SlideShare a Scribd company logo
1 of 30
Download to read offline
Poverty Mapping
An overview of methods,
based on a Malawi analysis
Todd Benson
International Food Policy Research Institute
June 2009 [t.benson@cgiar.org]
2
Poverty Mapping
 Small-area estimation poverty mapping involves:
1. Discovering relationships between household
characteristics and the welfare level of households.
 Through analysis of a detailed household survey.
2. Applying model of these relationships to data on the
same household characteristics contained in a
national census.
 Predict the welfare level of all households in census.
 Resulting estimates of aggregate welfare and poverty are
spatially disaggregated to a much higher degree than is
possible using survey.
3
Poverty headcount estimates from
survey analysis & poverty mapping
Individual poverty
headcount
by district
Blantyre
City
Zomba
Munic.
Lilongwe
City
Mzuzu
under 55%
55 to 75%
above 75%
National poverty
headcount: 65.3%
Poverty
mapping,
local gov’t
ward esti-
mates (848)
Poverty
analysis of
HH survey,
district esti-
mates (29)
4
Malawi – Poverty monitoring
 Will use Malawi examples here.
 IFPRI provided technical support to Malawi Poverty
Monitoring System, 1999-2002. Key technical
products were:
 Poverty analysis of 1997/98 Malawi Integrated Household
Survey (IHS).
 Determinants of poverty analysis.
 Poverty map – based on IHS and 1998 Population and
Housing Census – focus here
 Detailed national spatial data sets developed.
 Malawi – An atlas of social statistics
5
Poverty mapping –
simple overview of analysis (1)
 Daily per capita household consumption &
expenditure for survey households is modeled on set
of household and local (cluster) characteristics.
 Per capita household consumption & expenditure is
our household welfare indicator and is used with a
basic-needs poverty line to determine whether
household is poor or non-poor.
 Developed through poverty analysis of household survey.
 Independent household variables chosen for model
are only those that also appear in national census.
 Also add cluster (spatial) characteristics – these need to be
available for entire country.
6
Poverty mapping –
simple overview of analysis (2)
 Having developed model(s) of household welfare
based on household characteristics from survey, use
those same characteristics from all households in
census to estimate welfare of all households in
population using the model(s).
 Analysis draws on:
 Rich data on welfare and household characteristics, but
coarse representativity (region or district-level, at best) of
household survey.
 Fine, comprehensive coverage, but limited HH character-
istics and no welfare information in national census.
 Survey and census should have been implemented
at same time, or within a few years of each other.
7
 Elbers, C.; J. Lanjouw; & P. Lanjouw. 2003. Micro-level estimation
of poverty & inequality. Econometrica. 71: 355-364.
 Greater detail found in Elbers, Lanjouw, & Lanjouw. 2002.
Micro-Level Estimation of Welfare. Research Working
Paper 2911. World Bank, Development Research Group.
 Since 2001, method has been applied in several dozen
countries around the globe.
 Malawi IHS-derived models applied to census data used the
Poverty and Inequality Mapper module (SAS-based program).
 This program since updated with PovMap 2.0. See
http://iresearch.worldbank.org/PovMap/PovMap2/PovMap2Main.asp.
 Other software tools needed are a statistical program (Stata was
used for Malawi work) and, ideally, GIS.
Small-area estimation
poverty mapping method
8
Malawi poverty map - data
 1997-98 Malawi Integrated Household Survey (IHS-1)
 Representative at level of district & urban center.
 6,586 survey households.
 IHS used for national poverty analysis and
production of a poverty profile of Malawi in 2000.
 Detailed information on consumption & expenditure,
as well as range of other HH characteristics.
 1998 Population & Housing Census.
 2.25 million households
 Limited range of variables.
 Basic demography, education, employment, housing.
9
Poverty mapping –
Summary steps in analysis
1. Data preparation.
 The same variables from the two data sets –
census and IHS – are subject to means
comparison tests.
 To make sure that they really are the same variable –
similar definition and measurement.
 Sometimes referred to as the ‘Stage 0’ step.
 Independent variables for the model are those
variables which are found in both the household
survey and the census.
 Possible that survey may not contain all census variables.
 For example, in case of Malawi, survey did not contain the
housing quality variables in the census.
10
Summary steps (cont.)
2. Initial models developed from IHS household survey data using
a backward stepwise regression procedure.
 Dependent variable: natural log (ln) of household welfare indicator
 household welfare indicator: Daily per capita consumption and
expenditure of IHS survey household expressed in spatially-
deflated April 1998 Malawi Kwacha.
 ln of welfare indicator has more normal distribution than that of real
indicator, so preferred for model development.
 Aim is to develop models with greatest predictive power.
• While do want model coefficients to make theoretical sense, this is
not fundamental to analysis.
