Poverty Measurement

1
Poverty Measurement

• Prepared for African Development Bank Training
  Course on the Economics of Poverty

Hammamet, Tunisia, May 23-25, 2005
Jonathan Haughton Suffolk University,
Boston jonathan.haughton_at_suffolk.edu
2
Goals Objectives

• Understand poverty analysis
• Not the same as actually doing it. Know whats
  possible the pitfalls how to write TOR.
• Outline
• What is poverty?
• Why measure it?
• What welfare indicator?
• What poverty line?
• What summary measure?
• Robustness
• Poverty profiles what and why?
• Determinants of poverty
• Poverty reduction policies
• International poverty comparisons

3
What is poverty?

• Pronounced deprivation in well-being.
• une importante privation de bien-être
• Command over commodities. Monetary
• Specific poverty e.g. food poor, house poor
• Sen Capability to function in society. Low
  capability if
• Low income (monetary poverty)
• Low education (specific)
• Poor health (specific)
• Insecurity/vulnerability (insurance needed)
• Low self esteem (empowerment needed)
• Lack of rights (e.g. freedom of speech)
• Here, poverty is multi-dimensional.
• Note All are quite closely correlated.

4
AfDB on poverty

• poverty is defined as a condition where the poor
  have insufficient funds and political power to
  maintain an acceptable standard of living.
• Circular!
• Followed by mention of need to walk far to fetch
  water, wood.
• and then a reference to basic needs.
• The broader the definition, the less the focus,
  and the harder to ensure accountability.
• Source African Development Bank/African
  Development Fund, Bank Group Policy on Poverty
  Reduction, February 2004, p.4.

5
Why measure poverty?

• Not obvious and its expensive.
• Keep poor on the agenda (Ravallion)
• Target interventions (e.g. geographically)
• cibler les interventions
• Monitor evaluate projects and policies (e.g.
  microcredit)
• Evaluate the AfDB and World Bank!
• Re 1. and 2.
• See role of PRSP (Poverty Reduction Strategy
  Paper) process
• but stronger on description than diagnosis or
  action overhyped?

6
How measure poverty?

• 3 steps
• Define indicator of welfare (e.g. income,
  expenditure, malnutrition)
• indicateur de bien-être
• Set poverty line
• seuil de pauvreté
• Summarize the information.

7
Survey Data

• Poverty measures are based on survey data.
• i.e. its a sample statistic
• Implication should give intervals, not point
  estimates.
• Issues
• Sample frame (plan de sondage)
• Often under-represents homeless, unregistered
• Cross section or panel
• Stratified cluster sampling is common
• sondage aléatoire stratifié et par grappe
• Cheaper, but less precise
• Requires use of weights
• What level of detail in questions?
• E.g. Tobacco in Vietnam

8
Best practice LSMS

• Living Standards Measurement Surveys
• Multi-topic questionnaires
• Both income expenditures, useful
• Anthropometric measures useful
• Multiple questionnaires
• Quality control
• But expensive small samples (lt5,000) less
  frequent complex.
• Good ref Grosh and Glewwe opus.

9
What welfare indicator?

• Income Consumption ?net wealth
• Haig Simons
• Revenu consommation changement de la valeur
  nette
• Income
• Understated in LDCs (forget fear tax man
  illegal income hard to measure livestock)
• Expenditure
• Also understated (beer, tobacco)
• How deal with durables? Housing? Festivals?
• But closer to permanent income

10
(No Transcript)
11
Adult equivalents? échelle déquivalence

• Commonly use expenditure/capita
• But
• Needs differ, by age, gender, activity, etc.
• Economies of scale in consumption
• Solution adult equivalent scale
• E.g. OECD AE 1 0.7(NA-1) 0.5NC
• Or AE (NA a NC)?
• e.g. AE (NA 0.7 NC)0.8
• Nice idea rarely used. What weights?

12
Other welfare indicators

• Calories. 2,100?
• Food/expenditure (Engel found a robust
  relationship)
• Nutritional status (e.g. height for age z-scores)
• For community
• Life expectancy Infant Mortality Rate school
  enrollments

13
Calorie Equivalence Scale for East Africa
14
Does choice of welfare measure matter?
Maybe, overall yes, if identifying poor
churning.
15
What poverty line? (Difficult!)

• Relative (e.g. bottom quintile) popular in rich
  countries (except US)
• But doesnt allow one to track poverty over time
• Doesnt allow international comparisons
• Doesnt work with monitoring evaluation
• Yet line must be socially appropriate
• Absolute (adjusted for inflation)
• Cost of basic needs
• Food energy intake
• Subjective

16
Cost of basic needs

• Cost of 2,100 (or so) Calories
• What price per Calorie?
• Gives food poverty line
• Add a non-food component
• US multiply food by 3
• Vietnam non-food of those in 2000-2200 Calorie
  bracket
• Adjust for prices over regions, time
• Getting appropriate price data can be hard
• Often use CPI, but coverage limited
• How adjust for different consumption baskets
  (e.g. Indonesia)?

