Session 6 d merz iariw 28 8-2014 correa paper discussed by merz

Paper by A.F. Correa: An individual-centered approach to multidimensional poverty.
An individual-centered approach to multidimensional poverty -
The case of Chile, Colombia, Ecuador and Peru.
Paper: Andrea Franco Correa*
Discussion: Joachim Merz**
Session 6D: Poverty Measurement and the Durations of Poverty Spells II
Organizer: Lars Osberg (Dalhousie University, Canada)
*UNU-MERIT/Maastricht Graduate School of Governance, Keizer Karelplein 19, 6211 TC Maastricht, the Netherlands,
franco@merit.unu.edu
*‘*Leuphana University Lüneburg, Faculty of Economics, Research Institute on Professions (Forschungsinstitut Freie Berufe
(FFB), Scharnhorststr. 1, 21332 Lüneburg, Germany, merz@uni.leuphana.de, www.leuphana.de/ffb
1 Discussion by Joachim Merz

Paper‘s concern
Selection of the unit of analysis: Individual vs. Household
Multidimensional Poverty Index is the Adjusted Headcount Ratio (AHR) of the
Alkire-Foster (2009) family of indicators.
The used index follows Sen’s capability framework
Index with four dimensions: living conditions, health, education and labor;
aggregated and for three age groups
Comparing results of an individual-level and household-level Multidimensional
Poverty Index for Chile, Colombia, Ecuador and Peru:
Ranking for different poverty cut-offs yields
Individual level: Chile, followed by Colombia, Peru and Ecuador, best MDP scores
Household level: Peru and Ecuador reversed ranking

2 Selection of the unit of analysis: Individual vs. Household
Reasons in favor for the individual instead of household level:
1. Many dimensions of well-being are only relevant at the individual level,
and these dimensions may vary over a person's lifetime (economic insecurity
literature, Osberg and Sharpe (2014))
2. The internal decisions of a household are not, generally, democratic. They reflect a
benevolent dictator's will, who is the main decision-maker for a particular
household (Bolt and Bird, 2003).
Resources are not distributed equally among household members
3. HH-level samples excludes some population (military, homeless … foreigners (?))
4. EU in favor for individuals “all those living in the European Union”
5. Sen (1992) emphasizes the importance of the space to define the space within
which the capabilities of individuals are assessed. Choosing a space (like income)
is choosing the unit of analysis

Summarizing reasons in favor for the individual instead of
household level:
A household is a complex unit with specific and different needs of their single
members.
In particular, the individual approach allows different poverty levels for different
household members, e.g. different poverty levels for different ages (as in the
empirical analysis).

3 Multidimensional Poverty Measurement with Counting
Methods
… counting the dimensions in which individuals are considered poor.
Counting methods to measure multidimensional poverty are better to handle
non continuous data.
Particularly, the later applied Adjusted Headcount Ratio (AHR) converts
original categories or continuous variables in dichotomous variables in order to
calculate the censored headcount ratio.
That is to “count" in how many indicators a person or a household is considered
deprived and censored those which are lower than a predetermined threshold.
Three steps:
a) Identification of the dimensionally poor,
b) Identification of the multidimensionally poor, and
c) Aggregation to find the social poverty function.

a) Identification of the dimensionally poor
Indicator Deprived of a matrix of achievements
, ( ) n d X
for x z z depriviation cut offs
< - = í ij j j
ïî
³ = ìï
= x z
Vector of depriviation counts
=å <
C x z w
with weights
;
1 ( )
ij j 0 ( 1,..., ; 1,..., dim )
ij j
I
for x z i n individuals j d ensions
[ ]
å = > " Î
1
1 . . 0, 1,
d
j j
j
w s t w j d
=
1
( )
d
i ij j j
j
=

b) Identification of the multidimensionally poor
i C ³ k
Multidimensionally poor individual:
with k as the minimum number that an individual should have.
Poverty Identification Matrix
ì 1
³
for c k
Y = í < î
Individual Poverty Function
with as a breadth of poverty measure
( , , , )
0
i
i
i
X Z W k
for c k
( , , , ) ( , , , ) ( , , ) i i i i p X Z W k = Y X Z W k ´ g X Z W
( , , ) i g X Z W

c) Aggregation to find the social poverty function.
Social Poverty Function
( , , , ) 1 ( , , , )
= å
P X Z W k p X Z W k
i i i
N
N =
1
i
The paper uses the Adjusted Headcount Ratio (AHR) of the Alkire-
Foster family of poverty measures (Alkire and Foster, 2011; Alkire, 2011).
The AHR is composed of two sub-indices:
- the headcount ratio (H) that measures the ratio of multidimensionally
deprived individuals and
- the Average Deprivation (A), the measure of breadth.
Formulas not in the paper (here from Alkire and Foster 2011,
Alkire 2011):