• Hence, stepwise regression used.
 Candidate HH independent variables on next slide.
 Also did a close examination of potentially problematic cases in the
household survey data.
 Used a range of regression diagnostic statistics for this.
 Dropped 63 out of 6,586 cases in IHS dataset as a result.
11
Candidate household
variables
HHSIZSQ squared household size
AGEHHH age of head of household HOUSRNT hh rents house in which it lives (0/1)
AGESQ squared age of hh head KIDBORN tot. kids ever born to fertile women in hh
BIKE hh owns a bicycle (0/1) KIDDEAD prop. kids born in hh now dead
BRTHRTE yrs. between births for women in hh M0_05 males aged 0 to 5
COOKEL hh cooks over electricity or gas (0/1) M15_29 males aged 15 to 29
COOKWD hh cooks over firewood (0/1) M30_49 males aged 30 to 49
EMPLYEE head of household employee (0/1) M50UP males aged 50 and up
EMPLYR head of household employer (0/1) M6_14 males aged 6 to 14
F0_05 females aged 0 to 5 MMAXCL maximum class attained by males
F15_29 females aged 15 to 29 NETENRL primary age children in primary school
F30_49 females aged 30 to 49 NOMARRY household head not married (0/1)
F50UP females aged 50 and up NONFAM members who are not of nuclear family
F6_14 females aged 6 to 14 OTHROCC hh mem. w/ other occupation
FAMBU hh has a family business (0/1) PRFNLIT hh gets lighting from paraffin (0/1)
FEMALE females in household PRIMAGE primary age child in hh (0/1)
FEMHHH female headed household (0/1) PROF hh mem. w/ prof, admin, clerical occup.
FINPRIM total members finished primary SALES hh members with sales occupation
FMAXCL maximum class attained by females SECIND hh members in secondary industry
FRTWMN woman in hh in fertile years (15-45) SERVICE hh members with services occupation
HHHED educational level of head of hh in years TAPH2O hh gets water from tap (0/1)
HHSIZE household size TERIND hh members in tertiary industry
12
Summary steps (cont.)
3. Enumeration area (cluster) variables added to
each model to improve predictive power and to
deal with some econometric problems.
 Candidate enumeration area variables.
 Census variable means for EA.
 Some distance and agricultural production variables
developed using Geographic Information System (GIS).
 Stepwise regression is again used to select for
inclusion in model.
13
GIS distance variable
generation
 Straight-line
distance to
market center
 left – market
centers
 center –
distance to
market
centers
 right – EA
centerpoints
on distance
layer (data
extraction)
14
Candidate enumeration area
variables
AVGMXED Average maximum education level
in households in EA
MARKET Straight-line distance (km) to
nearest market centre
BOMA Straight-line distance (km) to
nearest boma (district HQ)
MN_YLD Mean maize yield in EA over 20
years
DIFF Difference maize yield in EA in '97
& '98 from long term mean
POPDENS Population density (persons/sq.
km.)
EAIMPTL Proportion of households in EA
with improved toilets
ROAD Straight-line distance (km) to
nearest primary or secondary road
EANTENR Net enrollment rate in EA ROOMPC Avg. rooms per person in EA
EAPRMHS Proportion of hhs in EA with
houses of permanent materials
URBAN Straight-line distance (km) to
nearest urban centre
HELTHFA Straight-line distance (km) to
nearest health facility
15
Summary steps (cont.)
 23 separate models developed.
 One for each of 22 IHS analytical strata, plus stratum of
enumeration areas which, although rural, are urban in
character.
 Relatively common set of independent
variables seen across models
 Rural models: age-sex group variables, HH head education,
highest ed. attainment by female, bicycle ownership, female
headship, and more.
 Urban models: fewer age-sex group variables, HH head
education, cook on wood, EA variables on sanitation,
education, rooms per capita.
16
Summary steps (cont.)
4. Assessment made of resultant models made up of
both household and EA variables.
 Final models developed at this step.
5. The error in each base household model is then
modeled in order to control for heteroskedasticity.
6. IHS-derived models then applied to the census data.
 Using special program for poverty mapping.
 Bootstrap method used to generate:
 Point and variance estimates of welfare indicator for a range of
spatial groupings of households.
 Poverty measures, with standard errors.
 Inequality measures, with standard errors.
17
Performance of the poverty
mapping models
 23 models - Developing so many models is definitely NOT
‘best practice’.
 Adjusted-R2 for base models range from 0.248 to
0.594, with mean of 0.380.
 Urban models have higher R2s.
 Assessment:
 The 11 models in Mozambique poverty maps had R2s
ranging from 0.26 to 0.55.
 Other countries with sharp economic divides that are defined
on racial lines (or other household and individual
characteristics) tend to have higher R2s.
• South Africa (around 0.70), Ecuador (0.50).
18
Poverty map validation –
national & regional
 Poverty headcount results compared between IHS
poverty analysis & poverty mapping analysis.
 Differences at national and regional levels are not
statistically significant.
IHS PovMap
National 65.3% 64.3%
Regional
Southern 68.1% 65.4%
Central 62.8% 63.9%
Northern 62.5% 61.1%
19
Poverty map validation – district
20
 District level