17
Example Cambodia poverty lines
18
Example Vietnam
19
Practice in Africa varies widely
20
Subjective poverty lines

• E.g. Social weather stations, Philippines
• What income level do you personally consider to
  be absolutely minimal? That is to say that with
  less you could not make ends meet.
• Datt
• Self-rated poverty lines are high (60 vs.
  official 25)
• Self-rated pov. line has risen rapidly over time,
  so no trend in poverty rate, despite economic
  growth
• Line given by poor is only slightly below line
  given by rich
• Urban line 2x rural line (vs. price differential
  of 1.2-1.5)
• In short not very satisfactory.

21
What poverty measure?

• Headcount lindice numérique P0 (1/N) ?
  I(yiltz)
• Clear, easy to understand
• Ignores depth of poverty
• Assumes equal distribution w/in hhold.
• Poverty gap lindice de lécart de pauvreté P1
  (1/N) ? Gi/z
• Sum of gaps is min. cost of elim. poverty
• P1 is ratio of min. to max. cost of elim. Poverty
• Poverty severity lindice de lintensité de la
  pauvreté
• P2 (1/N) ?(Gi/z)2
• Sen-Shorrocks-Thon PSST P0P1P(1GP)
• Decomposable US 0.12, Canada 0.06. Both about
  0.10 in 1970.
• It doesnt matter too much I like P0 and PSST.

22
(No Transcript)
23
e.g. Madagascar poverty
24
Robustness fiabilité

• Sampling error erreurs de sondage
• Good practice to report it bootstrapping.
• Measurement error
• 5 underestimation of expenditure
• 10 overstatement of P0
• Equivalence scales
• Doesnt matter much
• Choice of poverty line
• Use stochastic dominance tests to see if it
  matters
• Often helpful to report two lines.

25
Poverty Profiles profils de pauvreté

• Sets out major facts on poverty (who are the
  poor?)
• By geography
• By community characteristics
• By household characteristics
• By individual characteristics
• Graphs and tables
• See World Bank checklist for ideas
• The starting point of Poverty Reduction Strategy
  Papers

26
e.g. Cambodia graph
Headcount index conditional probabilities of
being poor Share measure contribution to
poverty
27
e.g. Malawi table
28
World Bank Checklist, 1-10

• Does poverty vary widely between different areas
  in the country?
• Are the most populated areas also the areas where
  most of the poor live?
• How is income poverty correlated with gender,
  age, urban and rural, racial, or ethnic
  characteristics?
• What are the main sources of income for the poor?
• On what sectors do the poor depend for their
  livelihood?
• What products or servicestradables and
  non-tradablesdo the poor sell? A tradable good
  is one that is, or easily might be, imported or
  exported. The prices of such goods are
  influenced by changes in the world price and the
  exchange rate.
• To what extent are the rural poor engaged in
  agriculture? In off-farm employment?
• How large a factor is unemployment?
  Underemployment?
• Which are the important goods in the consumption
  basket of the poor? How high is the share of
  tradables and non-tradables?
• How is income poverty linked to malnutrition or
  educational outcomes?

29
World Bank Checklist, 11-19

• What are fertility characteristics of the poor?
• To what public services do the poor have access?
  What is the quality of these services?
• How important are private costs of education and
  health for the poor?
• Can the poor access formal or informal credit
  markets?
• What assetsland, housing, and financialdo the
  poor own? Do property rights over such assets
  exist?
• How secure is their access to, and tenure over,
  natural resources?
• Is environmental degradation linked to poverty?
• How variable are the incomes of the poor? What
  risks do they face?
• Are certain population groups in society at a
  higher risk of being poor than others are?
  Households that are at a high risk of being poor,
  but are not necessarily poor now, are considered
  to be vulnerable.

30
e.g. Cambodia, education
31
e.g. Indonesia, sectors
32
Determinants of poverty

• Goal Make sense of the poverty profile/analytics
• Identify variables of relevance
• Participatory Poverty Assessment (PPA) can be
  very helpful here
• They should be exogenous
• Theory/economics is helpful here
• Use regression analysis to determine relative
  importance of variables.
• Usually multiple regression sometimes nonlinear
  occasionally logit or probit.
• The challenge moving from analysis to action
  (Whites missing middle)

33
(No Transcript)
34
(No Transcript)
35
Proximate vs. Deep Causes

• These are the immediate correlates of poverty
  are they causes?
• The search for deep causes continues
• E.g. In some countries female-headed households
  are poorer. Gender per se, or nature of social
  relations?
• E.g. If less poverty in urban areas, why dont
  more people move?
• We still lack a general theory of poverty (and
  perhaps always will)

