Measuring Poverty: Inequality Measures

Charting Inequality
Share of Expenditure of Poor
Dispersion Ratios
Lorenz Curve
Gini Coefficient
Theil Index
Comparisons
Decomposition

Poverty Measures, Lao PDR

Poverty in Lao PDR 1997/98

Dept of Poverty in Lao PDR 1997/98

Lecs II

Lecs II
Percentage
57.9 - 71.3
49.7 - 71.3
39.2 - 49.7
13.5 - 39.2

Severity of Poverty in Lao PDR 1997/98

Lecs II

Percentage
17 to 24.7
11.9 to 17
9.5 to 11.9
2.8 to 9.5

Percentage
7.1 to 12.1
4.3 to 7.1
3.3 to 4.3
0.8 to 3.3

Income Distribution

Types of analysis

Functional distribution

Functional distribution
Size distribution
income accrued to factors of production such as
land, labor, capital and entrepreneurship

Size distribution

income received by different households or

individuals
3

What is Inequality?

Dispersion or variation of the distribution of

income/consumption or other welfare indicator

Equality everyone has the same income

Inequality certain groups of the population have
higher incomes compared to other groups in the
population

Why measure inequality? (1)

Indicator of well-being

Position of individual relative to rest of population

Position of subgroup relative to other subgroups

Different measures, different focus

Poverty measures (HC, PGI, SPGI, etc) focus on the

situation of individuals who are below the poverty
line the poor.
Inequality is defined over the entire population, not
only for the population below a certain poverty line.
5

Why measure inequality? (2)

Inequality is measured irrespective of the mean

or median of a population, simply on the basis
of the distribution (relative concept).
Inequality can be measured for different
dimensions of well-being:
consumption/expenditure and income, land,
assets, and any continuous and cardinal
variables.
6

Charting Inequality: Histogram

Divide population
into expenditure
categories
Example: 20% of
households are in
category 4

Example: Income Classes

Example: Bar Chart, Income Classes

Percentage of families falling in each class

.2

Cumulative Distribution
.4
.6

.8

Example: CDF of Per Capita Expenditure

200000

400000
600000
800000
Per capita Total Expenditure

1000000

10

Distribution: Quintile and Deciles

11

Expenditure/Income-iles

Divide population into groups ranked from

poorest to richest based on expenditure (or
income)
Divide into 5 groups: income or expenditure
quintiles

Lowest 20% or first quintile poorest

Highest 20% or fifth quintile richest

Divide into 10 groups: income or expenditure

deciles
12

Expenditure per capita by Quintile, Viet Nam (1993)

Quintile
Poorest

Richest

First: Lowest
Second: Low-middle
Third: Middle
Fourth: Mid-upper
Upper: Fifth
All

Per Capita
% of Total
Expenditure Expenditure
518
756
984
1,338
2,540
1,227

8.4
12.3
16.0
21.8
41.4
100.0
13

Share of Income of Poorest, Korea

Income decile

2000

2001

2002

2003

1st

2.9

2.9

3.0

2.7

2nd

4.7

4.6

4.7

4.8

3rd

5.8

5.7

5.8

6.1

4th

6.9

6.8

6.9

7.1

5th

7.9

7.8

7.9

8.1

6th

9.1

9.1

9.2

9.3

7th

10.5

10.5

10.5

10.7

8th

12.2

12.3

12.4

12.5

9th

14.7

15.0

15.1

15.0

10th

25.4

25.4

24.6

23.8

14

Inequality Measures Based on -iles

Share of income/consumption of lowest ile

Dispersion ratios

15

Share of Consumption of the Poorest

Definition: Total consumption/income of the poorest

group, as a share of total consumption/income in the
population.
m

C( x )

y
i 1
N

y
i 1

Where
N is the total population
m is the number of individuals in the lowest x %.

16

Poorest Quintiles Share in National Income

or Consumption (UNSD, 2005)

17

Dispersion Ratio

Definition: measures the distance between

two groups in the distribution of expenditure (or
income or some other characteristic)
Distance: average expenditure of the richest
group divided by the average expenditure of
the poorest group
Example:
average expenditure of fifth quintile
Dispersion ratio=
average expenditure of first quintile

18

Dispersion Ratios: Examples

Expenditure decile

Median

1st

37,324

2nd

47,289

3rd

54,397

4th

62,929

5th

74,775

6th

89,478

7th

108,633

8th

129,890

9th

172,011

10th

267,214

(1)
10th:1st

(2) 10th :1st &

2d
(Kuznets
ratio)
19

Lorenz Curve and Gini Ratio

20

Lorenz Curve

21

Lorenz Curve: Interpretation (1)

If each individual
had the same
consumption (total
equality), Lorenz
curve would be the
line of total
equality.
If one individual had
all the consumption,
Lorenz curve would
be the curve of
total inequality.

