1. Multidimensional poverty in India: Has growth been
pro-poor on multiple dimensions?
Uppal Anupama
Punjabi University
Discussant: Flaviana Palmisano
University of Luxembourg
33th IARIW Conference
Rotterdam, the Netherlands, August 24-30, 2014
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 1 / 22
2. Motivations
Interest in growth and its distributional implications has been growing
steadily in both the academic and policy debate.
The pro-poor growth literature has been developed to assess the impact of
growth on poverty. Among the main contributions: Ravallion and Chen
(2003), Son (2004), Bourguingon (2004), Duclos (2009), Kakwani and
Pernia (2000).
They are income-based (some exceptions are Grosse et al. (2008) and
Klasen (2008)):
I insensitive to any change of non-income poverty.
I income pro-poor growth does not automatically mean that non-income
poverty has also been reduced.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 2 / 22
3. The role of multidimensional poverty
Interest in multidimensional poverty measurement has also been growing
steadily in both the academic and policy debate.
Broad acknowledgment that poverty is about more than just low incomes.
Poor people themselves often allude to non-income dimensions as crucial to
their perception of poverty.
Within the academic discussion, a number of approaches proposed to
analyze it, including, among others, Bourguignon and Chakravarty (2003),
Tsui (2002), Alkire and Foster (2011), Chakravarty, Deutsch and Silber
(2008), Duclos, Sahn and Younger (2006).
Dominant issue also within the broader policy debate. For instance, the
UNDPs Human Development Report 2010 gave prominence to the
Multidimensional Poverty Index (MPI) of Alkire and Santos (2010), which
was reported for over 100 countries.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 3 / 22
4. Research question and aims
How is it possible to evaluate pro-poor growth accounting for multiple
dimensions of deprivation?
'Hence, there is a need to have a rational synergy between the pro-poor
growth indicators and multi-dimensional poverty indicators.' (page 4)
Provide a methodology to evaluate pro-poor growth in multiple (cardinal
and ordinal) dimensions.
Empirical investigation of the dierence between income based pro-poor
growth and pro-poor growth based on multiple non-income indicators, using
Indian data.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 4 / 22
5. Data
The analysis is based on the 61th (2004-05) and 66th (2009-10) waves of
the NSSO dataset on consumption expenditure for measuring income as well
as non-income poverty.
8 dimensions of deprivations are considered: expenditure, education,
lighting, dwelling unit, ownership of land, regular salary income, cooking
fuel, number of meals per day. How do you choose them?
The poverty line of these dimensions has been
6. xed according to the MDG
indicators.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 5 / 22
7. Measuring multidimensional poverty
The Alkire and Foster (2008) framework (it is Alkire and Foster (2011)) is
used to measure multidimensional poverty:
I Adjusted headcount ratio: M0 = HA (H: number of the poor identi
8. ed
using the dual cuto approach; A: fraction of possible dimensions in
which the average poor person is deprived)
I Adjusted poverty gap: M1 = M0G (G: average poverty gap across all
instances in which poor persons are deprived)
I Adjusted FGT measure: M2 = M0S (S: average severity of
deprivations, across all instances in which poor persons are deprived)
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 6 / 22
9. Measuring pro-poor growth
Extend the GIC framework (Ravallion and Chen 2003) to non-income
indicators
I Rate of pro-poor growth (RPPG, Ravallion and Chen 2003):
RPPG =
Z H
0
g(p)dp/H (1)
g(p) = yt+1(p)yt (p)
yt (p) (coordinate of the GIC)
I Rank individuals according to each non-income indicators (8 rankings);
I Calculate the population centiles based upon this ranking;
I Calculate the GIC and RPPG for each dimension.
This type of exercise gives indication on how growth behaves for each
dimension which may further speci
10. es the direction of public spending for
any poverty removal strategy.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 7 / 22
11. Issues
Ranking based on dierent scales of attainments: shifting of one rank in the
lower orders may not mean the same thing as shifting of one rank in higher
orders (e.g. for education, the shift from below primary to primary may not
improve the living standard of a person as compared to the shift from
graduation to post-graduation) ) solution: assign higher weights to higher
order of education.
