Sectoral Decompositions of Changes in Poverty

(Ravallion - Huppi Decomposition)

• Definitions and Concepts 
• Limitations 
• Notes / Extensions 
• Quick Results (with ADePT and STATA) 
• Data Requirements 
• Helpful Tips 
• Annotated Examples 
• References / Related Papers 

Definitions and Concepts

The sectoral decomposition introduced by Ravallion and Huppi (1991) quantifies the relative contributions of changes in poverty within sectors and of inter-sectoral population shifts to changes in aggregate poverty. Sectors in this decomposition are typically defined as (i) urban and rural, (ii) sectors of employment, or (iii) sub-national regions.

With this method, the change in a poverty measure (e.g., headcount index, poverty gap, or squared poverty gap) is decomposed into three components: (i) intra-sectoral, (ii) inter-sectoral, and (iii) interaction:

where t₀ is the initial year of the period, t_n is the final year of the period, and k is the sector, P is the poverty measure, and s is the population share.

• The intra-sectoral component represents the change in poverty attributable to changes in poverty rates, holding the population share constant at the initial level. In other words, this is the change in poverty that would have occurred if the population shares in each sector did not change.
• The inter-sectoral (population shift) component represents the change in poverty attributable to changes in population shares in each sector, holding the poverty level within a sector constant. Depending on how sectors are defined, this component represents poverty changes resulting from people shifting either physical locations between poor and rich areas (e.g., between urban and rural areas or between regions) or shifting employment sectors.
• The interaction component represents the change in poverty attributable to both changes in population shares and poverty levels in sectors. This component can be “interpreted as a measure of the correlation between population shifts and intrasectoral changes in poverty” (Ravallion and Huppi).

Limitations

Notes / Extensions

Since a household may derive income from more than one sector of employment (e.g. agriculture, industry, service), assigning a household to a single sector often requires the analyst to make some arbitrary decisions on how this should be done. In most cases, the sector that generates the greatest amount of income for the household is often assigned to a household. [Sensitivity analysis may be useful if there are concerns regarding the classification.]

Complementing this decomposition with estimations of sectoral growth rates and sectoral growth elasticities of poverty can indicate how growth is translating into poverty reduction in the different sectors. For more details, see the section on growth elasticity of poverty and Ravallion and Datt (2002).

Quick Results with ADePT or Stata

• Stata: The command “sedecomposition” will perform this decomposition quickly and display the results in a single table. The Stata ado file (written by Michael Lokshin and Martin Ravallion) must be installed on your computer for the command to work. See the annotated examples for step-by-step details for installation and the “sedecomposition” command.
• ADePT (Automated DEC Poverty Tables): ADePT is a Stata based software program that automates the economic analysis typical for poverty assessments and produces a package of graphs and tables with standard poverty and inequality statistics or individual outputs a la carte. This decomposition is included in the ADePT Poverty module and results are displayed in Table 3.6 (by regions) and 3.6a (for urban-rural). For more details on installation and use, refer to the ADePT User’s Guide and website.

Data Requirements

• 2 or more comparable household surveys
• Per capita (or per adult equivalent) expenditure or income variables with measures in real terms (e.g., constant local currency).
• Poverty line variables (or numbers) also in real terms.
• Sector variable (e.g., categorical variables for urban-rural, region of residence, or sector of employment).
• Survey sampling weights.

Helpful Tips

• The welfare measures and poverty lines for the two years should be comparable (e.g., consistent survey design and measures) and appropriately deflated (e.g., using the CPI or relevant poverty lines).
• The growth-inequality (Datt-Ravallion) decomposition can complement this sectoral decomposition.
• Residual can be averaged out if Shapley decomposition is performed. In practice, if you are presented with the results for only one of two possible ways of performing a two-way decomposition, simply divide the residual (interaction term) in half and allocate to each of the decomposition components.

Annotated Examples

Here is an example of the sectoral decomposition of poverty changes using the Uganda sample data with notes explaining the command line specification and the output. The data sets and variables are defined in the Sample Data section.

Make sure that the sedecomposition command (ado file) is installed on your copy of Stata; see the Installation of Stata ado files section for details.

