding, or the presence of older person(s) in the household, or underlying health conditions such as diabetes and heart conditions, or domestic violence.
Third, it is possible to explore different scenarios by randomly assigning additional deprivations to some identified vulnerable groups and using these microsimulations to re-assess the MPI. For example, if undernutrition among poor and vulnerable people rose by 25%, what would be the effect on multidimensional poverty? This was explored in the 2020 global MPI report.
The global MPI covers more than 100 developing countries. Will an MPI be created for developed countries?
In the light of the COVID-19 pandemic, an MPI for developed countries may become more urgent. The constraint is not the methodology, which can be easily extended to reflect different thresholds and aspects of multidimensional poverty, but rather the data.
There are no publicly available equivalently comparable data across high-income countries. The list of the 109 countries that the global MPI 2021 covers and country-specific summaries are available on the Country briefings page and through the MPI data tables page.
For examples of discussions of MPI in developed countries see the following discussions for Germany and the United States.
What are the main limitations of the global MPI?
The MPI has some drawbacks, mainly due to data constraints. The indicators include both outputs (such as years of schooling) and inputs (such as cooking fuel), data on outcomes like people’s functionings and capabilities are not available for all dimensions.
There are many aspects of health that the two included indicators do not cover. Furthermore, the nutrition indicator is the least comparable of all 10 indicators, because in countries with MICS surveys, data are only available for children under 5 years of age, whereas in DHS women’s nutritional data are usually available and sometimes some data for men. National surveys may cover a larger age range. Despite these limitations, the patterns that emerge are plausible and familiar.
In some cases, careful judgments are needed to address missing data. To be considered multidimensionally poor, households must be deprived in at least six living standard indicators or in three living standard indicators and one health or education indicator. This requirement makes the MPI less sensitive to minor inaccuracies.
The estimates presented here are based on publicly available data and cover various years between 2009 and 2020, which limits direct cross-country comparability.
Why are these indicators used in the global MPI? Why not indicators for environment, employment, or empowerment for instance?
To create the global MPI an initial list of dimensions and indicators was prepared following a long process of consultation. This list was then compared against available data for most countries included in the study, and we found that there were some data constraints. To build a meaningful index we are limited to data readily available in most surveys.
Environment: The Human Development Report in 2020 released research probing the interrelationship between climate change, the environment and poverty and continues to explore the technical challenges that this raises.
Employment: The data sources used, unfortunately, are not able to identify unemployed or underemployed persons, nor unsafe work or informal work. Given its importance, many national MPIs do add this dimension. In this year’s report, we overlay the MPI with separate employment data collected through high-frequency phone surveys during the pandemic.
Empowerment: The DHS surveys collect data on women’s empowerment for some countries, but not every DHS survey includes empowerment, and the other surveys do not have these data. Data on men’s empowerment or political freedom are missing.
Why does national data for the MPI date from so many different years? Is it unfair to compare countries if the statistics in one case are five years older than in another?
The 2021 MPI relies on the most recent and reliable data available since 2009. As in the case of all poverty measures including income and social statistics, surveys are taken in different years, and some countries do not have recent data. In order to facilitate clear analysis, the year of the survey is reported in the MPI tables. The difference in dates limits direct cross-country comparisons, as circumstances may have improved, or deteriorated, in the intervening years. Naturally, this is a stimulus for country governments to collect up-to-date surveys that reflect recent progress. The SDGs’ focus on data should, we hope, give rise to more frequent data for MPI estimations.
What are the main limitations to measuring MPI trends over time?
Analysis of trends in multidimensional poverty rely on harmonised versions of the global MPI. As such, all the limitations of the global MPI, also apply to our work on trends over time.
What are the main technical limitations to analysing ethnicity, race and caste as reported by household head?
The ethnicity/race/caste variable was constructed using data from Multiple Indicator Cluster Surveys (MICS, 23 countries), Demographic and Health Surveys (DHS, 14) and national household surveys (4). The operationalization of ethnicity, race and caste applied here is constrained by data. Available data refer to self-identification with a group. The number of reported groups varies widely across countries, and intragroup ethnic inequalities might be obscured by survey groupings. Most questions asked about ethnic group or tribe, but surveys in some countries focused on racial categories (Cuba), caste (India) or a combination of ethnic group and native language (Paraguay). Because of these differences, comparisons across countries should be made with caution.
What are the main technical limitations to analysing MPI by gender of the household head?
Across all surveys, gender is a binary variable (male or female), and household head is a self-reported category. Household members typically acknowledge the household head on the basis of age (older), gender (male) or economic status (main provider). The analysis provides a global account of multidimensional poverty by headship but is constrained by the mixed definition of headship used in the surveys.
What are the main technical limitations to using high frequency phone surveys for COVID-19 impact analysis?
The representativeness of the high-frequency phone surveys varies, and all samples are drawn exclusively from the subpopulation that owns a phone and are thus not representative of individuals without phones—that is, the samples are not nationally representative. Sampling frames were based on existing, representative and face-to-face household surveys from which respondent phone numbers were available; on lists of phone numbers from telecom providers; or on lists of randomly generated numbers (based on so-called random digit dialling). The statistics thus need to be interpreted with caution and should not be considered representative for country-level analyses or cross-country comparisons. Selection-coverage and selection-nonresponse biases apply. The estimates are expected to be somewhat conservative. Phone owners who were sampled in all cases are, on average, better off than the average respondent in a face-to-face survey on several characteristics. Actual deprivations might thus exceed the ones presented.