Transcript International Comparison Program (ICP)

PPPs for Regional and Global Poverty
Measurement: Current Methodology,
Issues and Options
D.S. Prasada Rao
School of Economics
University of Queensland
Brisbane, Australia
Outline
Role of ICP-PPPs in poverty measurement:
Issues
Price level differences across income
groups
Poverty PPPs: options for future ICP rounds
7/20/2015
2
Two approaches
Derive a global income distribution and estimate
poverty incidence through the proportion under a
specified poverty line
Specify an international poverty line (eg $ 1 a day)
and count the number of poor in each country
under this poverty line – the World Bank approach
Note: Both approaches need PPPs for converting
national aggregates into a common currency unit
7/20/2015
3
PPPs and Exchange Rates
(Selected Developing countries)
PPPs
Exchange Rates
1988
4.756
1993
6.997
1988
13.919
1993
30.493
Relative Price
Levels
1988
1993
0.342
0.229
China
(Urban)
Indonesia
1.038
1.414
3.72
5.762
0.299
0.245
453.5
626.1
1685.7
2087.1
0.269
0.300
Bangladesh
8.822
9.496
31.733
39.567
0.278
0.240
Country
India
Source: Milanovic (2002)
7/20/2015
4
Income Per Capita by Regions
(in US dollars)
Region
Using Exchange
Rates
1988
1993
Using PPPs
1988
1993
Africa
619
673
1320
1757
Asia
1422
2007
1927
2972
E. Europe
1889
1194
6355
4522
Latin America
1967
3027
4829
5923
W. Europe/N.Am
16255
20485
14773
19913
World
3649
4537
4442
5643
Source: Milanovic (2002)
7/20/2015
5
World Inequality
(Gini Coefficient)
Converter
1988
1993
PPP
0.628
0.660
Exchange Rate
0.782
0.805
Source: Milanovic (2002)
7/20/2015
6
PPPs and Global Poverty
The World Bank Approach
Regular estimates of global and regional poverty
Uses $1-a day and $2- a day as international
poverty lines
The $-poverty line is converted into local currency
using PPPs from the ICP
Estimate poverty incidence using national
accounts and HES/LSMS data
Approach since the 1990 World Development
Report
7/20/2015
7
Advantages
A simple approach that ensures inter-country
comparability
 Country practices are diverse
 Need for comparability from an international
perspective in the light of MDGs
$1-a-day poverty line approximates existing
country poverty lines (especially in developing
countries) quite well
Concept is easy to explain and create awareness
among the international community
World Bank estimates of poverty incidence are
widely used
7/20/2015
8
Which conversion factors to use?
Exchange rates
PPPs from the ICP for the GDP as a whole
– used in real per capita income
comparisons
PPPs for consumption
PPPs for food and beverages
7/20/2015
9
Nature of price surveys
Location of price surveys
 Urban/cities/rural
 Past ICP price surveys mainly located in
metropolitan areas
Outlets for price surveys
 Make sure the service component of outlets is
comparable
Annual average prices and issues of seasonality
Asterisk approach to the problem of
representativity
7/20/2015
10
Expenditure weights
National averages are used
Main sources: national accounts/household
expenditure surveys
Using average weights in the presence of skewed
distributions
Average shares may differ significantly from the
budget shares of the poor
Relative importance of food and non-food
items
Share of services
7/20/2015
11
Aggregation issues
PPPs within ICP are computed using:
 Transitive methods (comparisons between a
pair of countries are influenced by price data
from other countries)
 EKS and Geary-Khamis methods
Biases induced by aggregation methods and real
income comparisons
Interpretation of PPPs: Do they refer to purchasing
powers associated with a fixed bundle of goods
and services?
7/20/2015
12
Extrapolations to non-participating
countries and non-benchmark years
 Low participation in ICP benchmarks in the past and infrequent
benchmarks
 Extrapolations to non-benchmark countries are based on regression
methods
 PWT Methodology (versions 5.6 and 6.1)
 World Bank Methodology
 Extrapolations to non-benchmark years
 Stability of regressions over time
 Inconsistency between benchmarks
 Standard errors for extrapolated PPPs
 Big changes in poverty estimates (and regional composition) when
new ICP round PPPs are used.
7/20/2015
13
PPPs for Poor
Need to de-link issues of international poverty
lines and PPPs for poor
 Conceptual problems may arise (vacuousness!)
 PPPs for poor should be applicable to different
poverty lines ($1- or $2-a-day)
Need to recognise the circular nature of the work
 Need to identify the poor before PPPs
PPPs for poor for inter-country comparisons are
likely to be more difficult due to the influences of
local and regional influences on the consumption
of the poor
7/20/2015
14
PPPs for Poor: Strategy
Identify a basket of goods and services that can be
considered as a poverty basket.
Examine the feasibility and find sources of price
data associated with the poverty basket.
Find expenditure patterns of the poor.
Use a suitable methodology to combine data to
derive PPPs for the poor.
7/20/2015
15
Research efforts to date
Main focus has been to examine feasible
alternatives to a more complete poverty-specific
PPP compilation.
Is it possible to estimate price level differences
across income groups using household
expenditure survey data?
Do poor pay higher prices?
Feasibility of computing PPPs for poor from
household expenditure surveys.
7/20/2015
16
Data used
Unit Record Data (household data)
Rural and urban
Provinces
Unit values from survey data
Focus on food items
Methods
Average price and quantity for each
income group/region
Individual price and quantity data
7/20/2015
17
Index Numbers for 3 Income Groups
Using Household Expenditure Data
and unit values
Ethiopia
Uganda
Bottom 30%
1.00
1.00
Middle 60%
1.015
1.279
Top 10%
1.029
1.672
Rao (2003)
7/20/2015
18
Brazilian Study
Price Indexes by
Income Groups
Low
1.000
Middle
1.025
High
1.091
Source: Aten and Menezies (2002)
7/20/2015
19
Indian Study
Household Group
Rural India
Urban India
South
East
West
South
East
West
All Households
0.990
1.095
0.996
0.931
1.041
0.961
Households above
poverty line
0.974
1.084
0.998
0.925
1.043
0.952
Households below
poverty line
1.082
1.101
1.092
1.058
0.994
1.009
Source: Coondoo et al. (2003)
7/20/2015
20
India-Indonesia Comparisons
Conversion Rates
(1 Rupee = Number of Rupiah)
Exchange Rate:
182.42
Penn World Tables:
165.20
World Bank PPP:
140.00
Deaton Poverty PPP:
Rural prices
238.20
Rural +Urban
235.00 (approx)
Source: Deaton et al. (2004)
7/20/2015
21
PPPs based on poverty-specific price data
Scope and coverage
 Basically replicate the current ICP work but restrict to the
consumption module
 Identify poverty-specific item lists




