A presentation by Peter Davis and Bob Baulch from the 2009 BASIS Conference on "Escaping Poverty Traps: Connecting the Chronically Poor to the Economic Growth Agenda."
2. Introduction
• In poverty research, different methods
often lead to different findings
• In the study of poverty dynamics
differences may be magnified
• Differences in findings can lead us to:
– critically assess methods
– mix methods strategically to
strengthen research findings
– attempt to uncover drivers of change
more reliably
– and therefore be able to suggest
more effective interventions
3. The focus of this presentation
• What can we learn by
integrating quantitative and
qualitative assessments of
socio-economic mobility of
the same individuals and
households?
• The implications of these
lessons for:
– poverty-dynamics research
– interventions to reduce chronic
poverty
4. The CPRC-DATA-IFPRI Bangladesh
longitudinal study
• The study combined three IFPRI
evaluations which started in
1994, 1996 and 2000/03, and
used a mixture of quantitative
and qualitative methods
• In 2006-7 we resurveyed the
entire set of these households
(plus new households created
due to household division ) in
three phases (qual-quant-qual)
5. The 2006-7 Study’s 3 Phases
3 phases of data collection:
• Summer 2006: focus group discussions
investigating causes of decline and
improvement and the long term impact of
3 interventions (116 FGDs in 11 districts)
• Winter 2006-7: quantitative resurvey of
panel households (1787 core + 365 splits
in 14 districts)
• Spring-Summer 2007: life-history
interviews and village histories in 8
districts (161 households – 293
individuals)
6. Map of the Study Sites
Nilphamari (38)
Kurigram (39)
Tangail (39)
Kishoreganj (19)
Mymensingh (18)
Manikganj (72)
Jessore (36)
Cox’s Bazar (32)
Life-history districts
(number of interviews)
7. Poverty and Growth in the Study Sites
Microfinance
(1994-2006)
Agricultural
technology
(1996-2006)
Educational
transfers
(2000-2006)
Poverty headcount
Poverty in baseline
survey
60% 62% 71%
Poverty in 2006/2007 21% 13% 28%
Growth in Per Capita Expenditures
Over period 28.0% 43.5% 44.3%
Annualised 2.1% 3.7% 6.5%
8. Methods used to assess poverty transitions
1) Quantitative: transition matrices based on per capita
expenditures and the BBS upper poverty lines
2) Qualitative: Changes in individual well-being levels
Level
English Bangla Guideline
1
Very poor or
destitute
khub gorib,
na keye
chole
Suffering tangible harm to health because of poverty, generally due to
insufficient food. Usually landless or near landless
2
Poor gorib
Very vulnerable but eating reasonably well. Could easily move into 1 due to a
common shock. For a medium size household, usually less than an acre for a
medium sized household
3
Medium
madhom
A common shock would not result in tangible harm or going without food. Hold
household assets or generate household income equivalent to between one and
two acres of land for a medium-sized household.
4
Rich
dhoni
Hold household assets or generate household income equivalent to that
generated by two to ten acres for a medium-sized household.
5
Very rich
khub dhoni
Hold household assets or generate household income equivalent to that
generated by ten acres or more for a medium sized household.
9. Transition matrix
(from per capita expenditures)
First round
(1994,1996,
2000)
2006-7
Poor Non-Poor Total
Poor 394 1081 1475
Non-Poor 66 598 664
Total 460 1679 2139
10. Transition matrix
(from well-being levels)
First round
(1994,1996,
2000)
2006-7
Poor Non-Poor Total
Poor 170 14 184
Non-Poor 23 86 109
Total 193 100 293
11. Mismatches between Qual and Quant
Assessments of Poverty Dynamics
quantitative
expenditure-
based
categories
qualitative wellbeing
categories
(numbers of people)
PP PN NP NN Total
PP 50 3 4 9 66
PN 74 3 13 31 121
NP 20 0 2 4 26
NN 26 8 4 42 80
Total 170 14 23 86 293
12. Exploring the ‘mismatches’
1. Cases where per capita expenditure does not
accurately reflect the economic wealth of the
household
– Asset-based transitions have more matches
13. 1. Expenditure an imperfect indicator of wealth
Classifying quant transitions using land assets halves the mismatches
quantitative
asset-based
categories
qualitative matrix categories
(numbers of people)
PP PN NP NN Total
PP 99 8 0 14 121
PN 6 0 3 6 15
NP 41 2 7 6 56
NN 24 4 13 58 101
Total 170 14 23 86 293
14. Box 1: Expenditure is an imperfect indicator of wealth
(qual PP: quant NN)Circumstances:
• Woman (57)
• Sold land to live
while husband ill -
died in 1980
• Lives with son (29)
working as a
mason
• Son injured 1996-
2001
• 4 decimals of land
owned
• Own illness since
2004
1994 2007
Per Capita Expenditure 778 2538
Poverty line (BBS) 547 877
Land owned (decimals) 100 4
15. Exploring the ‘mismatches’
1. Expenditure is an imperfect indicator of wealth
– Asset-based transitions have more matches
2. Cases where households’ expenditures are close to the
poverty line in either, or both, survey rounds.
