Poverty Measurement Design Sabina Alkire, Jose Manuel Roche and - PowerPoint PPT Presentation

Aggregation – Adjusted Headcount Ratio Adjusted Headcount Ratio = M 0 = HA Domains c(k) c(k)/d   0 0 0 0 0   2 / 4   0 1 0 1 2 g 0 ( k )  Persons   4 / 4 1 1 1 1 4     0 0 0 0 0   A = average deprivation share among poor = 3/4 ฀

Aggregation – Adjusted Headcount Ratio Adjusted Headcount Ratio = M 0 = HA = m (g 0 (k)) Domains c(k) c(k)/d   0 0 0 0 0   2 / 4   0 1 0 1 2 g 0 ( k )  Persons   4 / 4 1 1 1 1 4     0 0 0 0 0   A = average deprivation share among poor = 3/4 ฀

Aggregation – Adjusted Headcount Ratio Adjusted Headcount Ratio = M 0 = HA = m (g 0 (k)) = 6/16 = .375 Domains c(k) c(k)/d   0 0 0 0 0   2 / 4   0 1 0 1 2 g 0 ( k )  Persons   4 / 4 1 1 1 1 4     0 0 0 0 0   A = average deprivation share among poor = 3/4 ฀

Aggregation: Adjusted Poverty Gap Adjusted Poverty Gap = M 1 = M 0 G = HAG Domains   0 0 0 0   0 . 42   0 0 1 g 1 ( k )  Persons   0 . 04 0 . 17 0 . 67 1     0 0 0 0   Average gap across all deprived dimensions of the poor: G / ฀

Aggregation: Adjusted Poverty Gap Adjusted Poverty Gap = M 1 = M 0 G = HAG = m (g 1 (k)) Domains   0 0 0 0   0 . 42   0 0 1 g 1 ( k )  Persons   0 . 04 0 . 17 0 . 67 1     0 0 0 0   Obviously, if in a deprived dimension, a poor person becomes even more deprived, then M 1 will rise. ฀ Satisfies monotonicity

Aggregation: Adjusted FGT Consider the matrix of squared gaps Domains   0 0 0 0   0.42 2 1 2 0 0   g 2 ( k )  Persons   0.04 2 0.17 2 0.67 2 1 2    0 0 0 0  ฀

Aggregation: Adjusted FGT Adjusted FGT is M  = m (g  (k)) Domains   0 0 0 0   0.42 2 1 2 0 0   g 2 ( k )  Persons   0.04 2 0.17 2 0.67 2 1 2    0 0 0 0  Satisfies transfer axiom ฀

Aggregation: Adjusted FGT Family Adjusted FGT is M a = m (g a ( t )) for a > 0 Domains   0 0 0 0   0 . 42 a 1 a g a ( k )    0 0 Persons   0 . 04 a 0 . 17 a 0 . 67 a 1 a     0 0 0 0   Theorem 1 For any given weighting vector and cutoffs, the methodology M ka =( ρ k ,M a ) satisfies: decomposability, replication invariance, symmetry, poverty and deprivation ฀ focus, weak and dimensional monotonicity, nontriviality, normalisation, and weak rearrangement for a >0; monotonicity for a >0; and weak transfer for a >1.

Some Applications  International MPI  National Poverty  Mexico, Colombia  Well-being  Bhutan’s GNH  Adaptations  Empowerment  Energy  Governance

Examples

UNDP’s 2010 Human Development Report first published the MPI (updated annually for countries with new data)

The MPI (UNDP 2010) • The MPI 2011 is an international index of acute multidimensional poverty developed at OPHI • In 2011 it covers 109 developing countries. • It was launched in 2010 in the Human Development Report , and updated in 2011, 2012 • It complements the $1.25/day poverty by bringing more dimensions into view • It is the first measure to reflect joint distribution of disadvantages.

Data: Surveys Demographic & Health Surveys (DHS - 54) Multiple Indicator Cluster Surveys (MICS - 32) World Health Survey (WHS – 17) Additionally we used 6 special surveys covering urban Argentina (ENNyS), Brazil (PNDS), Mexico (ENSANUT), Morocco ( ENNVM ), Occupied Palestinian Territory (PAPFAM), and South Africa (NIDS) Constraints: Data are 2000-2010, and not all have all 10 indicators.

MPI Dimensions Weights & Indicators

Measurement: Indicators & Cutoffs • Health – Child Mortality: If any child has died in the family – Malnutrition: If any interviewed adult in the family has low Body Mass Index; if any child is more than 2 standard deviations below the reference normal weight for age, WHO standards) [ WHS has male & female data but no child data; MICS has child data but no adult data; DHS has women 15-49 & child ]

Measurement: Indicators & Cutoffs • Education – Years of Schooling: if no person in the household has completed 5 years of schooling – Child Enrolment: if any school-aged child is out of school, where school-aged is an eight year period from the national starting age.