 In comparing the IHS poverty analysis to the
poverty mapping results, the headcounts for seven
out of the 31 districts are significantly different.
 Blantyre rural & Lilongwe rural are the most important of
these.
 In the other five districts, lack of congruence may be due to
problems with the IHS data or the IHS sampling scheme.
 In spite of the problems in these seven districts,
results are reasonable, overall.
 Following slides show results mapped at the more
local scale of the sub-district Traditional Authority.
Poverty map validation – district
(cont.)
21
Depth & severity of poverty –
Traditional Authority
22
Local government ward
 Poverty mapping method used can provide
estimates of poverty for populations of 500 to 1000
households and above.
 Census enumeration areas (EA) are too small,
being 250 households on average.
 Only intermediate spatial unit between TA and EA is
the local government ward.
 Following map shows poverty headcount for the
almost 850 local government wards.
 Such small-area maps potentially of value for sub-
district planning.
23
Poverty
headcount –
Local
government
ward
24
 Poverty estimates for
populations in each
aggregated EA.
 Poverty measures are
estimated with standard
errors.
 Mean s.e. for poverty
head-count for rural
aggregated EAs of 8.2
percent.
 In examining results of
analysis, the consider-
able imprecision in
underlying estimates
should be kept in
mind.
Error in poverty maps
25
Additional analyses based on
poverty map
 Assessment of how well
programs target the poor
 Determine poverty
characteristics of
population in program
areas and compare to
wider poverty prevalence.
 Ex-post analysis of
poverty targeting.
 Malawi Social Action
Fund public works
locations & WFP Food
for Assets & Develop-
ment project sites.
 See IFPRI’s FCND
discussion paper
no. 205
26
Geographic scale and poverty
mapping – even more local
 Constructed a ‘new’ geography for Malawi based
on amalgamations of two or three EAs.
 Subdivided the country into about 3,400 spatial units
with populations above 500 households in order to
calculate local welfare and poverty measures for
population in each.
 Primarily for analytical purposes, rather than for
planning.
 Used in the investigation of the spatial determinants
of poverty prevalence in rural Malawi.
27
Creating the Malawi poverty map
– Institutional considerations
 Poverty map creation was an activity of the Poverty
Monitoring System of the government of Malawi.
 Three principal institutions:
 National Statistical Office (NSO)
 National Economic Council (NEC - now Ministry of Economic
Planning & Development)
 Centre for Social Research (CSR) of the University of Malawi
 IFPRI provided technical assistance on poverty
analysis to all three.
 Poverty map was logical extension of the poverty
analysis work that NSO and NEC jointly led.
 However, NEC did not engage in actual analysis.
28
Creating the Malawi poverty map
– Institutional considerations (cont.)
 Poverty mapping work done by two statisticians from NSO, led by
IFPRI researcher.
 In retrospect, this was not ideal.
 Poverty mapping, to do confidently, requires relatively sophisticated
econometric understanding & abilities.
 Such abilities not best placed within a national statistical office.
 Their principal activities do not include such high-level analyses.
 Rather, should have involved university-based Malawian
econometrician(s), as well.
 Technical abilities required, retention & refinement of poverty mapping
skills, and analytical leadership for future poverty mapping efforts
nationally, best served by universities.
 So, Malawi case was a missed opportunity for sustainable capacity
strengthening in technical analysis for poverty mapping.
29
Creating the Malawi poverty map
– Data considerations
 Had contemporaneous IHS and census data sets.
 However, this temporal correspondence not
absolutely critical.
 Key concern is that nature of correlation between
household welfare and independent variables used in
poverty mapping model for each stratum will not have
changed significantly between the two time periods.
 Judgment call that takes into account the dynamism and
character of economic transformation across country.
30
Creating the Malawi poverty map
– Data considerations (cont.)
 Had two separate GIS shapefiles of enumeration areas. For both:
 IHS, sample design of which was based on EAs for 1987 Malawi census, and
 1998 Census.
 Facilitated generation of GIS-based cluster-level variables for inclusion in
stratum models and then for use in applying models to census data.
 Also permitted extensive spatial analysis of poverty mapping results.
 However, having EA digital maps not absolutely critical.
 Can use smaller-scale, higher geographical units, such as districts, for
generating cluster variables.
 GIS-derived, census-aggregates, or developed from other geographically-
comprehensive ancillary data.
 Loss of information since any intra-district variation in the data is lost.
 Econometrically also not ideal, as EA-level cluster variables in models assist in
dealing issues arising due to expected correlation of characteristics of survey
households in same EA.
 Results in some loss in explanatory power and performance of models, but an
acceptable solution.

