Reducing inequalities and poverty: Insights from Multidimensional Measurement Sabina Alkire 16 October 2012, 4 th OECD Forum, New Delhi

Reducing inequalities and poverty:

Insights from Multidimensional Measurement

Sabina Alkire 16 October 2012, 4th OECD Forum, New Delhi

2

Motivation

Measurement: usually income or consumption data.

Trends: reflect trends in nutrition, services, education?

No: direct and lagged relationships are more complex

Hence additional indicators required to study change.

3

Why Multidimensional Measures?Unidimensional measures such as MDGs are essential:

consumption poverty, primary school attendance, malnutrition, immunization, housing, drinking water, etc.

Value-added of multidimensional measures1) joint distribution of deprivations (what one person experiences)

a) focus on poorest of the poorb) address interconnected deprivations efficiently

2) signal trade-offs explicitly: open to scrutiny3) provide an overview plus an associated consistent dashboard

4

Why not?Won’t an ‘overview’ index lose vital detail and information?

Aren’t weights contentious and problematic?

How to contextualise the measure?

5

Why not?Won’t an ‘overview’ index lose vital detail and information?

AF methodology: can be broken down by dimension, group.

Aren’t weights contentious and problematic?

How to contextualise the measure?

6

Why not?Won’t an ‘overview’ index lose vital detail and information?

AF methodology: can be broken down by dimension, group.

Aren’t weights contentious and problematic?Weights are set anyway: budgets, policies, human resources. Sen: the need to set weights is no embarrassmentMeasures should be made robust to a range of plausible weights

How to contextualise the measure?

7

Why not?Won’t an ‘overview’ index lose vital detail and information?

AF methodology: can be broken down by dimension, group.

Aren’t weights contentious and problematic?Weights are set anyway: budgets, policies, human resources. Sen: the need to set weights is no embarrassmentMeasures should be made robust to a range of plausible weights

How to contextualise the measure?The dimensions, cutoffs and weights can be tailor-made.

Multidimensional Poverty Index (MPI)

The MPI implements an Alkire and Foster (2011) M0 measure that can use ordinal data. It was introduced by Alkire and Santos (2010) and UNDP (2010) for 100+ countries

A person is identified as poor in two steps:1) A person is identified as deprived or not in 10 indicators2) A person is identified as poor if their deprivation score >33%

How is MPI Computed?

The MPI uses the Adjusted Headcount Ratio M0:

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.

.

Formula: MPI = H × A

Useful Properties

10

Subgroup Consistency and DecomposabilityEnables the measure to be broken down by regions or social groups.

Dimensional BreakdownMeans that the measure can be immediately broken down into its component indicators. - Essential for policy

Dimensional MonotonicityGives incentives a) to reduce the headcount and

b) the intensity of poverty among the poor.

Changes in the Global MPI

from 2011 MPI Update

Alkire, Roche, Seth 2011

Changes over time in MPI for 10 countries

• MPI fell for all 10 countries

• Survey intervals: 3 to 6 years.

Mul

tidim

ensi

onal

Pov

erty

In

dex

(MPI

)

How and How much? Ghana, Nigeria, and Ethiopia

Let us Take a Step Back in Time

Ghana2003

Nigeria2003

Ethiopia2000

Ethiopia: 2000-2005 (Reduced A more than H)

Ghana2008

Nigeria2008

Ethiopia2005

Ghana2003

Nigeria2003

Ethiopia2000

Nigeria 2003-2008 (Reduced H more than A)

Ghana2008

Nigeria2008

Ethiopia2005

Ghana2003

Nigeria2003

Ethiopia2000

Ghana 2003-2008 (Reduced A and H Uniformly)

Ghana2008

Nigeria2008

Ethiopia2005

Ghana2003

Nigeria2003

Ethiopia2000

Pathways to Poverty Reduction

Ghana Nigeria Ethiopia-6

-5

-4

-3

-2

-1

0

AssetsCooking FuelFlooringSafe Drink-ing WaterImproved SanitationElectricityNutritionChild MortalitySchool Atten-danceYears of Schooling

Annu

aliz

ed A

bsol

ute

Cha

nge

in th

e Pe

rcen

tage

Who

is P

oor

and

Dep

rive

d in

...

Performance of Sub-national Regions

Ethiopia’s Regional Changes Over Time

Addis Ababa

Harari

Nigeria’s Regional Changes Over Time

South South

North Central

Looking Inside the Regions of Nigeria…

North Cen-tral

North East

North West

South East

South South

South West

-8.0

-7.0

-6.0

-5.0

-4.0

-3.0

-2.0

-1.0

0.0

1.0

2.0

3.0

AssetsCooking FuelFlooringSafe Drinking WaterImproved San-itationElectricityNutritionChild Mortal-ityYears of SchoolingSchool At-tendanceAn

nual

ized

Abs

olut

e C

hang

e in

th

e Pe

rcen

tage

Who

is P

oor

and

Dep

rive

d in

...

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Nigeria: Indicator Standard Errors

An Indian ExampleAlmost MPI 1999-2006

Alkire and Seth In Progress

25

India: Almost-MPI over time

We use two rounds of National Family Health Surveys for trend analysisNFHS-2 conducted in 1998-99NFHS-3 conducted in 2005-06

Less information is available in the NFHS-2 dataset; so we have generated two strictly comparable measures, with small changes in mortality, nutrition, and housing.

26

How did MPI decrease for India?

  1999 2006 Change

MPI-I 0.299 0.250 -0.049*

Headcount 56.5% 48.3% -8.2%*

Intensity 52.9% 51.7% -1.2%

27

How did MPI decrease for India?