Measurement: Indicators & Cutoffs • Standard of Living – Electricity (no electricity is deprived) – Drinking water (MDG definitions) – Sanitation (MDG definitions + not being shared) – Flooring (dirt/sand/dung are deprived) – Cooking Fuel (wood/charcoal/dung are deprived) – Assets (deprived if do not own a car/truck and do not own more than one of these: radio, tv, telephone, bike, motorbike, or refrigerator)

Identification: Who is poor? A person is multidimensionally poor if they are deprived in 33% of the dimensions at the same time. 33%

How do you calculate the MPI? • The MPI uses the Alkire Foster method: Formula: MPI = M 0 = H × A • H is the percent of people who are identified as poor, it shows the incidence of multidimensional poverty. • A is the average proportion of weighted deprivations people suffer at the same time. It shows the intensity of people’s poverty – the joint distribution of their deprivations. The MPI is appropriate for ordinal data, and satisfies properties like subgroup consistency, dimensional monotonicity, poverty & deprivation focus. MPI is like the poverty gap measure – but looks at breadth instead – what batters a person at the same time .

What is new? Intensity. The MPI starts with each person, and constructs a deprivation profile for each person. Some people are identified as poor based on their joint deprivations. The others are identified as non-poor. • Most multidimensional poverty measures look at deprivations one by one , not at the household level. • Counting measures do look at coupled deprivations but only provide a headcount , giving no incentive to target those who are deprived in most things at the same time or to reduce intensity.

53 53

54 54

55 55

56

57 57

58 58

59 59

60 60

Phuba Deprived in 67% of dimensions. It doesn’t tell the full story But it gives some idea. 61

MPI – Key Results

Global Results: These results are for 109 developing countries , selected because they have DHS, MICS or WHS data since 2000. Special surveys were used for Argentina, Brazil, Mexico, Morocco, Occupied Palestinian Territory, and South Africa They cover 5.3 billion people - 78.6% of the world’s population Of these 5.3 billion people, 31% of people are poor. That is 1.65 billion people. (2008 population figures taken from Population Prospects 2011 ; 2010 Revision).

Half of the world’s MPI Total Population in 109 MPI countries poor people live in South Asia, and 29% in Sub-Saharan Africa MPI poor people by region

Slovenia Income data ranges from 1992-2008; poverty data within 3 years of MPI. Belarus Kazakhstan 103 of our 109 Countries have Georgia income; only 71 have income Bosnia and Herzegovina The MPI Headcount Ratios and the $1.25/day Poverty Serbia Armenia MPI from 2000-2010. Russian Federation Albania Montenegro Latvia Thailand Uruguay Moldova Macedonia Ukraine Ecuador Uzbekistan Jordan Brazil Argentina Mexico Croatia Hungary Dominican Republic Kyrgyzstan Maldives Azerbaijan Sri Lanka Colombia Syrian Arab Republic Egypt Turkey Estonia Morocco China Paraguay South Africa Philippines Iraq Mongolia Tajikistan Viet Nam Peru Bolivia Indonesia Guatemala Bhutan Nicaragua Djibouti Ghana Honduras Sao Tome and Principe Lesotho Gabon Republic of Congo Swaziland Lao Kenya Pakistan Cambodia Yemen Cameroon India Nigeria Togo Haiti Bangladesh Gambia Cote d'Ivoire Chad Zambia Nepal Tanzania Madagascar Senegal Timor-Leste Benin Malawi Uganda DR Congo Comoros Sierra Leone Mozambique Rwanda Guinea Burkina Faso Liberia Burundi Central African Republic Mali Ethiopia Niger 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Intensity is highest in the poorest countries.

But there is variety… H in High-income countries 1-7%

H in High- and Upper Middle-income countries 1-40%

H in Middle- and High-income Countries 1-77%

H in Low-income Countries ranges from 5-92%

Ghana, Nigeria, and Ethiopia

Ethiopia’s Regional Disparities Ethiopia

Ethiopia’s Regional Disparities Afar Somali Dire Dawa Harari Addis Ababa

Nigeria’s Regional Disparities Nigeria

Nigeria’s Regional Disparities North East Nigeria South West

Ghana’s Regional Disparities Ghana

Ghana’s Regional Disparities Northern Ghana Greater Accra

CHANGES OVER TIME

Ghana, Nigeria, and Ethiopia

Let us Take a Step Back in Time Ethiopia 2000 Nigeria 2003 Ghana 2003

Ethiopia: 2000-2005 (Reduced A more than H) Ethiopia 2000 Ethiopia 2005 Nigeria Nigeria 2003 2008 Ghana 2003 Ghana 2008

