The Growth Semi-Elasticity
of Poverty Reduction
Explaining Heterogeneity
across Space and Time
Stephan Klasen and Mark Misselhorn
Background
• Many efforts to estimate a general growth
elasticity of headcount poverty
• Chen and Ravallion (1997) estimated growth
elasticity to be around 3;
• World Development Report 2000/2001: elasticity
between closer to 2 (Bhalla: 5!)
• Cross-Country Heterogeneity, sample and time
period differences
• Must be the case: Mathematical link between
growth, inequality change, and (absolute)
poverty reduction;
The Growth Semi-Elasticity of Poverty Reduction 2
Poverty Reduction and Growth
Source: Bourguignon 2003
The Growth Semi-Elasticity of Poverty Reduction 3
Decomposition
The Growth Semi-Elasticity of Poverty Reduction 4
Background
• Bourguignon (2003): Growth elasticity of poverty
reduction depends on:
– Initial inequality
– Location of poverty line (relative to mean incomes)
• Under assumption of lognormal income
distribution, can calculate these elasticities
precisely. Works empirically quite well (for
headcount, not so well for depth, severity).
• Allows quick prediction of poverty impacts of
growth and distributional change.
The Growth Semi-Elasticity of Poverty Reduction 5
This Paper
• Argues:
– Growth Semi-Elasticity (i.e. percentage point change
in poverty as a result of growth) more useful from a
policy perspective;
– Avoids some distortions in growth elasticity;
• Derives:
– Determinants of growth and distribution semi-
elasticity under the assumption of log-normal income
distribution;
• Applies
– Poverty spells database to show a better empirical fit
enabling a greater use of data, and better
interpretation and prediction.
The Growth Semi-Elasticity of Poverty Reduction 6
Decomposition Identity
The Growth Semi-Elasticity of Poverty Reduction 7
Decomposition Identity
• Growth elasticity:
– Higher, the lower initial inequality (at least as
long as poverty line smaller than mean
income).
– Higher in richer countries (low ratio z/y).
– By implication: increasing over time as
countries grow.
• Growth semi-elasticity:
– Higher, the lower initial inequality (if z
– Generally higher in high poverty countries.
The Growth Semi-Elasticity of Poverty Reduction 8
Growth Elasticity of P0
The Growth Semi-Elasticity of Poverty Reduction 9
Growth Semi-Elasticity of P0
The Growth Semi-Elasticity of Poverty Reduction 10
Why Semi-Elasticity?
• Percentage point changes in poverty easier to
understand and compare and more relevant for policy-
makers;
• ‚Bias‘ in Poverty Elasticity:
• Higher in richer countries;
• Growing with development;
• More affected by growth in richer countries;
• Empirical advantages:
• No need for arbitrary ‚sample selection‘;
• Better fit for all poverty measures (and thus better
predictive power).
The Growth Semi-Elasticity of Poverty Reduction 11
The Growth Semi-Elasticity of Poverty Reduction 12
Empirical Application
• Poverty spells database;
• Empirical Strategy (following Bourguignon, 2003):
• Dependent variable: Percent (percentage point) change in
poverty;
• ‚Naive‘ poverty-growth model;
• Augmented by change in inequality;
• Augmented by location of poverty line and initial Gini
(mathematical relation to log normal distribution parameter);
• Augmented by interactions with inequality change;
• Compared to calculated elsticities under lognormal assumption;
• Problem for Elasticity regressions:
• Must exclude data where % changes are very large or
undefined.
• Unable to explain poverty depth/severity measures.
The Growth Semi-Elasticity of Poverty Reduction 13
Relative Changes in P0
The Growth Semi-Elasticity of Poverty Reduction 14
Relatives Change in P1
The Growth Semi-Elasticity of Poverty Reduction 15
Relatives Change in P2
The Growth Semi-Elasticity of Poverty Reduction 16
Absolutes Changes in P0
The Growth Semi-Elasticity of Poverty Reduction 17
Absolutes Change in P1
The Growth Semi-Elasticity of Poverty Reduction 18
Absolutes Change in P2
The Growth Semi-Elasticity of Poverty Reduction 19
Application
• To understand past poverty reduction
performance:
– E.g. India‘s recent success versus lower success in
growing SSA economies (e.g. South Africa);
• To ‚predict‘ poverty-effect of policy based on
assumption about their growth and distributional
effects;
• To simulate growth and/or distributional change
requirements to reach the MDGs; or
alternatively: to project MDG success based on
assumed growth/distributional change patterns;
• Note: Measurement not policy tool !
The Growth Semi-Elasticity of Poverty Reduction 20
‚Predicting‘ Headcount Poverty
The Growth Semi-Elasticity of Poverty Reduction 21
Conclusion
• Semi-elasticities more policy-relevant and
not prone to ‚bias‘;
• Allow integration of more growth spells;
• Can explain changes in FGT-Measures
better;
• Could also be used for simulations about
growth and distributional change
requirements to achieve MDGs;
The Growth Semi-Elasticity of Poverty Reduction 22
Distribution Semi-Elasticity
The Growth Semi-Elasticity of Poverty Reduction 23