100
90
Cumulative % of consumption

80
70

l
a
t

60
50
40
30
20
10
0
0

to
of t y
z
n e al i
n
i
e
L qu
r
e
e
Lo rv
cu
Curve of total
20
40
60inequality
80
100
Cumulative % of population

22

Lorenz Curve: Interpretation (2)

The further away from

the line of total
equality, the greater
the inequality.
Example: Inequality is
greater in country D
than in country C.

100
90
80
70
60
50
40
30
20
10
0

C
D

50

100

23

Comparing Lorenz Curves

100

D
C

100

24

Lorenz Criterion

Whenever one Lorenz curve lies above another

Lorenz curve the economy with the first Lorenz curve
is more equal, and the latter more unequal

e.g. A is more equal; D is more unequal

When 2 curves cross, the Lorenz criterion states that

we need more information (or additional assumptions)
before we can determine which of the underlying
economies are more equal

e.g. curves B and C

25

Constructing Lorenz Curve, Example (1)

Quintile

Cumulative
Share of
Population (p)

% of Total
Expenditure

Cumulative
share of
expenditure (e)

First

20

8.4

8.4

Second

40

12.3

20.7

Third

60

16.0

36.7

Fourth

80

21.8

58.5

100

41.4

100.0

Fifth

26

Constructing Lorenz Curve, Example (2)

27

Gini Coefficient: Definition

Measure of how close to or far from a given

distribution of expenditure (or income) is to
equality or inequality
Varies between 0 and 1
Gini coefficient 0 as the expenditure/income
distribution absolute equality
Gini coefficient 1 as the expenditure/income
distribution absolute inequality
28

Gini Coefficient & Lorenz Curve (1)

Area between line of
equality and Lorenz
Curve (A)
A

If A=0 then G=0

(complete equality).

29

Gini Coefficient & Lorenz Curve (2)

Area below Lorenz

Curve (B)
If B=0 then G=1
(complete inequality).

30

Gini Coefficient & Lorenz Curve (3)

100
Cumulative % of consumption

90
80
70

l
a
t

60
50
40
30
20
10
0
0

to
of t y A
e li
z
n
n
a
i
e
L qu
r
e
e
Lo rv B
cu
Curve of total
20
40
60inequality
80
100

Gini coefficient (G) is

the ratio of the area
between the line of
total equality and the
Lorenz curve (A) to
the area below the
line of total equality
(A+B)

Cumulative % of population

31

Lorenz Curve and Gini Coefficient

e

32

Gini Coefficient: A Formula

Heres one. (There are other formulations.)

Cov y i ,fi
G = 2
1 N
yi
N i=1

Where:

N is population size
y is expenditure of individual
f is rank of individual in the distribution

33

Gini Coefficient: +s and s

(+) Easy to understand, in light of the Lorenz curve.

(-) Not decomposable: the total Gini of the total
population is not equal to the sum of the Ginis for its
subgroups.
(-) Sensitive to changes in the distribution, irrespective
of whether they take place at the top, the middle or the
bottom of the distribution (any transfer of income
between two individuals has an impact, irrespective of
whether it occurs among the rich or among the poor).
(-) Gives equal weight to those at the bottom and those
at the top of the distribution.
34

Measures of Inequality, Example

35

Poor people in Senegal get bigger share of

income than poor people in the US

36

General Entropy Indexes

1 1
GE ( ) 2

yi

1
i 1 y

represents the weight given to distances between

incomes at different parts of the income distribution

Sensitive to changes at the lower end of the distribution if

is close to zero
Equally sensitive to changes across the distribution if is 1
(Theil index)
Sensitive to changes at the top of the distribution if takes a
higher value.

37

GE(1) and GE(0)

GE(1) is Theils T index

yi
1 N yi
GE (1) ln( )
N i 1 y
y

GE(0), also known as Theils L, is called mean

log deviation measure :
1
GE (0)
N

y
ln( )

yi
i 1

38

The Theil Index: Definition

1
T
N

i 1

yi
yi
ln
y
y

where yi is expenditure of ith individual

y is average expenditure of population

Varies between 0 (total equality) and 1 (total

inequality). The higher the index, the more unequal
the distribution of expenditure (or income).
39

Theil Index: +s and s)