I How do you construct this weighting scheme? What is the rational
behind them? It could also be the opposite: think at the decreasing
marginal utility of income.
I Probably you could think about constructing a weighting function
which depends on the return to education for each level, giving
constant marginal utility to income
I The problem is more relevant when measuring education with years, for
instance the increase from 1 year to 2 years of education may mean
little if that person remains illiterate. An increase from 5 years to
completed primary (6 years) education might be much more valuable.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 8 / 22
12. Issues (2)
Some variables of non-income indicators do not vary much i.e. the variables
are bounded. Hence the GIC may result to be
at ) solution: use
conditional GIC in which the population is ranked by income indicator.
I These are two dierent aspects: 'in-built inertia' (Klasen, 2008) and
variables bounded above
I Using conditional GIC can lessen - but not solve - this problem when
the variable is bounded above
I As for variables that do not vary much, such as education after a
certain age, you could opt for evaluating pro-poorness considering
speci
13. c cohorts.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 9 / 22
14. Measuring pro-poor growth (2)
Kakwani and Pernia (2000):
KP = h/hg (2)
h actual growth elasticity of poverty, hg growth elasticity of poverty in the
counterfactual scenario with pure growth and no change in relative
inequality.
Poverty Equivalent Growth Rate (PEGR, Kakwani and Son 2008 (not
Kakwani and Pernia 2001!):
PEGR = KP g (3)
g is the mean income growth rate.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 10 / 22
15. Measuring pro-poor growth (2)
Introduce these formalizations and explain them in the text
These measures present the same drawbacks of the RPPG - as most of the
other measures - this is because poverty and inequality variations can be
expressed as dierent ways of aggregating g(p). In fact:
KP =
1
g
Z H
0
dP
dy
y (p)g(p)dp/
Z H
0
dP
dy
y (p)dp (4)
PEGR =
Z H
0
dP
dy
y (p)g(p)dp/
Z H
0
dP
dy
y (p)dp (5)
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 11 / 22
16. Issues (3)
Do we really need a composite index?... In case of developing economies, for
dierent dimensions, we have to depend upon dierent data sets. This poses
a problem as we would be dealing with dierent reference units... For
targeting policy, the separate calculations of these indicators across
dimensions are more important.
I The paper then analyzes pro-poorness in each single non-income
indicator
I Conceptual con
ict with the aim of the paper stated in the
introduction
I You are not looking anymore for a 'synergy' between pro-poor growth
and multidimensional poverty
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 12 / 22
17. Unidimensional poverty rates
Poverty has declined in all the dimensions (except regular salary income); this decline is the
highest for education in rural areas and dwelling unit in urban areas.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 13 / 22
18. Multidimensional poverty rates
Number of
Dimensions
Percentage of Population
2004-05 2009-10
Rural Areas Urban Areas Rural Areas Urban Areas
1 98.9 89.5 97.9 82.3
2 94.8 64.7 89.7 48.4
3 83.8 37.0 65.5 23.8
4 52.4 16.9 31.9 8.9
5 17.1 5.2 8.3 2.3
6 1.3 0.8 0.6 0.2
7 0.1 0.2 0.00 0.00
8 0.00 0.00 0.00 0.00
As we increase the number of dimensions that are needed in order to be classified as poor, the
head count ratio falls.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 14 / 22
19. Unidimensional poverty gap and severity of poverty
Dimensions
Rural Urban
2004-05 2009-10 2004-05 2009-10
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Uni-dimensional
Expenditure 0.075 0.030 0.044 0.017 0.050 0.021 0.032 0.013
Number of
Meals Per
Day
0.019 0.019 0.013 0.013 0.015 0.015 0.013 0.013
Education 0.629 0.524 0.418 0.387 0.424 0.318 0.244 0.221
Dwelling 0.007 0.002 0.007 0.003 0.014 0.005 0.001 0.001
Ownership
of Land
0.021 0.010 0.016 0.008 0.122 0.058 0.114 0.054
Regular
Salary
Income
0.419 0.198 0.427 0.202 0.271 0.128 0.284 0.134
Cooking
Fuel
0.609 0.423 0.599 0.420 0.208 0.143 0.180 0.128
Lighting 0.228 0.114 0.179 0.090 0.040 0.020 0.031 0.016
Poverty has declined for most of the dimensions, with the exception of regular salary income in
both areas and dwelling unit in rural areas.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 15 / 22
20. Multidimensional poverty gap and severity of poverty
Dimensions
Rural Urban
2004-05 2009-10 2004-05 2009-10
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Poverty
Gap
Severity
of
Poverty
Multi-dimensional
1 0.576 0.378 0.580 0.387 0.534 0.328 0.541 0.348
2 0.575 0.378 0.580 0.389 0.536 0.338 0.552 0.370
3 0.576 0.380 0.579 0.394 0.545 0.347 0.558 0.385
4 0.568 0.372 0.567 0.386 0.543 0.348 0.563 0.375
5 0.550 0.358 0.547 0.358 0.529 0.353 0.533 0.333
6 0.600 0.400 0.750 0.500 0.429 0.286 0.500 0.500
Poverty gap and severity of poverty have increased over time.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 16 / 22
21. Unidimensional vs multidimensional poverty change
Unidimensional poverty decreases over time according to all three indices.
