measuring and monitoring poverty for the MDGs

1
measuring and monitoring poverty for the MDGs

• estimating extreme poverty
• for
• international and national purposes
• Tim Conway, World Bank Cambodia Country Office
• UNSD / ESCAP Workshop on MDG Monitoring
• Bangkok, 14-16 January 2009

2
outline

• defining and measuring poverty a very brief
  recap
• measurement, description, explanation, and
  targeting
• national poverty estimates
• Choice of welfare measure
• Process of setting the poverty line
• the international poverty line (IPL)
• IDTs, MDGs and dollar-a-day
• the International Comparison Project 2005 and the
  new 2008 IPL
• summary of differences and similarities between
  old and new estimates
• current issues
• remaining issues with international poverty
  estimates
• which poverty line to use for what purposes?
• basic principles
• further reading
• sources on the IPL, IPC and global poverty
  measures

3
why measure poverty?
1 concept to measurement

• poverty reduction a critical goal of almost all
  modern states, particularly in the developing
  world
• and, since 2000, an international commitment (MDG
  1.1)
• reliable, consistent poverty measures support
  policy effectiveness and accountability by
  enabling
• analysis of the causes of poverty and formulation
  of appropriate policies to tackle these causes
• targeting of limited resources to where poverty
  is highest
• monitoring poverty trends to assess if policies
  are working
• evaluating the impact of policies, programmes or
  shocks

4
defining and measuring poverty is as much an art
as a science
1 concept to measurement

• In sum, there is no ideal measure of
  well-being. The implication is simple all
  measures of poverty are imperfect. This is not
  an argument for avoiding poverty measurement, but
  rather for approaching all measures of poverty
  with a degree of caution, and for asking in some
  detail how the measures are constructed.
• Introduction to poverty analysis (World Bank 2005
  p. 40, emphasis added)

5
measuring poverty 3 steps
1 concept to measurement

• choosing a welfare measure
• i.e. income or consumption
• defining a poverty line
• setting a minimum level of the chosen welfare
  measure, below which an individual is said to
  be poor
• collecting welfare data through a survey and
  comparing it to the poverty line to obtain
  estimates of
• what percentage of the population falls below the
  poverty line
• (the poverty headcount or poverty incidence)
• and by how much
• (the poverty gap and the squared poverty gap)

6
a short history of national poverty estimates
2 national poverty estimates

• the first household welfare surveys
• household budget surveys in England in late 1700s
  
• and in Germany in early 1800s
• Charles Booth credited with the first poverty
  survey (London 1889)
• colonial governments conducted surveys but
  sampling analysis often poor
• development of modern sampling theory in 1920s
  and 1930s
• national surveys and poverty monitoring since
  World War II
• Indian National Sample Survey (NSS) established
  in 1950s
• Foster-Greer-Thorbecke (FGT) poverty indices
  (1984)
• spread of PCs in late 1970s/early 1980s allowed
  rapid expansion of survey programmes (e.g. LSMS,
  DHS etc.)

7
means and ends functional relationships between
different welfare / poverty measures
2 national poverty estimates
Housing quality and consumer durables (Stock of
past investments)
Happiness, quality of life
market
savings
Human development outcomes (health, education
etc.)
current consumption value of housing and consumer
durables
Privately-purchased goods and services
Total household current consumption (starting
with food)
income
productive assets
Publicly-provided goods and services
Consumption of household-produced goods and
services
Household economic activity / production
state
8
choice of welfare measure variesexamples from
Asia
2 national poverty estimates
countries
welfare measure
per capita consumption
Afghanistan, Bangladesh, Bhutan, Cambodia, Lao
PDR, Thailand
per capita income
China, Philippines
household income
Malaysia
9
setting a national poverty line
2 national poverty estimates

• start by setting the food poverty line
• estimating food consumption required to obtain
  minimum healthy nutrition (typically 2,100
  calories per person per day)
• typically using a survey to look at the food
  consumption bundle of the quintile around the
  poverty line for composition
• Then estimate a minimum consumption level
  required to meet basic non-food needs
• clothing, housing, health care, education
• various alternative approaches to this
• Add the food and non-food allowances to obtain a
  total poverty line or, simply, the poverty line
• NB poverty lines often calculated separately for
  regions of the country with very different
  consumption bundles / price levels

10
collecting data generating poverty
measuressummary of some key survey design
issues and tradeoffs
2 national poverty estimates

• policy and political - timetable
• preparation of national development / poverty
  reduction strategies
• election cycle
• coverage
• spatial (changing coverage between rounds makes
  it hard to identify trend)
• thematic (very easy to overload a survey
  instrument)
• frequency required
• scale of disaggregation required
• consistency and comparability
• cost and available resources
• complexity, national capacity, and ownership
• include a panel component?