-12.0%

-10.0%

-8.0%

-6.0%

-4.0%

-2.0%

0.0%

Abso

lute

Cha

nge

in C

H R

atio

Indicator (Statistical Significance) [Initial CH Ratio]

28

Absolute Reduction in Acute Poverty Across Large States

-0.110 -0.090 -0.070 -0.050 -0.030 -0.010

Andhra Pradesh (*) [0.296]Kerala (*) [0.14]Tamil Nadu (*) [0.194]Maharashtra (*) [0.23]Orissa (*) [0.38]Karnataka (*) [0.253]Gujarat (*) [0.246]West Bengal (*) [0.336]Jammu & Kashmir (*) [0.214]Eastern States (*) [0.319]Himachal Pradesh (*) [0.149]Uttar Pradesh (*) [0.344]Rajasthan (*) [0.34]Madhya Pradesh () [0.358]Haryana () [0.187]Punjab (*) [0.114]Bihar () [0.443]

Absolute Change (99-06) in MPI-I

Stat

es (S

igni

fica

nce)

[MPI-

I in

199

9]

We combined Bihar and Jharkhand, Madhya Pradesh and Chhattishgarh, and Uttar Pradesh and Uttarakhand

Significant reduction in all states except

Bihar, MP and Haryana.

29

Change in MPI by casteM0-99 M0-06 Change H-99 H-06 Change A-99 A-06 Change

Scheduled Tribe 0.454 0.411 -0.043 79.7% 73.2% -6.5% 56.9% 56.1% -0.8%Scheduled Caste 0.378 0.308 -0.070 68.7% 58.3% -10.4% 55.0% 52.8% -2.2%OBCs 0.298 0.258 -0.040 57.4% 50.8% -6.5% 52.0% 50.7% -1.2%None Above 0.228 0.163 -0.065 45.0% 32.7% -12.3% 50.7% 49.8% -0.9%

Disparity Increases

MPI Poverty decreased least among the poorest. The STs (8.5% population share) are the poorest, but the change is lowest for them and for OBCs, who have a higher pop share. STs saw almost no reduction of mortality or

undernutrition.

MPI Poverty decreased most for SC and ‘None’.

30

Change in MPI by CasteM0-99 M0-06 Change H-99 H-06 Change A-99 A-06 Change

Scheduled Tribe 0.454 0.411 -0.043 79.7% 73.2% -6.5% 56.9% 56.1% -0.8%Scheduled Caste 0.378 0.308 -0.070 68.7% 58.3% -10.4% 55.0% 52.8% -2.2%OBCs 0.298 0.258 -0.040 57.4% 50.8% -6.5% 52.0% 50.7% -1.2%None Above 0.228 0.163 -0.065 45.0% 32.7% -12.3% 50.7% 49.8% -0.9%

Change in Censored

Headcount Ratio

-16%-13%-10%

-7%-4%-1%2%

Scheduled Tribe Scheduled Caste Other Backward Castes None Above

Least change in Mortality and Nutrition among ST

Deprivation Score

Ultra Poor: Changing Both Deprivation and Poverty Cutoffs

50%

Deprived

33%

No Deprivations

MPI POORMPI z Cutoffs

Ultra z Cutoffs Not Severe

k cutoffs

SeverelyPoorUltra Poor

32

Inequality Among the PoorIndia 1999-2006 Alkire and Seth

Year M0 H (MPI) High Intensity

High Depth

Intense & Deep

1999 0.299 56.5% 30.6% 37.9% 15.8%% of MPI poor 54.2% 67.1% 28.0%

2006 0.250 48.3% 24.7% 31.7% 12.5%% of MPI poor 51.1% 65.6% 25.9%

Change in MPI -.049 -8.2% -5.9% -6.2% -3.3%

33

Multidimensional Poverty Reduction in India, 1999-2006

• Multidimensional poverty declined across India, with an 8% fall in the percentage of poor.

• But disparity among the poor may have increased • Progress has been slowest for STs, for hh with uneducated

head of household, for Bihar MP and Rajasthan, and for Muslims.

• Subgroup decomposable indicators of inequality among the poor may be constructed, and their precise trends tracked.

• We are unable to update these results: new data are unavailable for India since 2005/6.

Why MPI post-2015, & National MPIs?1. Birds-eye view – trends can be unpacked

a. by region, ethnicity, rural/urban, etc

b. by indicator, to show compositionc. by ‘intensity,’ to show inequality

among poor2. New Insights:

a. focuses on the multiply deprived b. shows joint distribution of

deprivation. 3. Incentives to reduce headcount and intensity.4. Flexible: you choose indicators/cutoffs/values5. Robust to wide range of weights and cutoffs

35

Ultra-poverty Deprivation CutoffsSubset of MPI poor that are most deprived in each dimension

Indicator Acute Deprivation Cut-off ‘Ultra’ Cutoff

Nutrition Any adult or child in the household with nutritional information is undernourished (2SD below z score or 18.5 kg/m2 BMI) 3SD or 17 BMI

Child mortality Any child has died in the householdYears of schooling No household member has completed five years of schooling No SchoolingSchool attendance Any school-aged child is not attending school up to class 8Electricity The household has no electricity

Sanitation The household´s sanitation facility is not improved or it is shared with other households

Anything except bush/field

Drinking water The household does not have access to safe drinking water or safe water is more than 30 minutes walk round trip

Unprotected well and 45 Minutes

House The house is kachha, or semi-pucca and owns <1 acre or < 0.5 irrigated kaccha & no land

Cooking fuel The household cooks with dung, wood or charcoal.Wood, grass, Crops, dung

Assets The household does not own more than one of: radio, TV, telephone, bike, motorbike or refrigerator, and does not own a car or truck even one