Multidimensional poverty increases over time according to poverty gap and
severity of poverty.
Thus, contrasting results. This poses the question- has the growth been
pro-poor on multiple dimensions?
I Contrasting results which seems to justify a pro-poor growth analysis in
multidimensional framework. This is somehow dierent from what is
done in the following step of this analysis
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 17 / 22
22. Pro-poor growth in multiple dimensions of poverty
Dimensions Average
Growth
Rate (g)
Ravallion
and Chen
Index
Ravallion
and Chen
Index - g
Poverty Gap Severity of Poverty
Kakwani
and
Pernia
PEGR PEGR-g
Kakwani
and
Pernia
PEGR PEGR-g
Rural
Expenditure 0.217 0.157 -0.059 0.789 0.171 -0.046 0.702 0.152 -0.065
Number of
Meals Per
Day
0.006 -0.443 -0.449 51.54 0.325 0.319 26.176 0.165 0.158
Education 1.065 0.423 -0.642 0.517 0.550 -0.515 0.328 0.349 -0.716
Dwelling 0.001 -0.002 -0.003 17.56 0.021 0.020 -18.010 -0.021 -0.022
Ownership
of Land
0.006 0.155 0.150 21.095 0.115 0.110 10.61 0.058 0.052
Regular
Salary
Income
-0.016 -0.013 0.003 0.589 -0.010 0.007 0.289 -0.005 0.012
Cooking
Fuel
0.061 0.015 -0.046 0.222 0.014 -0.048 0.066 0.004 -0.057
Lighting 0.077 0.148 0.071 1.518 0.117 0.040 0.802 0.062 -0.015
Urban
Expenditure 0.303 0.147 -0.156 0.571 0.173 -0.130 0.529 0.160 -0.143
Number of
Meals Per
Day
0.001 0.209 0.208 130.04 0.118 0.117 64.79 0.059 0.058
Education 0.793 0.439 -0.354 0.629 0.499 -0.295 0.402 0.319 -0.474
Dwelling 0.012 0.358 0.346 26.558 0.307 0.296 12.207 0.141 0.130
Ownership
of Land
0.010 0.045 0.035 3.230 0.034 0.023 1.633 0.017 0.007
Regular
Salary
Income
-0.019 -0.031 -0.012 1.162 -0.022 -0.003 0.569 -0.011 0.008
Cooking
Fuel
0.061 0.239 0.179 1.542 0.094 0.033 0.701 0.043 -0.018
Lighting 0.010 0.130 0.119 10.281 0.106 0.096 4.437 0.046 0.036
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 18 / 22
23. Pro-poor growth in multiple dimensions of poverty by
social groups
Scheduled Tribes Scheduled Castes Other Backward Classes Others
Dimensions Average
Growth
Rate (g)
PPGR PPGR-g Average
Growth
Rate (g)
PPGR PPGR-g
Average
Growth
Rate (g)
PPGR PPGR-g
Average
Growth
Rate (g)
PPGR PPGR-g
Rural
Expenditure 0.233 0.227 -0.007 0.194 0.122 -0.072 0.188 0.153 -0.035 0.295 0.170 -0.125
Number of
Meals Per
Day
0.004 -0.685 -0.689 0.002 0.035 0.033 0.006 -0.335 -0.340 0.011 -0.955 -0.966
Education 1.324 0.415 -0.910 1.133 0.418 -0.715 1.095 0.423 -0.672 0.974 0.433 -0.541
Dwelling 0.001 0.083 0.081 0.001 -0.049 -0.050 0.001 -0.047 -0.049 0.001 0.059 0.058
Ownership
0.003 0.080 0.077 0.008 0.226 0.218 0.006 0.181 0.175 0.003 0.091 0.087
of Land
Regular
Salary
Income
-0.016 -0.012 0.004 -0.016 -0.013 0.004 -0.011 -0.009 0.002 -0.022 -0.019 0.003
Cooking
Fuel
0.088 0.027 -0.061 0.039 -0.004 -0.043 0.066 0.013 -0.053 0.082 0.045 -0.038
