Motivation
description
Transcript of Motivation
Analysis of Inequality across Multi-dimensionally Poor and Population
Subgroups for Counting Approaches
Suman Seth and Sabina Alkire
Development Studies Association Annual Conference 2013University of Birmingham
November 16, 2013
Motivation
Poverty measurement tools may affect policy design and policy incentive – Incidence
– Intensity
– Inequality
Three I’s of poverty measurement (Jenkins and Lambert 1997)
2
Concerns for Inequality
Inequality among the poor– Consideration of inequality in poverty measurement has
been prominent since Sen (1976)
Disparity across population subgroups– Horizontal Inequality (Stewart 2000)
3
Counting Approach
Counting measures– Townsend (1979), Atkinson (2003)
– Adjusted headcount ratio (Alkire and Foster): • several national and international (MPI, WEAI) adaptations
Capturing inequality is natural for cardinal dimensions– Can reflect inequality within each dimension
Not so straightforward for ordinal or binary dimensions– But can be captured across deprivation counts
4
Example: Concern for Inequality
Deprivation counts
(0, 2, 4, 6, 8, 10)
(0, 2, 2, 4, 8, 10) (0, 2, 2, 5, 8, 9)
Next period I Next period II
Similar reduction in incidence and intensity, but these two situations are not the same
5
6
Consideration for Inequality
Other Approach: Use a separate inequality measure?
An example: Use of standard deviation in child poverty– Delamonica and Minujin (2007), Roche (2013)
Advantage:– Additional information besides incidence and intensity
– If decomposable, can observe inequality decomposition within and between group
– Can be used with intuitive poverty measure such as Adjusted
Headcount Ratio
7
Consideration for Inequality
Q: Which inequality measure to use?– Eepends on which properties we want the measure to satisfy
Three key properties– Absolute inequality: if every poor’s deprivation count rises
or decreases by same number, inequality should not change
– Additive Decomposability: within-group + between group
– Within-group Mean Independence: Total within-group does not change if no change in inequality within any subgroup
8
Consideration for Inequality
The Inequality Measure?
The only absolute inequality measure that satisfies these properties is a positive multiple of variance
V(x) = i(xi – (x))2/n
where, V(x): positive multiple of variance of distribution x
(x): mean of distribution x
n: population size of distribution x
> 0
Chakravarty (2001), Bosmans and Cowell (2011),Chakravarty and Tyagarupananda (1998)
9
Illustrations using the Global Multidimensional Poverty Index (MPI)
10
Example on India: Castes (DHS 2 & 3)
11
1999Intensity
(MPI)Share of Poor
Inequality (Poor)
caste A_caste poor_shr var_depr_caste_p
Scheduled Tribe 57.0% 12.6% 0.110 Scheduled Caste 55.0% 22.1% 0.107 Other Backward Classes
52.1% 33.3% 0.095
General 50.6% 32.0% 0.089 India 52.9% 100% 0.1002006ST 56.3% 12.9% 0.115 SC 52.6% 22.9% 0.098 OBC 50.8% 42.1% 0.090 General 49.7% 22.0% 0.092 India 51.7% 100% 0.097
Example on India: Castes (DHS 2 & 3)
12
1999Intensity
(MPI)Share of Poor
Inequality (Poor)
caste A_caste poor_shr var_depr_caste_p
Scheduled Tribe 57.0% 12.6% 0.110 Scheduled Caste 55.0% 22.1% 0.107 Other Backward Classes
52.1% 33.3% 0.095
General 50.6% 32.0% 0.089 India 52.9% 100% 0.1002006ST 56.3% 12.9% 0.115 SC 52.6% 22.9% 0.098 OBC 50.8% 42.1% 0.090 General 49.7% 22.0% 0.092 India 51.7% 100% 0.097
Inequality among the poor fell for SC and OBC, but not for ST
Indian States (DHS 2 & 3)
13
Arunachal Pradesh
Bihar
RajasthanTripuraAssamOrissa
NagalandHimachal Pradesh
Uttar Pradesh
HaryanaMeghalaya
Madhya PradeshNew DelhiKerala West Bengal
GujaratGoaMaharashtra
Jammu Sikkim
Manipur PunjabTamil Nadu
Karnataka
MizoramAndhra Pradesh
-0.040
-0.030
-0.020
-0.010
0.000
0.010
0.020
-0.120 -0.100 -0.080 -0.060 -0.040 -0.020 0.000 0.020 0.040
Ch
ange
in In
equ
alit
y in
Dep
riva
tion
Sco
res a
mon
g th
e P
oor
Change in MPI
Cross Country Comparison
Two countries: different MPIs but similarly unequal
14
Country YearHeadcount
Ratio MPI
Average Deprivation
Count (Poor)Inequality
(Poor)
Colombia 2010 5.4% 0.022 40.9% 0.041
Lesotho 2009 35.3% 0.156 44.1% 0.042
Demographic Health Surveys
Disparity in Intensity vs. Disparity in Poverty
Between group inequality among poor is not sufficient for disparity in poverty between groups– Sub-national Disparity (Alkire, Roche, Seth 2011)
Example:
c = (0,0,0,6,6,6,6,6,7,7), cA = (0,0,6,6,7) and cB = (0,6,6,6,7)
c' = (0,0,0,6,6,6,6,6,6,6), c'A = (0,0,0,6,6) and c'B = (6,6,6,6,6)
Overall inequality, within group inequalities, between group inequalities among the poor – all lower in c' than in c
Disparity in poverty between subgroups?
15
Cross Country Comparisons
Similar inequality among the poor but very different sub-national disparity
16
Country Year MPIInequality
(Poor)
Total Within-Group
Between Group
Between MPI
Bolivia 2008 0.089 0.044 0.042 0.002 0.006Zimbabwe 2011 0.172 0.045 0.044 0.001 0.021
Demographic Health Surveys
Concluding Remarks
We discuss how inequality can be captured among the poor in counting approach through a positive multiple of variance
A separate measure can provide more information besides a poverty measure than just ranking
Although debated (Kanbur, 2006), change in inequality decomposition can provide important information
17
Concluding Remarks
It is also important to capture disparity between subgroup’s poverty
Same result if counted in achievement or deprivation space
Future research– Compute the standard error to understand statistical
significance of comparisons
– Deeper analysis across countries to understand inequality decompositions and causes of change in inequality
18