More Related Content

Similar to Povertymappingmalawitoddbenson 091211011812-phpapp01

Poverty Mapping, An overview of methods,based on a Malawi analysis
Poverty Mapping, An overview of methods,based on a Malawi analysisPoverty Mapping, An overview of methods,based on a Malawi analysis
Poverty Mapping, An overview of methods,based on a Malawi analysisguest9970726
 
Poverty Mapping: An overview of methods, based on a Malawi
Poverty Mapping: An overview of methods, based on a MalawiPoverty Mapping: An overview of methods, based on a Malawi
Poverty Mapping: An overview of methods, based on a Malawiessp2
 
PRESENTATION URBAN POVERTY.pptx
PRESENTATION URBAN POVERTY.pptxPRESENTATION URBAN POVERTY.pptx
PRESENTATION URBAN POVERTY.pptxchristian aja
 
Csilla FILO: A rural area of social processes present in the Sellye region, i...
Csilla FILO: A rural area of social processes present in the Sellye region, i...Csilla FILO: A rural area of social processes present in the Sellye region, i...
Csilla FILO: A rural area of social processes present in the Sellye region, i...Territorial Intelligence
 
The mexican family life survey within the surveys
The mexican family life survey within the surveysThe mexican family life survey within the surveys
The mexican family life survey within the surveysUNDP Policy Centre
 
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. Dawas
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. DawasIAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. Dawas
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. DawasStatsCommunications
 
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...essp2
 
Multidimensional Poverty For Monitoring Development Progress
Multidimensional Poverty For Monitoring Development ProgressMultidimensional Poverty For Monitoring Development Progress
Multidimensional Poverty For Monitoring Development ProgressUNDP Eurasia
 
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...IFPRIMaSSP
 
Equity in the Basic Education Opportunities in Egypt
Equity in the Basic Education Opportunities in EgyptEquity in the Basic Education Opportunities in Egypt
Equity in the Basic Education Opportunities in EgyptEman Refaat
 
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...Nabraj Lama
 
DLHS III - Dr. Suraj Chawla
DLHS III - Dr. Suraj ChawlaDLHS III - Dr. Suraj Chawla
DLHS III - Dr. Suraj ChawlaSuraj Chawla
 
Socioeconomic staus and Social Security Measures
Socioeconomic staus and Social Security MeasuresSocioeconomic staus and Social Security Measures
Socioeconomic staus and Social Security MeasuresDrAnup Kumar
 
Building a Better National Targeting System for Improving Social Safety Net P...
Building a Better National Targeting System for Improving Social Safety Net P...Building a Better National Targeting System for Improving Social Safety Net P...
Building a Better National Targeting System for Improving Social Safety Net P...Paul Mithun
 
Alex zscheile project
Alex zscheile projectAlex zscheile project
Alex zscheile projectajzscheile
 
!!Power point TTP Butajira.pptx
!!Power point TTP Butajira.pptx!!Power point TTP Butajira.pptx
!!Power point TTP Butajira.pptxGetahunAlega
 

Similar to Povertymappingmalawitoddbenson 091211011812-phpapp01 (20)

Poverty Mapping, An overview of methods,based on a Malawi analysis
Poverty Mapping, An overview of methods,based on a Malawi analysisPoverty Mapping, An overview of methods,based on a Malawi analysis
Poverty Mapping, An overview of methods,based on a Malawi analysis
 
Poverty Mapping: An overview of methods, based on a Malawi
Poverty Mapping: An overview of methods, based on a MalawiPoverty Mapping: An overview of methods, based on a Malawi
Poverty Mapping: An overview of methods, based on a Malawi
 
PRESENTATION URBAN POVERTY.pptx
PRESENTATION URBAN POVERTY.pptxPRESENTATION URBAN POVERTY.pptx
PRESENTATION URBAN POVERTY.pptx
 
The Donor Footprint and Gender Gaps
The Donor Footprint and Gender GapsThe Donor Footprint and Gender Gaps
The Donor Footprint and Gender Gaps
 
Csilla FILO: A rural area of social processes present in the Sellye region, i...
Csilla FILO: A rural area of social processes present in the Sellye region, i...Csilla FILO: A rural area of social processes present in the Sellye region, i...
Csilla FILO: A rural area of social processes present in the Sellye region, i...
 
The mexican family life survey within the surveys
The mexican family life survey within the surveysThe mexican family life survey within the surveys
The mexican family life survey within the surveys
 
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. Dawas
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. DawasIAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. Dawas
IAOS 2018 - Global Multidimensional Poverty Index in Jordan, M. Dawas
 
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is There a "Girl Effect"? Expe...
 
Multidimensional Poverty For Monitoring Development Progress
Multidimensional Poverty For Monitoring Development ProgressMultidimensional Poverty For Monitoring Development Progress
Multidimensional Poverty For Monitoring Development Progress
 
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...
Parental Aspirations for Children's Education: Is there a "Girl Effect"? Expe...
 