Nigeria 2003-2008 (Reduced H more than A) Ethiopia 2000 Ethiopia 2005 Nigeria Nigeria 2003 2008 Ghana 2003 Ghana 2008

Ghana 2003-2008 (Reduced A and H Uniformly) Ethiopia 2000 Ethiopia 2005 Nigeria Nigeria 2003 2008 Ghana 2003 Ghana 2008

Pathways to Poverty Reduction Assets 0 in the Percentage Who is Poor and Deprived in... Cooking Fuel -1 Flooring Annualized Absolute Change Safe Drinking -2 Water Improved Sanitation -3 Electricity Nutrition -4 Child Mortality -5 School Attendance Years of -6 Schooling Ghana Nigeria Ethiopia

0.7 0.6 0.5 0.4 0.3 0.2 Nigeria: Indicator Standard Errors 0.1 0

Ethiopia’s Regional Changes Over Time Harari Addis Ababa

Nigeria’s Regional Changes Over Time North Central South South

Inside the Regions of Nigeria 3.0 2.0 Assets Percentage Who is Poor and Deprived in... 1.0 Cooking Fuel Annualized Absolute Change in the 0.0 Flooring -1.0 Safe Drinking Water -2.0 Improved Sanitation -3.0 Electricity -4.0 Nutrition -5.0 Child Mortality -6.0 Years of -7.0 Schooling School -8.0 Attendance North North North South South South Central East West East South West

Robustness checks to ‘value judgements’ (choice of parameters)

Robustness to poverty cutoff k= 20% to 40% • 90% of the possible pairs of countries have a dominance relation for k 2 to 4. That is, we can say that one country is unambiguously poorer than another regardless of whether we require to be poor in 20, 30 or 40% of the weighted indicators.

Robustness to Weights MPI Weights 1 MPI Weights 2 MPI Weights 3 Equal weights: 50% Education 50% Health 33% each 25% Health 25% Education (Selected 25% LS 25% LS Measure) 50% Education Pearson 0.992 MPI 25% Health Spearman 0.979 Weights 2 25% LS Kendall (Taub) 0.893 50% Health Pearson 0.995 0.984 MPI 25% Education Spearman 0.987 0.954 Weights 3 25% LS Kendall (Taub) 0.918 0.829 50% LS Pearson 0.987 0.965 0.975 MPI 25% Education Spearman 0.985 0.973 0.968 Weights 4 25% Health Kendall (Taub) 0.904 0.863 0.854 Number of countries: 109

Robustness to Weights Summary: • High Correlations: 0.97 and above • High Rank Concordance: 0.90 and above • 85% of all possible pairwise comparisons are robust

Uses of an MPI  Complements monetary poverty measures  Gives a ‘high resolution’ lens on poor people’s lives  An overview and a ‘dashboard’  Changes over time – can change relatively quickly  Provides incentives to reduce intensity and incidence .  Can be used to identify the poorest  Adaptable for National Poverty Measures (or M&E)  Research and Policy: large agenda is ongoing

Value Judgement 1: Selection of Data - existing data - internationally comparable - legitimacy: MDGs - malnutrition rather than income - updated infrequently

Value Judgement 2: Selection of Dimensions Dimensions are ‘notional’ – affect wts, communication Follows precedent: HDI, HPI Possible because of data

Value Judgement 3: Selection of Indicators Data constrained Reflect MDGs (consensus) Relatively Comparable *note: technicalities in construction are hardly normatively justified but matter a lot – e.g. hh

Value Judgement 4: Selection of Weights Weights are ‘nested’ – 1/6 on Health and equal then equal. Education Indicators This is easy to understand, and often 1/18 on standard used. of living indicators. Robust to a range of plausible weights Most controversial.

Identification: Who is poor? A person is multidimensionally poor if they are deprived in 33% of the dimensions at the same time. 33%

Colombia’s National MPI: 5 Dimensions, 15 Variables, Nested Weights 0.2 0.2 0.2 0.2 0.2 Educational Housing & Public Childhood & Youth Work Health Conditions Services Schooling School Absence of Improved Water Coverage Attendance long-term unemploy- ment Access to health Sanitation At the right Illiteracy care given a level Formal work necessity 0.1 Flooring Access to infant 0.1 0.1 services Exterior Walls Overcrowding No Child Labour Poverty cutoff = 0.05 0.04 33%

Mexico’s National Measure: 6 social deprivations (1/12) + income (1/2) Current income per capita Population Wellbeing Income Six Social Rights: • Education Territorial • Health • Social Security • Housing • Basic Services 4 6 5 3 2 1 0 Deprivations • Food Social Rights