11
origins of the international poverty line, late
1990s
3 international poverty estimates

• International community sought to define a list
  of global development priorities
• first the OECD-DAC International Development
  Targets (IDTs)
• and then, at the Millennium General Assembly in
  2000, the MDGs
• consolidated from targets agreed at UN
  conferences during the 1990s, most targets were
  for human development
• an indicator for absolute poverty was set on the
  basis of what, in local currency unit
  equivalents, a dollar could buy in the US in
  1990, i.e.
• US1 per capita per day, 1990 purchasing power
  parity (PPP)
• this had been used in World Development Report
  (WDR) 1990 on poverty
• chosen because among 33 countries for which WDR
  1990 had survey data, poverty lines for the
  poorest ranged from 0.75-1 pc per day, 1985 PPP
• this became the international poverty line (IPL)
• and the basis for MDG 1 to cut, by 2015, the
  percentage of the worlds population living in
  absolute poverty to half its 1990 level

12
over time, estimates of global poverty have
changed
3 international poverty estimates

• This is due to
• changing definition of the IPL (in US per capita
  per day)
• 1990 1 1985 PPP
• 2000 1.08 1993 PPP
• 2008 1.25 2005 PPP
• reflecting
• updated (and better) data in relative prices
• new household surveys (which have increased
  coverage of the population of the developing
  world)
• plus some methodological refinements
• However, the basic principles and steps remain
  the same

13
revising the international poverty line
3 international poverty estimates

• 1998
• UN Evaluation of the ICP (the Ryten Report)
  identified concerns in ICP 1993-6
• 2000
• 2000/01 World Development Report on Poverty (WDR
  2000/01) compared to WDR 1990,
• drew on 1993 ICP price data
• and a larger sample of household surveys
• modified the IPL slightly to 1.08 per capita per
  day, 1993 PPP
• but the dollar-a-day label stuck
• 2002-7
• Deaton (2002)criticised World Bank (2002) for
  mixing methodologically inconsistent estimates
  from different sources to come up with a
  misleading picture of trend
• Chen Ravallion (2004) attempted to address
  these concerns
• And updated this analysis with new household
  surveys, bringing estimates up to 2004 (Chen
  Ravallion 2007)
• 2008
• new and better PPPs for more countries from the
  2005 ICP
• found that cost of living in developing countries
  higher than had previously been thought
• new IPL set at 1.25 pc per day 2005 PPP

14
calculating international poverty estimates-
basic principles, illustrated with new 2008
estimates
3 international poverty estimates

• Define a representative international poverty
  line in a reference currency (US )
• Rank countries by real mean per capita
  consumption
• Convert mean national per capita consumption
  values into using purchasing power parity (PPP)
  exchange rates (in 2008, using PPPs from the 2005
  round of the ICP)
• Take the poorest countries in this list (in 2008,
  the poorest 15 countries)
• convert their national poverty lines into using
  PPPs
• And average these the mean of these standardised
  poverty lines from the poorest countries (1.27
  pc per day) is taken as the definition of
  absolute poverty
• Convert the IPL (in 2008, 1.25 pc per day) into
  all local currency units (LCU)
• Again, using PPPs from the most recent ICP data
  year (in 2008, using 2005 PPPs)
• Deflate this LCU poverty line value to the
  year(s) for which have survey data
• using the official national CPI data for each
  country
• NB weights in this index may or may not match
  consumer budget shares at the poverty line
• Use this poverty line to generate poverty
  estimates for that year
• Where necessary, use interpolation to generate
  poverty estimates for reference years in the IPL
  reporting series (i.e. every three years from
  1981 to 2005)
• Use pc growth rates from national accounts data
• projected poverty estimates back from the survey
  immediately after the reference year
• and project forward from the survey immediately
  before it
• Obtain the mean of these two projected values
  take this as the value for that year

15
three key improvements in the 2008 revisions
3 international poverty estimates

• better data on relative prices from ICP 2005
• updated poverty line to reflect new PPPs
• better data on household consumption

16
better data on relative prices
3 international poverty estimates

• The 2005 round of the ICP improved considerably
  on 1993
• covered more countries
• 1968 (first) ICP 10
• 1993 ICP 117
• 2005 ICP 146 including China for the first
  time
• more detailed descriptions of goods and services
  to ensure stricter comparison, and larger number
  of data points for each item in each country
• new methods to measure and price difficult cases
• government compensation housing and public
  sector productivity
• more ring comparisons (18) to link regional PPP
  estimates to global prices
• Improvements reflected
• Adoption of recommendations from the Rysten
  Report
• Better funding