Equity in the Basic Education Opportunities in Egypt
Equity in the Basic Education Opportunities in EgyptEquity in the Basic Education Opportunities in Egypt
Equity in the Basic Education Opportunities in Egypt
 
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...
Possible Decision Rules to Allocate Quotas and Reservations to Ensure Equity ...
 
DLHS III - Dr. Suraj Chawla
DLHS III - Dr. Suraj ChawlaDLHS III - Dr. Suraj Chawla
DLHS III - Dr. Suraj Chawla
 
Ray mcrhrd trg 31.05.2011
Ray mcrhrd trg 31.05.2011Ray mcrhrd trg 31.05.2011
Ray mcrhrd trg 31.05.2011
 
Community satisfaction on UHEP
Community satisfaction on UHEPCommunity satisfaction on UHEP
Community satisfaction on UHEP
 
A study of the value of local bus services to society
A study of the value of local bus services to societyA study of the value of local bus services to society
A study of the value of local bus services to society
 
Socioeconomic staus and Social Security Measures
Socioeconomic staus and Social Security MeasuresSocioeconomic staus and Social Security Measures
Socioeconomic staus and Social Security Measures
 
Building a Better National Targeting System for Improving Social Safety Net P...
Building a Better National Targeting System for Improving Social Safety Net P...Building a Better National Targeting System for Improving Social Safety Net P...
Building a Better National Targeting System for Improving Social Safety Net P...
 
Alex zscheile project
Alex zscheile projectAlex zscheile project
Alex zscheile project
 
!!Power point TTP Butajira.pptx
!!Power point TTP Butajira.pptx!!Power point TTP Butajira.pptx
!!Power point TTP Butajira.pptx
 

Recently uploaded

Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...
Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...
Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...ThinkInnovation
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一F sss
 
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一F La
 
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Serviceranjana rawat
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改atducpo
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130Suhani Kapoor
 
{Pooja: 9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...
{Pooja:  9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...{Pooja:  9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...
{Pooja: 9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...Pooja Nehwal
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...Florian Roscheck
 
RA-11058_IRR-COMPRESS Do 198 series of 1998
RA-11058_IRR-COMPRESS Do 198 series of 1998RA-11058_IRR-COMPRESS Do 198 series of 1998
RA-11058_IRR-COMPRESS Do 198 series of 1998YohFuh
 
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Callshivangimorya083
 
Schema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfSchema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfLars Albertsson
 
GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]📊 Markus Baersch
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFAAndrei Kaleshka
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort servicejennyeacort
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...soniya singh
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)jennyeacort
 
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAmazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAbdelrhman abooda
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubaihf8803863
 

Recently uploaded (20)

Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...
Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...
Predictive Analysis - Using Insight-informed Data to Determine Factors Drivin...
 
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
办理学位证中佛罗里达大学毕业证,UCF成绩单原版一比一
 
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一
办理(Vancouver毕业证书)加拿大温哥华岛大学毕业证成绩单原版一比一
 
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service
(PARI) Call Girls Wanowrie ( 7001035870 ) HI-Fi Pune Escorts Service
 
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
代办国外大学文凭《原版美国UCLA文凭证书》加州大学洛杉矶分校毕业证制作成绩单修改
 
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
VIP Call Girls Service Miyapur Hyderabad Call +91-8250192130
 
{Pooja: 9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...
{Pooja:  9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...{Pooja:  9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...
{Pooja: 9892124323 } Call Girl in Mumbai | Jas Kaur Rate 4500 Free Hotel Del...
 
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...From idea to production in a day – Leveraging Azure ML and Streamlit to build...
From idea to production in a day – Leveraging Azure ML and Streamlit to build...
 
RA-11058_IRR-COMPRESS Do 198 series of 1998
RA-11058_IRR-COMPRESS Do 198 series of 1998RA-11058_IRR-COMPRESS Do 198 series of 1998
RA-11058_IRR-COMPRESS Do 198 series of 1998
 
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
꧁❤ Greater Noida Call Girls Delhi ❤꧂ 9711199171 ☎️ Hard And Sexy Vip Call
 
Schema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdfSchema on read is obsolete. Welcome metaprogramming..pdf
Schema on read is obsolete. Welcome metaprogramming..pdf
 
GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]GA4 Without Cookies [Measure Camp AMS]
GA4 Without Cookies [Measure Camp AMS]
 
How we prevented account sharing with MFA
How we prevented account sharing with MFAHow we prevented account sharing with MFA
How we prevented account sharing with MFA
 
Call Girls in Saket 99530🔝 56974 Escort Service
Call Girls in Saket 99530🔝 56974 Escort ServiceCall Girls in Saket 99530🔝 56974 Escort Service
Call Girls in Saket 99530🔝 56974 Escort Service
 
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
9711147426✨Call In girls Gurgaon Sector 31. SCO 25 escort service
 
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
High Class Call Girls Noida Sector 39 Aarushi 🔝8264348440🔝 Independent Escort...
 
