Assessing the Poverty Impact of Economic

Growth:The Case of Indonesia

B. Essama-Nssah and Peter J. Lambert

World Bank Poverty Reduction Group and

University of OregonOctober, 2006

2

Introduction Context

Poverty Reduction: A fundamental objective of development (MDGs) Hence we need

a metric for assessing effectiveness in terms of this objective.

meaningful ways to assess poverty impact of shocks and policies.

Economic growth: Potentially a powerful instrument of poverty

reduction. Yet, same rate of growth can lead to different levels

of poverty reduction in different circumstances.

Issue When to declare a growth pattern pro-poor?

3

General approach

Logic of social impact evaluation understood as an assessment of changes in individual and social welfare attributable to a shock or policy.

Three basic dimensions:

Identification of individual situations. Aggregation of individual situations into an

indicator of a social state. Ranking of social states

4

Specification

Identification: Individual situations represented by point elasticity

of income w.r.t. a change in aggregate income. Aggregation:

Use members of the class of additively separable poverty measures (e.g. Watts, FGT).

Ranking: A pro-poor growth pattern achieves a reduction in

poverty over and above that which is feasible in a benchmark case (either desirable or counterfactual).

We choose as benchmark the amount of poverty reduction attainable under distributional neutrality (equiproportionate growth).

5

Case of Indonesia A long reputation of high achievements in growth

and poverty reduction, even in the face of adverse conditions (Ravallion and Huppi 1991; World Bank 1996).

Is the observed poverty reduction more or less than we would have observed had growth been distributionally neutral?

We answer this question using 1993-2002 data.

We also unbundle the pattern of growth, to identify the contributions of income components to the overall outcome.

6

Focus of the rest of the presentation:

Evaluation Framework A General Impact Indicator Poverty Focus

Empirical Results Data Profile of Poverty and Inequality Pro-Poorness

Conclusion

7

Evaluation Framework

Identification

Let x be an individual characteristic (e.g. income) that is responsive to growth (or policy change).

Let f(x) be the frequency density function for x among the population.

Let u(x) be any individual attribute (e.g. poverty), a function of x.

8

Aggregation Average x:

Average u(x) :

Impact Growth rate of mean of x:

Growth pattern (elasticity of individual income with respect to y):

q(x) is a normalized growth incidence curve (Ravallion & Chen, 2003).

xmdxxxfy

0)(

xm dxxfxuW0

)()(

y

y

ydy

xdx

yd

xdxq

/

/

)ln(

)ln()(

9

Impact, continued

Change in average u(x) induced by change in y:

As a proportion:

As an elasticity:

Component breakdown (x=x1 +x2 xq(x)=x1 q(x1)+x2 q(x2)):

xm dxxfxqxux

WW

W0

)()()(

xm

wy dxxfxqxuxWyd

Wd0

)()()(1

)ln(

)ln(

xx mmwy dxxfxqxux

Wdxxfxqxux

W 0 220 11 )()()(1

)()()(1

10

Poverty Focus

x is now income or expenditure and mx is the poverty line, call it z.

The function is an indicator of individual deprivation.

Overall poverty

)|()( zxxu

z

dxxfzxPW

0 )()|(

11

Poverty Impact

Aggregate poverty elasticity:

- Decomposition by income components:

z

Pwy dxxfxqzxxP

q

0 )()()|(

1)(

zz

P dxxfxqzxxP

dxxfxqzxxP

q0 220 11 )()()|(

1)()()|(

1)(

12

Indicator of pro-poorness:

where q0 is the growth pattern associated with distribution

neutrality, i.e. q0(x) = 1 for all x.

- Indicator measures the difference between poverty reduction under distribution neutrality and amount produced by observed growth pattern.

- Typically, the growth elasticity ΦP(q) is negative, since positive growth induces poverty reduction for all incomes below the poverty line.

If the pro-poorness index πP(q) is positive, the observed growth pattern leads to more poverty reduction than the benchmark case

)]()([)( 0 qqPq PPP

z

dxxfxqzxx0

)(]1)()}[|({

13

Factoring growth into scale and distribution contributions

Elasticity:

Pro-poorness:

growth is pro-poor if distributional component of poverty elasticity is negative.

PPzz

P dxxfxqzxxP

dxxfzxxP

q

0

0 )(]1)()[|(

1)()|(

1)(

PP Pq )(

14

Ratio Comparisons

Instead of an additive comparison, one can consider the ratio of the actual elasticity to the benchmark elasticity. Growth is pro-poor if this ratio is greater than 1.

Ratio comparison for Watts index leads to Ravallion and Chen’s (2003) “mean growth rate for the poor” measure

Ratio comparisons also underlie the Kakwani et al (2004) measure called “poverty equivalent growth rate”.

Decomposability not established for ratio measures

15

Empirical Results

Two types of datasets for Indonesia used in this study

Distribution of household expenditure per decile (in 1993 PPP dollars) from World Bank Global Poverty Monitoring database.

SUSENAS household surveys (1999, 2002) used to achieve decomposition of pro-poorness across income/expenditure components

Poverty line: about 2 dollars a day.

16

Poverty and Inequality in Indonesia

1993-1996 Poverty falls while inequality increases. Likely outcome of successful adjustment to oil shock

1996-1999 Poverty increases while inequality falls Likely outcome of the 1997 Asian financial crisis.

1999-2002 Poverty falls, inequality goes up. Signs of recovery: macroeconomic stability associated

with reduced vulnerability to external shocks.