Need to use regional and sub-regional lists
Lists may be based on the ICP product lists
Use sizes similar to typical purchases of the poor
May need input of poverty researchers from regions
 Price surveys
 Rural and urban surveys
 Focus on outlets typically used by the poor (general markets)
 Integrate these surveys with the ICP price surveys
 Expenditure weights
 Use HES/LSMS and estimates of national poverty incidence estimates to
identify the patterns of the poor
7/20/2015
22
Income group price differences:
Basic Findings
 Possible to derive adjustment factors for the PPPs derived
from the ICP
 Results did not show that prices paid by poorer households
are higher than those paid by richer households
 Adjustment for quality differences may alter this conclusion
 Results derived appear to be sensitive to the aggregation
methodology used
 CPD and its variants provide more consistent results
 Results do not appear to be sensitive the use of aggregated
data by income groups
7/20/2015
23
Option 1: Reweight Basic Heading
PPPs
 Compute basic heading PPPs from regular ICP price
surveys
 Derive PPPs for poor using expenditure patterns of the
poor
 HES/LSMS may not refer to the ICP benchmark
 Price data:
 Item lists may not be representative of the consumption of the poor
 Outlets surveyed
 Size of purchases of the poor
 This approach addresses only one of the current criticisms
7/20/2015
24
Option 2: Adjust ICP PPPs
 Based on the premise that current ICP PPPs refer to the prices paid by
the upper income households
 Need to measures differences in levels of prices paid by higher and
lower income groups
 HES/LSMS are the principal source of price and weights data
 Problems





7/20/2015
Unit values from HES refer to broadly specified items
No measure of quality differences
HES data have details of food consumption but not for non-food items
Price imputations for consumption from own production
HES/LSMS may not correspond to the ICP benchmarks
25
Option 3: Use Price Data from Loose
specifications using SPD approach
Structured Production Descriptions (SPD)
approach is under consideration for the next round
ICP
SPD is a description that lists important
characteristics of products in a narrow cluster.
Section 1:
 Name of product
 Product cluster
 Basic heading
7/20/2015
26
SPDs - continued
 Section 2: Establishment/Outlet (characteristics)
 Section 3:
 Size/ no. of units
 Imported/domestically produced
 Section 4: Various characteristics (brand name, etc.)
 Tight specification (target specifications - close to the
matched product approach)
 Loose specification
 Increased coverage – use of CPI lists
 Methods: Hedonic and CPD approaches
7/20/2015
27
Option 4: Use poverty-specific price
data
 Basically replicate the current ICP work but restrict to the
consumption module
 Identify poverty-specific item lists




Need to use regional and sub-regional lists
Lists may be based on the CPI product lists
Use product sizes similar to typical purchases of the poor
Need the input of poverty researchers from regions
 Price surveys
 Rural and urban surveys
 Focus on outlets used by the poor (general markets)
 Integrate these surveys with CPI price surveys
 Expenditure weights
 Use HES/LSMS and estimates of national poverty incidence estimates to
identify the patterns of the poor
 Linking regional PPPs
 Ring country approach may be more difficult
 May need spatial linking based on similarities in consumption baskets
7/20/2015
28
Outstanding Analytical issues
Provide a conceptual framework for the use
of $1 or $2 a day approach.
Treatment of inherent circularity in the
process.
Linking regional poverty PPPs with the US
dollar.
Updating $1-a-day poverty line over time.
7/20/2015
29
Example
In 2005:
1US$ = 55 Pesos
PPP
Suppose prices double in both US and the
Philippines over the period 2001 to 2015
Suppose real incomes are unchanged in both
countries - no change in poverty status
In 2015: 1US$ = 50 Pesos PPP
What about poverty in the Philippines according
to $1-a-day poverty line?
 Poverty reduced by 50% in the Philippines????
 Meets MDG No.1!!!
7/20/2015
30
Benefits
 ADB and the World Bank: Estimates of poverty incidence
based on sound methodology
 Less volatile estimates of poverty incidence
 Standard errors for the estimates
 ICP: Enthusiastic participation from countries
 Increased coverage of developing countries
 Possibility to integrate ICP work with the work of NSOs
 Participating countries:




7/20/2015
Statistical capacity building (improved HES etc)
Regional price comparisons
CPIs for the poor
Structure for poverty monitoring (useful in the context of MDGs
and PRSPs)
31