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
Call Us ➥97111√47426🤳Call Girls in Aerocity (Delhi NCR)
 
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptxAmazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
Amazon TQM (2) Amazon TQM (2)Amazon TQM (2).pptx
 
E-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptxE-Commerce Order PredictionShraddha Kamble.pptx
E-Commerce Order PredictionShraddha Kamble.pptx
 
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls DubaiDubai Call Girls Wifey O52&786472 Call Girls Dubai
Dubai Call Girls Wifey O52&786472 Call Girls Dubai
 

Povertymappingmalawitoddbenson 091211011812-phpapp01

  • 1. Poverty Mapping An overview of methods, based on a Malawi analysis Todd Benson International Food Policy Research Institute June 2009 [t.benson@cgiar.org]
  • 2. 2 Poverty Mapping  Small-area estimation poverty mapping involves: 1. Discovering relationships between household characteristics and the welfare level of households.  Through analysis of a detailed household survey. 2. Applying model of these relationships to data on the same household characteristics contained in a national census.  Predict the welfare level of all households in census.  Resulting estimates of aggregate welfare and poverty are spatially disaggregated to a much higher degree than is possible using survey.
  • 3. 3 Poverty headcount estimates from survey analysis & poverty mapping Individual poverty headcount by district Blantyre City Zomba Munic. Lilongwe City Mzuzu under 55% 55 to 75% above 75% National poverty headcount: 65.3% Poverty mapping, local gov’t ward esti- mates (848) Poverty analysis of HH survey, district esti- mates (29)
  • 4. 4 Malawi – Poverty monitoring  Will use Malawi examples here.  IFPRI provided technical support to Malawi Poverty Monitoring System, 1999-2002. Key technical products were:  Poverty analysis of 1997/98 Malawi Integrated Household Survey (IHS).  Determinants of poverty analysis.  Poverty map – based on IHS and 1998 Population and Housing Census – focus here  Detailed national spatial data sets developed.  Malawi – An atlas of social statistics
  • 5. 5 Poverty mapping – simple overview of analysis (1)  Daily per capita household consumption & expenditure for survey households is modeled on set of household and local (cluster) characteristics.  Per capita household consumption & expenditure is our household welfare indicator and is used with a basic-needs poverty line to determine whether household is poor or non-poor.  Developed through poverty analysis of household survey.  Independent household variables chosen for model are only those that also appear in national census.  Also add cluster (spatial) characteristics – these need to be available for entire country.
  • 6. 6 Poverty mapping – simple overview of analysis (2)  Having developed model(s) of household welfare based on household characteristics from survey, use those same characteristics from all households in census to estimate welfare of all households in population using the model(s).  Analysis draws on:  Rich data on welfare and household characteristics, but coarse representativity (region or district-level, at best) of household survey.  Fine, comprehensive coverage, but limited HH character- istics and no welfare information in national census.  Survey and census should have been implemented at same time, or within a few years of each other.
  • 7. 7  Elbers, C.; J. Lanjouw; & P. Lanjouw. 2003. Micro-level estimation of poverty & inequality. Econometrica. 71: 355-364.  Greater detail found in Elbers, Lanjouw, & Lanjouw. 2002. Micro-Level Estimation of Welfare. Research Working Paper 2911. World Bank, Development Research Group.  Since 2001, method has been applied in several dozen countries around the globe.  Malawi IHS-derived models applied to census data used the Poverty and Inequality Mapper module (SAS-based program).  This program since updated with PovMap 2.0. See http://iresearch.worldbank.org/PovMap/PovMap2/PovMap2Main.asp.  Other software tools needed are a statistical program (Stata was used for Malawi work) and, ideally, GIS. Small-area estimation poverty mapping method
  • 8. 8 Malawi poverty map - data  1997-98 Malawi Integrated Household Survey (IHS-1)  Representative at level of district & urban center.  6,586 survey households.  IHS used for national poverty analysis and production of a poverty profile of Malawi in 2000.  Detailed information on consumption & expenditure, as well as range of other HH characteristics.  1998 Population & Housing Census.  2.25 million households  Limited range of variables.  Basic demography, education, employment, housing.
  • 9. 9 Poverty mapping – Summary steps in analysis 1. Data preparation.  The same variables from the two data sets – census and IHS – are subject to means comparison tests.  To make sure that they really are the same variable – similar definition and measurement.  Sometimes referred to as the ‘Stage 0’ step.  Independent variables for the model are those variables which are found in both the household survey and the census.  Possible that survey may not contain all census variables.  For example, in case of Malawi, survey did not contain the housing quality variables in the census.