Lighting 0.145 0.207 0.061 0.093 0.144 0.052 0.073 0.143 0.070 0.055 0.143 0.088
Growth is never (relative) pro-poor in the dimension of expenditure, cooking fuel and education even though the average rate
of growth of this particular dimension is the highest among all the dimensions for all social groups. It is always (relative) pro-poor
for ownership of land and lighting.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 19 / 22
24. Pro-poor growth in multiple dimensions of poverty by
social groups (2)
Scheduled Tribes Scheduled Castes Other Backward Classes Others
Dimensions Average
Growth
Rate (g)
PPGR PPGR-g Average
Growth
Rate (g)
PPGR PPGR-g
Average
Growth
Rate (g)
PPGR PPGR-g
Average
Growth
Rate (g)
PPGR PPGR-g
Urban
Expenditure 0.746 0.132 -0.614 0.302 0.134 -0.168 0.354 0.172 -0.182 0.279 0.140 -0.140
Number of
Meals Per
Day
-0.006 0.541 0.547 -0.001 -0.305 -0.305 0.001 0.281 0.280 -0.003 0.192 0.195
Education 1.088 0.441 -0.647 1.011 0.438 -0.573 0.896 0.437 -0.459 0.703 0.442 -0.261
Dwelling 0.025 0.297 0.271 0.013 0.362 0.349 0.010 0.394 0.383 0.011 0.333 0.322
Ownership
0.005 0.019 0.014 0.011 0.045 0.035 0.006 0.026 0.020 0.015 0.068 0.052
of Land
Regular
Salary
Income
0.030 0.051 0.021 -0.024 -0.039 -0.015 -0.023 -0.033 -0.010 -0.014 -0.025 -0.011
Cooking
Fuel
0.105 0.291 0.186 0.093 0.139 0.046 0.097 0.272 0.176 0.038 0.326 0.287
Lighting 0.040 0.264 0.224 0.021 0.121 0.100 0.016 0.194 0.177 0.002 0.018 0.016
Growth is never (relative) pro-poor in the dimension of expenditure and education even though the average rate of growth of
this particular dimension is the highest among all the dimensions for all social groups. It is always (relative) pro-poor for
dwelling, ownership of land, cooking fuel and lighting.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 20 / 22
25. Pro-poor growth in multiple dimensions
'Thus, even though the overall poverty rates have declined over time...
growth had not been pro-poor for all population groups and in all
dimensions. Therefore, for any policy stance there is a need to target these
areas' (page 11)
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 21 / 22
28. ed, in
particular how this paper diers from Klasen (2008) and Grosse et al.
(2008).
The methodology needs a more accurate description (for instance, do you
use any weighting procedure for computing MP? This is a relevant issue in
multidimensional poverty, but it is not mentioned in your work. )
Probably you should include s.e. and make some robustness check.
The layout of the paper needs to be revised.
The text needs a careful editing.
There are some important but missing references. The bibliography needs to
be cleaned up, revised and made consistent with the references in the main
text.
Flaviana Palmisano Rotterdam, the Netherlands August 24-30, 2014 22 / 22