(+) Gives more weight to those at the bottom of the

income distribution.
(+) Can be decomposed into sub-groups: the
population Theil is the weighted average of the index
for each sub-group where the weights are population
shares of each sub-group
(-) Difficult to interpret
(-) Sensitive to changes in the distribution, irrespective
of whether they take place at the top, the middle or the
bottom of the distribution (any transfer of income
between two individuals has an impact, irrespective of
whether it occurs among the rich or among the poor).
40

Atkinsons Index

This class also has a weighting parameter

(which measures aversion to inequality)
The Atkinson class is defined as:
1
A 1
N

yi

i 1 y
N

(1 )

Ranges from 0 (perfect equality) to 1

41

Criteria for Goodness of Measures

Mean independence If all incomes are doubled,

measure does not change.
Population size independence If population size
changes, measure does not change.
Symmetry If two individuals swap incomes, the
measure does not change.
Pigou-Dalton transfer sensitivity Transfer of income
from rich to poor reduces value of measure.
Decomposability It should be possible to break down
total inequality by population groups, income source,
expenditure type, or other dimensions.
42

Checklist of Properties
Property

Dispersion Gini Theil

Mean independence
Population size independence
Symmetry
Pigou-Dalton Transfer Sensitivity
Decomposability
43

Inequality Comparisons

Extent and nature of inequality among certain groups of

households. This informs on the homogeneity of the
various groups, an important element to take into
account when designing interventions.
Nature of changes in inequality over time. One could
focus on changes for different groups of the population to
show whether inequality changes have been similar for
all or have taken place, say, in a particular sector of the
economy.
Other dimensions of inequality: land, assets, etc
44

Example: Inequality Changes over Time

Year
1985
1988
1991
1994
1997
2000

Poverty
Rate
48
40
40
36
32
34

Gini
Coefficient
0.4466
0.4446
0.4680
0.4507
0.4872
0.4818

46

Example: Gini Ratios, Indonesia

48

Decomposition of Inequality

49

At One Point in Time (1)

Inequality decompositions are typically used to estimate

the share of total inequality in a country which results
from different groups, from different regions or from
different sources of income.
Inequality can be decomposed into between-group
components and within-group components. The first
reflects inequality between people in different subgroups (different educational, occupational, gender,
geographic characteristics). The second reflects
inequality among those people within the same subgroup.
50

Example, Viet Nam (1993)

51

Decomposition of Inequality, Egypt

52

At One Point in Time (2)

Inequality decompositions can be calculated for the

General Entropy indices, but not for the Gini coefficient.
For future reference, the formula is:
k

I IW I B v . f
j 1

1
j

.GE ( ) j 2

j 1

yj

f j .
y

where fi is the population share of group j (j=1,2, k),

vj is the income share of group j;
yj is the average income in group j.
53

Changes over Time (1)

Changes in the number of people in various

groups or allocation effects
Changes in the relative income (expenditure) of
various groups or income effects
Changes in inequality within groups or pure
inequality effects.

54

Changes over Time (2)

The formula can get complicated, and is

typically used for GE(0) only, as follows:
k

j 1

j 1

j 1

GE (0) f j GE (0) j GE (0) j f j log( j ) f j ( v j f i ) log( ( y )) j

Pure inequality
effects

All ocation effects

j 1

Income effects

where is the difference operator, j is the mean income of group j relative to the
overall mean ( j (y j )/ (y)), and the over bar represents averages.

55

Poverty Changes over Time (1)

Poverty is fully determined by the mean income

or consumption of a population, and the
inequality in income or consumption in the
population.
Changes in poverty can result from changes in
mean income/consumption growth or from
changes in inequality.
56

Poverty Changes Over Time (2)

14
12

8
6
4

10

mean = 100

Original
distribution
Higher mean
(grow th)

poverty line = 50

10

Inequality effect

Share individuals (%)

Share individuals (%)

12

mean = 130

poverty line = 50

14

mean = 100

Growth effect

Original
distribution
Low er
inequality

8
6

20 40 60 80 100 120 140 160 180 200 220 240

Income

20 40 60 80 100 120 140 160 180 200 220 240

Income

57

Poverty Changes Over Time (3)

Decomposition can be done as follows:

P [ P ( 2 , Lr ) P ( 1 , Lr )] [ P ( r , L 2 ) P ( r , L1 )] Rr
Growth effect

Inequality effect

Residual

Where P ( t , Lt ) is the poverty measure corresponding to

mean income in period t of t and a Lorenz curve Lt .

58

Conclusions & Recommendations

Inequality is a difficult concept to measure.

For analysis, use several measures:

Lorenz curve
Gini coefficient
Dispersion ratios
Share of expenditure of the poorest x%
Theil Index

Analysis

Comparisons across subgroups

Comparisons over time
59