  • 10. 10 Summary steps (cont.) 2. Initial models developed from IHS household survey data using a backward stepwise regression procedure.  Dependent variable: natural log (ln) of household welfare indicator  household welfare indicator: Daily per capita consumption and expenditure of IHS survey household expressed in spatially- deflated April 1998 Malawi Kwacha.  ln of welfare indicator has more normal distribution than that of real indicator, so preferred for model development.  Aim is to develop models with greatest predictive power. • While do want model coefficients to make theoretical sense, this is not fundamental to analysis. • Hence, stepwise regression used.  Candidate HH independent variables on next slide.  Also did a close examination of potentially problematic cases in the household survey data.  Used a range of regression diagnostic statistics for this.  Dropped 63 out of 6,586 cases in IHS dataset as a result.
  • 11. 11 Candidate household variables HHSIZSQ squared household size AGEHHH age of head of household HOUSRNT hh rents house in which it lives (0/1) AGESQ squared age of hh head KIDBORN tot. kids ever born to fertile women in hh BIKE hh owns a bicycle (0/1) KIDDEAD prop. kids born in hh now dead BRTHRTE yrs. between births for women in hh M0_05 males aged 0 to 5 COOKEL hh cooks over electricity or gas (0/1) M15_29 males aged 15 to 29 COOKWD hh cooks over firewood (0/1) M30_49 males aged 30 to 49 EMPLYEE head of household employee (0/1) M50UP males aged 50 and up EMPLYR head of household employer (0/1) M6_14 males aged 6 to 14 F0_05 females aged 0 to 5 MMAXCL maximum class attained by males F15_29 females aged 15 to 29 NETENRL primary age children in primary school F30_49 females aged 30 to 49 NOMARRY household head not married (0/1) F50UP females aged 50 and up NONFAM members who are not of nuclear family F6_14 females aged 6 to 14 OTHROCC hh mem. w/ other occupation FAMBU hh has a family business (0/1) PRFNLIT hh gets lighting from paraffin (0/1) FEMALE females in household PRIMAGE primary age child in hh (0/1) FEMHHH female headed household (0/1) PROF hh mem. w/ prof, admin, clerical occup. FINPRIM total members finished primary SALES hh members with sales occupation FMAXCL maximum class attained by females SECIND hh members in secondary industry FRTWMN woman in hh in fertile years (15-45) SERVICE hh members with services occupation HHHED educational level of head of hh in years TAPH2O hh gets water from tap (0/1) HHSIZE household size TERIND hh members in tertiary industry
  • 12. 12 Summary steps (cont.) 3. Enumeration area (cluster) variables added to each model to improve predictive power and to deal with some econometric problems.  Candidate enumeration area variables.  Census variable means for EA.  Some distance and agricultural production variables developed using Geographic Information System (GIS).  Stepwise regression is again used to select for inclusion in model.
  • 13. 13 GIS distance variable generation  Straight-line distance to market center  left – market centers  center – distance to market centers  right – EA centerpoints on distance layer (data extraction)
  • 14. 14 Candidate enumeration area variables AVGMXED Average maximum education level in households in EA MARKET Straight-line distance (km) to nearest market centre BOMA Straight-line distance (km) to nearest boma (district HQ) MN_YLD Mean maize yield in EA over 20 years DIFF Difference maize yield in EA in '97 & '98 from long term mean POPDENS Population density (persons/sq. km.) EAIMPTL Proportion of households in EA with improved toilets ROAD Straight-line distance (km) to nearest primary or secondary road EANTENR Net enrollment rate in EA ROOMPC Avg. rooms per person in EA EAPRMHS Proportion of hhs in EA with houses of permanent materials URBAN Straight-line distance (km) to nearest urban centre HELTHFA Straight-line distance (km) to nearest health facility
  • 15. 15 Summary steps (cont.)  23 separate models developed.  One for each of 22 IHS analytical strata, plus stratum of enumeration areas which, although rural, are urban in character.  Relatively common set of independent variables seen across models  Rural models: age-sex group variables, HH head education, highest ed. attainment by female, bicycle ownership, female headship, and more.  Urban models: fewer age-sex group variables, HH head education, cook on wood, EA variables on sanitation, education, rooms per capita.
  • 16. 16 Summary steps (cont.) 4. Assessment made of resultant models made up of both household and EA variables.  Final models developed at this step. 5. The error in each base household model is then modeled in order to control for heteroskedasticity. 6. IHS-derived models then applied to the census data.  Using special program for poverty mapping.  Bootstrap method used to generate:  Point and variance estimates of welfare indicator for a range of spatial groupings of households.  Poverty measures, with standard errors.  Inequality measures, with standard errors.
  • 17. 17 Performance of the poverty mapping models  23 models - Developing so many models is definitely NOT ‘best practice’.  Adjusted-R2 for base models range from 0.248 to 0.594, with mean of 0.380.  Urban models have higher R2s.  Assessment:  The 11 models in Mozambique poverty maps had R2s ranging from 0.26 to 0.55.  Other countries with sharp economic divides that are defined on racial lines (or other household and individual characteristics) tend to have higher R2s. • South Africa (around 0.70), Ecuador (0.50).