17

Profile of Poverty and Inequality in Indonesia, continued

1993 1996 1999 2002 Headcount 61.53 50.49 55.29 52.41 Poverty Gap 21.03 15.33 16.56 15.68 Squared Poverty Gap 9.16 6.02 6.49 6.09 Watts 28.12 19.73 21.33 20.12 Gini 32 36 31 34

Source: Authors’ calculations

18

Pro-Poorness in Indonesia Additive comparisons

Ratio Comparisons

Watts Poverty Gap Squared Poverty Gap 1993-1996 -0.22 -0.14 -0.09 1996-1999 -0.20 -0.14 -0.08 1999-2002 -0.039 -0.022 -0.015 1993-2002 -0.13 -0.09 -0.05 1996-1999 (Counterf actual) 0.20 0.14 0.08

Source: Authors’ calulations

Watts Poverty Gap Squared Poverty Gap 1993-1996 0.64 0.65 0.64 1996-1999 0.60 0.61 0.58 1999-2002 0.74 0.74 0.73 1993-2002 0.79 0.77 0.80 1996-1999 (Counterf actual) 1.40 1.39 1.42

Source: Authors’ calculations

19

Growth pattern in Indonesia for 1999-2002

Aggregate Pattern of Growth (1999-2002)

0.6

0.8

1.0

1.2

1.4

1.6

0 10 20 30 40 50 60 70 80 90 100

Cumulative Percentage of the Population

20

Components of growth for 1999-2002

I ncidence of Growth on Expenditure Components (1999-2002)

-2

-1

0

1

2

3

4

0 10 20 30 40 50 60 70 80 90 100

Cumulative Percentage of Population

RiceOther FoodEducation

HealthOther Non-Food

21

Pro-poorness components in Indonesia for 1999-2002

Watts Poverty Gap Squared Poverty Gap Rice -0.173 -0.125 -0.062 Other Food 0.044 0.037 0.015 Education 0.005 0.003 0.002 Health 0.009 0.006 0.003 Other Non-Food 0.077 0.057 0.027

Aggregate -0.039 -0.022 -0.014

Source: Authors’ calculations

22

Distributional component of growth

Shapley Decomposition

distributional effect alleviated some of the negative impact of the 1997 economic crisis.

Overall Growth Distribution 1993-1996 -5.70 -8.83 3.13 1996-1999 1.23 4.97 -3.74 1999-2002 -0.88 -2.97 2.09 1993-2002 -5.35 -6.87 1.52

23

Component Analysis, 1999-2002

Outcome driven mainly by what happened to expenditure on rice, with some help from expenditure on other food items.

Rice represents 26 percent of total expenditure for the poor.

Total food expenditure including rice is about 73 percent of total household expenditure for the poor.

24

Winners and losers among the poor

Actual growth pattern for 1999-2002 crosses benchmark pattern twice before the headcount ratio (55 percent).

q(x) below benchmark up to 20th percentile, and between 43rd and 55th percentiles (these are the losers)

winners located between 20th and 43rd percentiles.

25

Appraisal of gains and losses Pro-poorness at percentiles:

Pro-Poorness at Percentiles (1999-2002)

-5

-4

-3

-2

-1

0

0 10 20 30 40 50 60 70 80 90 100

Watts

Poverty Gap

Squared Poverty Gap

Cumulative Percentage of the Population

26

The gains versus the losses

Percentile pro-poorness curves lie below zero: economic growth not pro-poor at any percentile up to the headcount.

according to our metric, benefits enjoyed by the poor between the 20th and the 43rd percentiles not high enough to compensate for losses experienced by those who came before.

27

Conclusions Pro-poor growth analysis fits nicely within the logic of

social evaluation

An assessment of changes in individual and social welfare attributable to the process of economic growth.

Measurements depend critically on underlying value judgments. Our metric for pro-poorness is defined by the

following choices: (1) individual outcomes represented by point elasticity of

income; (2)social welfare measured by poverty measures in the

additively separable class; (3) standard of comparison based on distributional

neutrality.

28

Application of this approach to data for Indonesia reveals that:

Overall poverty reduction achieved over period 1993-2002 far below what distributionally neutral growth would have achieved.

Focusing on the 1999-2002 period:

Some poor people gained from the economic growth that occurred over that period, but these gains do not measure up to the losses suffered by the rest of the poor.

The behavior of categories of expenditures over the same period reveals that the weak performance is due mainly to changes in food expenditure.

29

References Essama-Nssah, B. (2005). A unified framework for pro-poor growth analysis.

Economics Letters, vol. 89, pp. 216-221.

Essama-Nssah, B. and Lambert P.J. (2006). Measuring the Pro-Poorness of Income Growth within an Elasticity Framework. World Bank Policy Working Paper No. 4035 (October)

Kakwani, N.C., S. Khandker and H.H. Son (2004). Pro-poor growth: concepts and measurement with country case studies. Working Paper Number 2004-1, International Poverty Center, Brasil.

Kraay, Art. (2006). When is growth pro-poor? Evidence from a panel of countries Journal of Development Economics 80, 198-227.

Ravallion, Martin and Huppi Monika. 1991. Measuring Changes in Poverty: A Methodological Case Study of Indonesia during an Adjustment Period. The World Bank Economic Review, Vol. 5, No. 1:57-82.

Ravallion, M. and S. Chen (2003). Measuring pro-poor growth. Economics Letters, vol. 78, pp. 93-99.

World Bank (1996). Indonesia: dimensions of growth. Report No.15383-IND. Country Department III, East Asia and Pacific Region. Washington, D.C.: The World Bank.

30

End.