17
updated international poverty linewhy 1.25?
3 international poverty estimates

• Ravallion, Chen and Sangraula (2007) created a
  dataset of 74 countries for which it is possible
  to plot both
• national poverty lines and
• the log of mean per capita consumption
• (both converted into 2005 US using the new
  PPPs)
• The 15 poorest countries that is, the 15 with
  lowest per capita consumption had a mean
  poverty line (in 2005 PPP terms) of 1.27 per
  capita per day
• This line is taken as a good definition of
  absolute poverty

18
better data on household consumption
3 international poverty estimates

• Expanding dataset of countries and surveys
  between poverty estimation exercises
• And expanding number of total surveys, so more
  survey years (data points) for each country
• total 675 surveys
• As a result, more recent estimates are able to
  draw on increasing recent coverage of the
  population of the developing world

19
of population represented by a survey conducted
2 years before or after the reference date
3 international poverty estimates
Source Chen and Ravallion 2008 p. 30
20
How do the new changes affect findings?new
estimates suggest poverty higher than previously
thought,but global and regional poverty trends
essentially the same
3 international poverty estimates

• Using the 2008 revisions
• estimated global poverty level in 2005 rises from
  1bn to 1.4bn
• but the trend that results is largely the same as
  under the old estimates
• applying the new PPPs and poverty line to earlier
  years raises poverty estimates for those years,
  too

21
trend using the new line is similar to that using
the oldfrom 1981 to 2005
3 international poverty estimates

• global poverty incidence has halved (from 52 to
  26)
• driven largely by rapid falls in east Asia
• and to a lesser degree south Asia
• while poverty incidence is unchanged in
  sub-Saharan Africa

22
remaining issues with the IPL
4 current issues

• urban bias in the ICP price data
• especially in China also 11 other countries
  partially in India
• use rural-urban differentials in national poverty
  lines to calibrate these cases
• relevance of ICP consumption bundles (items and
  weights) to the poor a case for PPPs for the
  poor?
• ADB (2008) suggests such PPPs would reduce
  poverty estimates in Asia
• for alternative conclusions, see Deaton and
  Dupriez (2008)
• using national CPI to deflate poverty from survey
  years to IPL series years
• especially over periods of relative price shifts
• data issues
• a hole in the Indian data reflecting
  methodology changes in NSS 1999
• more broadly, differences in national survey
  design (e.g. recall period recall cf diary
  number of items in consumption list) and
  implementation can create significant differences
  in how consumption is recorded

23
which poverty line to use?
4 current issues

• Note that national and international estimates
  draw on the same data
• i.e. data on pc consumption or income, derived
  from a household survey
• what differs is where the poverty line is drawn
  through this distribution
• The international poverty line is a valuable tool
  for
• Comparing levels of absolute poverty between
  countries
• Aggregating national headcounts to track poverty
  at regional and global levels
• Providing standardised national poverty estimates
  which can be used in cross-country econometric
  analysis to obtain insights into (for example)
  the relationships between growth, equality and
  poverty reduction
• However, it is a fairly blunt tool
• increasingly sophisticated (greater coverage of
  countries in the ICP larger number of household
  surveys efforts to construct poverty-specific
  PPPs)
• but comparisons over time or between any two
  countries will still be imprecise
• And sometimes not credible
• For monitoring and analysing poverty at the
  national level, almost always better to use a
  national poverty line

24
capacity and ownership of poverty estimates
4 current issues

• ownership issues apply to both and national
  poverty lines
• defining a poverty line and calculating welfare
  measures are very complex
• as much an art as a science
• choices are informed by accepted general
  principles and knowledge of international good
  practice
• but there is no blueprint, no single, correct
  method
• the statistical skills required to set a poverty
  line and estimate the welfare measure may be in
  very short supply
• particularly in low income and / or post-conflict
  countries,

25
capacity, ownership and donor practice
4 current issues

• ownership and capacity problems may be
  exacerbated by donors
• urgent need for data leads to heavy reliance upon
  expatiate experts
• inconsistency in support responsibility for the
  survey and / or analysis may pass between donors
  and / or from one consultant to another
• Often very poor archiving of data and
  documentation on methodology
• timing of surveys is driven by supply (donors)
  rather than domestic demand
• Often affects not just analysis
• how data is processed to estimate poverty
• but even what data is collected and with what
  quality
• variations in survey design and implementation
• Sometimes, this results in two or more
  alternative national poverty estimates in
  addition to international poverty line estimates
• Particularly problematic in post-conflict
  countries
• choices made in the first, baseline survey can
  constrain later trend analysis

26
challenges in interpreting alternative poverty
estimates estimates for Cambodia currently in
circulation
4 current issues
27
monitoring poverty some basic principles
4 current issues