  • 18. 18 Poverty map validation – national & regional  Poverty headcount results compared between IHS poverty analysis & poverty mapping analysis.  Differences at national and regional levels are not statistically significant. IHS PovMap National 65.3% 64.3% Regional Southern 68.1% 65.4% Central 62.8% 63.9% Northern 62.5% 61.1%
  • 20. 20  District level  In comparing the IHS poverty analysis to the poverty mapping results, the headcounts for seven out of the 31 districts are significantly different.  Blantyre rural & Lilongwe rural are the most important of these.  In the other five districts, lack of congruence may be due to problems with the IHS data or the IHS sampling scheme.  In spite of the problems in these seven districts, results are reasonable, overall.  Following slides show results mapped at the more local scale of the sub-district Traditional Authority. Poverty map validation – district (cont.)
  • 21. 21 Depth & severity of poverty – Traditional Authority
  • 22. 22 Local government ward  Poverty mapping method used can provide estimates of poverty for populations of 500 to 1000 households and above.  Census enumeration areas (EA) are too small, being 250 households on average.  Only intermediate spatial unit between TA and EA is the local government ward.  Following map shows poverty headcount for the almost 850 local government wards.  Such small-area maps potentially of value for sub- district planning.
  • 24. 24  Poverty estimates for populations in each aggregated EA.  Poverty measures are estimated with standard errors.  Mean s.e. for poverty head-count for rural aggregated EAs of 8.2 percent.  In examining results of analysis, the consider- able imprecision in underlying estimates should be kept in mind. Error in poverty maps
  • 25. 25 Additional analyses based on poverty map  Assessment of how well programs target the poor  Determine poverty characteristics of population in program areas and compare to wider poverty prevalence.  Ex-post analysis of poverty targeting.  Malawi Social Action Fund public works locations & WFP Food for Assets & Develop- ment project sites.  See IFPRI’s FCND discussion paper no. 205
  • 26. 26 Geographic scale and poverty mapping – even more local  Constructed a ‘new’ geography for Malawi based on amalgamations of two or three EAs.  Subdivided the country into about 3,400 spatial units with populations above 500 households in order to calculate local welfare and poverty measures for population in each.  Primarily for analytical purposes, rather than for planning.  Used in the investigation of the spatial determinants of poverty prevalence in rural Malawi.
  • 27. 27 Creating the Malawi poverty map – Institutional considerations  Poverty map creation was an activity of the Poverty Monitoring System of the government of Malawi.  Three principal institutions:  National Statistical Office (NSO)  National Economic Council (NEC - now Ministry of Economic Planning & Development)  Centre for Social Research (CSR) of the University of Malawi  IFPRI provided technical assistance on poverty analysis to all three.  Poverty map was logical extension of the poverty analysis work that NSO and NEC jointly led.  However, NEC did not engage in actual analysis.
  • 28. 28 Creating the Malawi poverty map – Institutional considerations (cont.)  Poverty mapping work done by two statisticians from NSO, led by IFPRI researcher.  In retrospect, this was not ideal.  Poverty mapping, to do confidently, requires relatively sophisticated econometric understanding & abilities.  Such abilities not best placed within a national statistical office.  Their principal activities do not include such high-level analyses.  Rather, should have involved university-based Malawian econometrician(s), as well.  Technical abilities required, retention & refinement of poverty mapping skills, and analytical leadership for future poverty mapping efforts nationally, best served by universities.  So, Malawi case was a missed opportunity for sustainable capacity strengthening in technical analysis for poverty mapping.
  • 29. 29 Creating the Malawi poverty map – Data considerations  Had contemporaneous IHS and census data sets.  However, this temporal correspondence not absolutely critical.  Key concern is that nature of correlation between household welfare and independent variables used in poverty mapping model for each stratum will not have changed significantly between the two time periods.  Judgment call that takes into account the dynamism and character of economic transformation across country.
  • 30. 30 Creating the Malawi poverty map – Data considerations (cont.)  Had two separate GIS shapefiles of enumeration areas. For both:  IHS, sample design of which was based on EAs for 1987 Malawi census, and  1998 Census.  Facilitated generation of GIS-based cluster-level variables for inclusion in stratum models and then for use in applying models to census data.  Also permitted extensive spatial analysis of poverty mapping results.  However, having EA digital maps not absolutely critical.  Can use smaller-scale, higher geographical units, such as districts, for generating cluster variables.  GIS-derived, census-aggregates, or developed from other geographically- comprehensive ancillary data.  Loss of information since any intra-district variation in the data is lost.  Econometrically also not ideal, as EA-level cluster variables in models assist in dealing issues arising due to expected correlation of characteristics of survey households in same EA.  Results in some loss in explanatory power and performance of models, but an acceptable solution.