Example 1

. cd C:\yourdir\yoursubdir [substitute the location where you placed the sample data files]
. use ugahh92, clear [Note: “ugahh92.dta” is the data for the initial year of the period]
. sedecomposition using ugahh02 [aweight=iwe], sector(urban) var1(welfare) var2(welfare) pline1(spline) pline2(spline)

-------------------------------------------------------------------------------------------------------------------------------------------
Sectoral Decomposition of a Change in Poverty: HeadCount
-------------------------------------------------------------------------------------------------------------------------------------------

To allow you to replicate and understand the results in the table above, the population shares and poverty rates for the rural and urban sectors for each of the years are shown in the table below.

Recall that the three components of this sectoral decomposition are the following:

a) Headcount poverty index in 1992.

b) Headcount poverty index in 2002.

c) Rural intra-sectoral effect: If the rural population share had remained constant at 87.58% but rural poverty rates had changed from 1992-2002, rural poverty would have decreased by 15.44 percentage points.

d) Urban intra-sectoral effect: If urban poverty rates changed but the urban population share had remained constant at 12.42% from 1992-2002, urban poverty would have decreased by 1.79 percentage points.

e) Total intra-sectoral effect: This is the sum of the individual intra-sectoral effects in lines (c) and (d),

f) Population-shift effect:

This is also referred to as the inter-sectoral effect and represents the change in poverty that would have occurred if poverty rates had remained constant at initial levels but population shares had changed. In this case with only two sectors, the difference in shares will be of the same magnitude but have opposite signs. Thus, a negative population-shift effect indicates a shrinking population share in the sector with the higher initial poverty rate – the rural sector. However, this component is not a major factor as it only accounts for 2.4% of the total change in poverty.

g) Interaction effect: This term accounts for the population shifts and poverty changes together. This size of this effect is negligible in this case.

h) Change in poverty: This the actual change in poverty and is equal to the sum of (e)-(g).

Example 2

We can also decompose changes in the poverty gap or poverty gap squared by specifying “pg” or “pgs” at the end of the command.

. sedecomposition using ugahh02 [aweight=iwe], sector(urban) var1(welfare) var2(welfare) pline1(spline) pline2(spline) pg

-------------------------------------------------------------------------------------------------------------------------------------------
Sectoral Decomposition of a Change in Poverty: Poverty Gap
-------------------------------------------------------------------------------------------------------------------------------------------

Example 3

We can also decompose changes in poverty by, say, administrative regions or sectors of employment. The example below shows the decomposition of the poverty gap by region.

. sedecomposition using ugahh02 [aweight=iwe], sector(region) var1(welfare) var2(welfare) pline1(spline) pline2(spline) pg

-------------------------------------------------------------------------------------------------------------------------------------------
Sectoral Decomposition of a Change in Poverty: Poverty Gap
-------------------------------------------------------------------------------------------------------------------------------------------

References / Related Papers

Ravallion, M. and M. Huppi (1991). "Measuring Changes in Poverty: A Methodological Case Study of Indonesia during an Adjustment Period." World Bank Economic Review, 5, pp. 57-82.

Chatti, R. and A. El Lahga (2008). "On the Contribution of Sectoral Natural Population Growth to the Aggregate Poverty Change" Journal of Population Economics 21(1): pp. 183-190.

(Link above is for SpringerLink subscribers. Other versions are available: Cahiers de recherche PMMA No. 2007-14) 

Suryahadi, A., D. Suryadarma, and S. Sumarto. "The Effects of Location and Sectoral Components of Economic Growth on Poverty: Evidence from Indonesia," Journal of Development Economics, 89(1): 109-117.

(Link above is for Science Direct subscribers.)

Sectoral elasticities of poverty

Ravallion, M. and G. Datt (2002). "Why has Economic Growth been more Pro-Poor in Some States of India than Others?" Journal of Development Economics, 68(2), pp. 381-400.

(Link above is for Science Direct subscribers. Other versions are available: World Bank Policy Research Working Paper # 2663) 

Growth-Inequality Decomposition | Sectoral Poverty Decomposition |
Growth Incidence Curve | Rate of Pro-Poor Growth | Growth Elasticity of Poverty |
Installation of Stata ado files | Sample data

Back to Measuring the Growth-Poverty Link