• If you want to measure change, dont change your
  measure
• consistency is critical to analysing trend or
  monitoring policy impact
• periodically updating lines national or
  international will be necessary
• but should be infrequent, and done carefully
• changes in survey design or poverty methodology
  should be a matter of inclusive debate (technical
  choices have political consequences)
• and carefully documented
• Not everything that can be counted counts not
  everything that counts can be counted (Albert
  Einstein)
• national poverty measures need to reflect a
  social and political consensus
• need to triangulate poverty line measures with
  other sources, including qualitative
• I would rather be roughly right than precisely
  wrong. (JM Keynes)
• look at standard errors and sensitivity
• focus on the big questions (is poverty going down
  or up? Which groups are poorest?)
• and choose the combination of measures that best
  answers these

28
2008
5 further reading

• Poverty data a supplement to World Development
  Indicators 2008. in World Bank 2008
  (forthcoming) 2008 World Development Indicators
• UNDP International Policy Center (ICP)
  (http//www.undp-povertycentre.org/theme.do)
• several short One-Pager notes under the
  Poverty theme
• No. 69 Where the line is drawn a rejoiner to
  Ravallion. Thomas Pogge October 2008
• No. 66 Global poverty reassessed a reply to
  Reddy. Martin Ravallion September 2008
• No. 65 The new global poverty estimates
  digging deeper into a hole. Sanjay Reddy
  September 2008
• No. 54 A consistent measure of real poverty a
  reply to Ravallion. Thomas Pogge May 2008
• No. 53 Which poverty line? a response to Reddy.
  Martin Ravallion May 2008
• The bottom 1.4 billion The Economist 28th
  August 2008.
• World Bank updates estimates poverty estimates
  for the developing world http//go.worldbank.org/C
  9GR27WRJ0
• - World Bank page, November 3rd 2008 with summary
  of new estimates and links to various materials
• The developing world is poorer than we thought,
  but no less successful in the fight against
  poverty. Shaohua Chen and Martin Ravallion
  World Bank Policy Research Working Paper 4703
  August 2008 http//go.worldbank.org/5V41Z1WRL0
• Briefing notes in English, Chinese, Arabic,
  French, Spanish and Russian
• Press release August 26th
• PovcalNet online Poverty Aanalysis Tool
  http//go.worldbank.org/NT2A1XUWP0

29
1998-2007
5 further reading

• Absolute poverty measures for the developing
  world, 1981-2004. Shaohua Chen and Martin
  Ravallion March 2007 World Bank.
• http//mdgs.un.org/unsd/mdg/Resources/Attach/Capa
  city/Chen2020Ravallion20(2007).pdf
• How have the worlds poorest fared since the
  early 1980s? Shaohua Chen and Martin Ravallion
  fall 2004 World Bank Research Observer.
• http//mdgs.un.org/unsd/mdg/Resources/Attach/Capa
  city/Chen2020Ravallion20(WBRO202004).pdf
• Purchasing power parity for international
  comparison of poverty sources and methods.
  Sultan Ahmad 2003 World Bank.
  siteresources.worldbank.org/ICPINT/Resources/pover
  ty_PPP_Data_rev2.doc
• Counting the worlds poor problems and possible
  solutions. Angus Deaton fall 2001 World Bank
  Research Observer Vol. 16 No. 2
  http//imagebank.worldbank.org/servlet/WDSContentS
  erver/IW3P/IB/2002/01/18/000094946_02010804013046/
  Rendered/PDF/multi0page.pdf
• World Development Report (WDR) 2000/01 attacking
  poverty. 2001 World Bank OUP.
  http//go.worldbank.org/L8RGH3WLI0
• Evaluation of the International Comparison
  Project (the Ryten Report). UN 1998.
  http//unstats.un.org/unsd/methods/icp/8-e.pdf

30

• extra / alternative slides

31
calculating estimates against the IPL
3 international poverty estimates

• Convert the IPL (1.25 pc per day) into the
  national currency in 2005 using PPPs from the new
  ICP
• Adjust this value to the year(s) for which have
  survey data for that country using official
  national CPI data
• Note the weights in this index may or may not
  match well with consumer budget shares at the
  poverty line
• Use this poverty line to generate poverty
  measures for that year

32
poverty-specific PPPs?
3 international poverty estimates

• Two aspects of the problem
• PPPs are based on prices for consumption items in
  all countries in the ICP so, developing country
  prices influenced by OECD consumption baskets
• the poor and non-poor may face different prices
  for the same level of consumption within the same
  country
• Deaton and Dupriez (2008) suggest PPPs for the
  poor may not have much effect on estimates
• However, ADB (2008) has attempted to construct
  PPPs specific to the goods and service consumed
  by the poor for a large set of Asian countries
  (although not China)
• And found on balance that these would suggest a
  lower poverty rate in Asia than would the World
  Banks 1.25 pc per day, 2005 PPP estimates
• Ongoing research agenda

33
challenges in interpreting alternative poverty
estimates estimates for Cambodia currently in
circulation
4 current issues