Headcount ratio (H) – Incidence of (multidimensional) poverty
H = q q = number of the poor; n = population; (0 £ H £
1)
n
Percentage of (multidimensionally) poor
Average breadth (A) – Intensity of multidimensional poverty
n
å
A C k depriviation score k d A
= = 1
= £ £
C k
( )
( ) ; ( / 1)
i
i
i
q
Average depriviation share among the poor (q);
Percentage of weighted dimensions in which the average poor is deprived
Adjusted Headcount Ratio (AHR) (Alkire & Foster 2011; Alkire 2011:
Average depriviation share among all (n)
0 M = MPI = AHR(?)
AHR = H * A

4 Data
Selected countries: Chile, Colombia, Ecuador and Peru (ca. 28% of South America’s population)

5 Selecting the capability set
Following Sen's capability approach, the measure of multidimensional poverty
requires a list of functionings.
These functionings are understood ”not in commodities that people can
consume but in terms of what people are or do like being healthy, reading or
writing" (Deneulin, 2006, p 4).
Capabilities are combinations of functionings for which an individual makes a
free choice.
The paper defines four categories or dimensions of functionings, based on the
concept of achievements: living conditions, health, education and labor.
It is not possible to measure capabilities with the available information. The
paper assumes that the reached achievements are the desired result of a
voluntary choice.

5.1 Factor analysis to select domains
To detect an underlying structure of the dimensions.
“This proves that …the indicators share a common factor” (?)
5.2 Indicators Proposed
Indicators, following the capability approach, are included if the following
conditions are fulfilled:
1. The indicators express an achievement for individuals in the sense of
creating or improving human capabilities,
2. The indicators do not express subjective well-being,
3. The indicators are not a measure of material or income poverty, and
4. The indicators are comparable among the four datasets.

5.3 Deprivation and Poverty Cut-offs
Cut-offs are described in terms of deprivation (lack of fulfillment).
The particular poverty cut-off (k = 30%) is based on the threshold
defined by the UNDP.
Results for
raw (uncensored results by indicator) and
censored (when the second cut-off (the k=30%) is imposed)
Individual age groups
- children (<18 years),
- adults (between 18 and 59 years) and
- elderly (>=60 years)

Figure 2: Adults (Figure for Children and the Elderly in the paper)

6 Multidimensional Poverty Index – Results
Results for
- Headcount ratio (H),
- Average intensity (A) and the
- Adjusted Headcount Ratio (AHR)
with
- different values of the poverty cut-off (k), and
- differences among the values of the national AHR:
=å - =
D AHR AHR w q group indicator
, ,
1
) (?)
Q
AB q A q B q
q
=
6.1 Individual Multidimensional Poverty

Figure 5: Adults (Figure for Children and the Elderly in the paper)

k - Interpretation (my wording):
A person’s depriviation must be at least k% of the weighted indicators to be considered
AHR poor: Ci ³ k
.
The greater k, the more severe poverty is regarded (MPI, Alkire Roche Santos Seth
2011)
Headcount ratio
Why small ks result in large Headcount ratios?
There are less people with a very severe poverty (high k, small H),
but more people with a less severe poverty (small k, high H)
Household level:
•Electricity - Measure of the lack of basic services inside the housing - unit All
The greater the lack of fulfillment, the greater k, the less number of peoples are severely poor k
Individual level:
•Literacy Measures the ability of an individual to read and write - Adults and Elderly:
The greater the lack of literacy ability, the greater k, the less people are severely poor
Average intensity: the larger k, the greater is the depriviation measured by
i C ³ k

Figure 7: Total
Chile remains as the least multidimensional
deprived country (0.0077), followed by
Colombia (0.033), Peru (0.1382) and Ecuador
(0.1573), for the particular case of k=30%.
At higher values of k, the ordering does not
prevail. The ranking is reverted for the last two
countries (Ecuador and Peru).