36
Poverty Reduction Policies

• Growth is good for the poor
• Dollar Kraay 236 episodes, 80 countries, 4
  decades. Elasticity of 1 and R2 of 0.87.
• How grow? End war macro stability then?
• The national ownership inconsistency
• So why write poverty into every project?
• Can one do more? Maybe
• Opportunity
• via credit, schools, health
• Empowerment linsertion Hard to define
• via information, participation/inclusion,
  accountability
• Income security
• via insurance, disaster response, transfers, land
  tenure

37
International Poverty Comparisons

• Two levels of measurement
• Comparing country averages
• GDP/capita or similar
• Composite indexes such as HDI
• Comparing poverty from country to country

38
How poor are poor countries?

• Usually use GDP/capita PIB/capita
• GNI/capita better, but harder to get
• E.g. Ireland, Lesotho, Saudi Arabia
• Only compares country averages
• Imperfectly correlated with wellbeing
• The marry your housekeeper problem
• Environment
• Security, stress
• Resource exhaustion
• Need to use PPP (purchasing power parity)
• Can use social indicators
• IMR life expectancy literacy education access
  to water, sewers.
• Data quality often poor (e.g. how measure
  literacy?)
• MDGs 48, mainly outcome indicators. Loss of
  focus?

39
On PPP
40
Human Development Index

• UNDP invention
• Put emphasis on dimensions of poverty other than
  monetary
• Caveat not comparable over time
• HDIj (1/3)(Life expectancy 25)/(85-25)
• (1/3)(ln(GDPPPP /cap)-ln(100)/(ln(400
  00)-ln(100))
• (1/3)(2/3)(adult literacy 0)/(100
  0)
• (1/3)(school enrolment
  0)/(100 0)
• Obscurantist highly correlated with GDP/capita

41
Computing HDI

• Kenya (HDR 2000)

• HDIj (1/3)(51.3 25)/(85-25)
• (1/3)(ln(980)-ln(100)/(ln(40000)-ln(1
  00))
• (1/3)(2/3)(80.5 0)/(100 0)
• (1/3)(50 0)/(100 0)
  0.508

42
The dollar a day standard

• Actually 1.08/day in 1993 PPP, equivalent to
  about 1.34/day now
• To get local-currency value of poverty line
• Use PPP e.r. to get local value of 1.08 in 1993
• Update using local CPI inflation
• Problems
• PPP e.r. and CPI may be poorly correlated
• Some PPP e.r.s are inaccurate imputed
  econometrically, not measured directly
• PPP refers to all goods services, not just
  those of poor yet is it hard to find comparable
  poverty line baskets across countries
• Better to use food poverty measure?

43
(No Transcript)
44
Is world poverty falling?

• Bhalla Argues MDGs on poverty already met!
• Take distribution consumption from survey data
• Gross up to be consistent with national accounts
• But
• Survey data systematically under-represents high
  incomes expenditures (non-response bias
  under-reporting bias)
• Bhallas approach assumes elasticity of low
  expenditure w.r.t. total expenditure to be 1 OK
  perhaps in long-run, not in short-run
• By any standard, Africa is lagging in poverty
  reduction

45
Targeting the poor (1)Cibler les pauvres

• Means testing (by income or expenditure)
• E.g. US. Usually not feasible in LDCs.
• Indirect indicator targeting (e.g. by size of
  landholding, children, etc.)
• Anh Vietnam, indicators (mosquito net?
  Blankets?)
• Sahn asset index based on DHS data (education,
  housing quality, utensils)
• African poverty falling, early 1990s
• Ravallion skeptical (1993 Jamaica study on rapid
  appraisal questions)
• Self-targeting programs
• Maharashtra food for work but hard to design

46
Falling poverty in Africa?
47
Targeting the poor (2)

• Subsidize food mainly eaten by the poor (e.g.
  rough bread in Egypt)
• Invest in pro-poor technology
• Geographic targeting
• Works because spatial poverty is persistent
• Clear criteria
• Relatively easy to implement (e.g. community
  development grants)

48
Implementing Geographic targeting

• How identify poor areas to target?
• Household surveys too small to do this
• Use survey data to explain poverty using a few
  variables in a regression
• Apply regression to census data, to create a
  detailed poverty map
• Done in Vietnam, Ecuador, not Africa (yet)
• How accurate, especially at fine level of
  aggregation?
• NB Vietnam uses local authorities to identify
  poor households
• Ravallion finds limited cost effectiveness at
  regional level in Indonesia

49
Concluding thoughts

• Measurement of poverty is not trivial
• Implication Ask for details about data sources,
  assumptions made, etc.
• Proximate causes of poverty are relatively
  straightforward to identify
• Poverty profiles do this well
• The analysis of deep causes is harder
• Keep an eye on the academic literature, coupled
  with the commonsense of practitioners evaluate.
• There are many policy instruments, deserving of
  more attention in Africa
• Taxation
• Government spending (health, education, etc.)
• Targeting of various types