6.2 Household Level Multidimensional Poverty Index
Same index, same indicators and cut-offs, equal weights to all 16 indicators.
But unit of analysis: household.
“A household is considered deprived in one particular indicator when at least
one member is deprived.“

Figure 8: Multidimensional Poverty Index for Households
H ranking:
1 Chile,
2 Colombia,
3 Ecuador,
4 Peru
Same as with
the individual
level as the unit
of analysis

Conclusion
Rraw and censored headcount: the levels of Chile are comparatively lower that those
of the rest of the countries.
AHR: Chile is the country with the best performance in this ranking.
Individual approach:
Elderly: showed that the elderly, as opposed to children, is the most multidimensional
deprived population subgroup.
Driven factors: the minimum years of education, and the access to sources of income
after retiring age (either pension or labor).
Ecuador and Peru: poor coverage of health insurance and the limited access of effective
health care in case of need.
Children and adults: living conditions are the main source of deprivations. Lack of piped
water and adequate facilities for sanitation is known for increasing probabilities of
contagious diseases which are still a major cause of death. Conflicts with the ownership
of the land are of particular interest in the Colombian case. Crowdedness is a conflicting
aspect in the case of Peru.
Household level: Chile was still the least multidimensional poor country followed by
Colombia, Ecuador and Peru. The ranking of the two last countries was reverted from the
individual case.

Discussion

Methods
Adjusted Headcount Ratio (AHR)
After preparation of counting essentials, final formula missing
Aggregated National AHR: index not sufficient defined (q=groups?)
Factor Analysis
Should show the underlying structure of dimensions.
Which dimensions? The four ones: living conditions, health, education and labor. P. 13:
sanitation? Factor analysis: What for?
Poverty indicator cut-offs
Just given in the appendix, problems?
Poverty cut-off k
“The decision about the poverty cut-off induces arbitrariness in the calculation of the
overall result and could cause contrary conclusions (Lasso de la Vega, 2009).
Interestingly, in the empirical case presented in this paper, it is not possible to establish a
consistent ranking between Ecuador and Peru when a k higher than 50% was
considered.”
Though k simulated, k=30% for discussion: reason?
Illustrative examples would help

Empirical Results
Domains (3?), index dimensions(4), indicators (16)
Missing explanation of actual relations
Discussing results
Mostly only replication of figure results, missing reasons behind (only in the conclusion
some sentences)
Average depriviation (A): only little mentioning, underlying weights not discussed
Driving factors unfolded
What‘s about problems?
Household as the unit of analysis:
“A household is considered deprived in one particular indicator when at least one
member is deprived.“
Paper: only one footnote 20: „The most popular application of household-level MPI, performed by the
UNDP and the OPHI, considers for instance in the dimension of education the complete opposite
condition stated here. The cut-offs defined is having at least one household member who has
complete five years of schooling and having at least one school-age child (up to grade 8) who is not
attending school. (UNDP, 2013, Technical Notes p 7).“
Missing: Discussion of concept and consequences.

Policy concern
Missing discussion of targeted policy possibilities based on AHR;
Which k?
Which crucial indicators behind?
Which recommendations?
What‘s about conflicting/interacting/interdependent indicators/dimensions/goals?
Overall
Interesting paper with a lot of work behind.
Interesting the country specific ranking.
Good: Showing the k sensitivity of results.
Good: An application of Sen‘s Capability approach to construct indicators
Policy: Eye-catching multidimensional figure.
Wish: Give more practical insights in the functioning and effects of the single index bricks.

Thank you for your attention
An individual-centered approach to multidimensional poverty -
The case of Chile, Colombia, Ecuador and Peru.
Paper: Andrea Franco Correa*
Discussion: Joachim Merz**
Session 6D: Poverty Measurement and the Durations of Poverty Spells II
Organizer: Lars Osberg (Dalhousie University, Canada)
*UNU-MERIT/Maastricht Graduate School of Governance, Keizer Karelplein 19, 6211 TC Maastricht, the Netherlands,
franco@merit.unu.edu
*‘*Leuphana University Lüneburg, Faculty of Economics, Research Institute on Professions (Forschungsinstitut Freie Berufe
(FFB), Scharnhorststr. 1, 21332 Lüneburg, Germany, merz@uni.leuphana.de, www.leuphana.de/ffb

Session 6 d merz iariw 28 8-2014 correa paper discussed by merz

Recommended

Recommended

More Related Content

Similar to Session 6 d merz iariw 28 8-2014 correa paper discussed by merz

Similar to Session 6 d merz iariw 28 8-2014 correa paper discussed by merz (20)

More from IARIW 2014

More from IARIW 2014 (20)

Recently uploaded

Recently uploaded (20)

Session 6 d merz iariw 28 8-2014 correa paper discussed by merz