– High numbers of households near the poverty lines
mean small changes in expenditure can cause
transitions
16. 2. Proximity to poverty lines:
Distribution of per capita expenditures
and poverty lines0.001.002
ProportionofHouseholds
0 1000 2000 3000 4000
Monthly per capita expenditure
1996
0.001.002
ProportionofHouseholds
0 1000 2000 3000 4000
Monthly per capita expenditure
2007
Agricultural Technology Sites
17. Box 3: Proximity to poverty lines
(qual PP quant NP)
1994 2007
Per cap. Expenditure 796 690
Poverty line (BBS) 547 877
Household members 3 4
Land owned (decimals) 13 3
Circumstances:
• Man 26
• Married in 1996
• Split from
parents in 2001
• Lives with wife
and 2
daughters
• Only one
household
member the
same as in
1994
• Day labourer
• Own one cow
18. Exploring the ‘mismatches’
1. Expenditure is an imperfect indicator of wealth
– Asset-based transitions have more matches
2. Proximity to poverty lines
– High numbers of households near the poverty lines
mean small changes in expenditure can cause
poverty transitions
3. Non-monetary aspects of ill-being were not detected in
the expenditure-based measurement
-domestic violence, disability, illness, or vulnerability
19. Box 4: Non-monetary aspects of illbeing not detected
(qual PP but quant PN)
Circumstances
• Man (45) living with
his wife (36), 2
daughters, 2 sons
• Drives a van gari
• One disabled daughter
• Own chronic illness
since 2002
• Dowry problems for
eldest daughter
1996 2007
Per Capita Expenditure 312 931
Poverty line (BBS) 551 773
Household size 6 6
Land owned (decimals) 174 12
20. Exploring the ‘mismatches’
1. Expenditure is an imperfect indicator of wealth
– Asset-based transitions have more matches
– Liberal spenders versus frugal spenders
2. Proximity to poverty lines
– High numbers of households near the poverty lines
mean small changes in expenditure can cause
transitions
3. Non-monetary aspects of ill-being were not detected in
the expenditure based measurement
-domestic violence, disability, illness, or vulnerability
4. Cases where changes in household size (often due to a
‘split’) led to changed household economies of scale
21. Box 5: mismatch caused by diseconomies of scale
qual PP but quant PN
Circumstances
• Woman (56) living
with her husband
(64)
• Income from selling
snacks
• 10 decimals of
homestead land,12
trees
• 2 daughters and 3
sons separated
• Land sold to pay for
daughter’s dowries
1994 2007
Per Capita Expenditure 412 906
Poverty line (BBS) 501 799
Household size 7 2
22. Exploring the ‘mismatches’
1. Expenditure is an imperfect indicator of wealth
– Asset-based transitions have more matches
– Liberal spenders versus frugal spenders
2. Proximity to poverty lines
– High numbers of households near the poverty lines
mean small changes in expenditure can cause
transitions
3. Cases where some non-monetary aspects of ill-being
were not detected in the expenditure based
measurement (such as the impact of domestic violence,
disability, illness, or vulnerability)
4. Cases where changes in household size (often due to a
‘split’) led to changed household economies of scale
5. Cases where recall errors affected qualitative
assessments
23. Sequential reduction in mismatches
Individual (per cent) Cumulative (per cent)
Total mismatches 196 (66.9) 196 (66.9)
Wealth not expenditures 93 (47.4) 103 (35.2)
Proximity to poverty line 60 (30.6) 69 (23.5)
Non-monetary aspects
of ill-being
43 (21.9) 60 (20.5)
Changes in household
size
33 (16.8) 46 (15.7)
Qualitative recall errors 16 (8.2) 42 (14.3)
24. Trajectory patterns
Direction Pattern Depiction
Number
of Cases
Weighted
Percent of
Cases
Stable Smooth 8 1.47
Improving Smooth 3 1.43
Declining Smooth 2 0.36
Stable Saw-tooth 135 44.98
Improving Saw-tooth 76 26.15
Declining Saw-tooth 30 6.90
Declining Single-step 2 0.48
Declining Multi-step 37 18.22
Total 293 100
25. Lessons from integration
• Movement across monetary poverty lines can happen
with little tangible change in people’s well-being
• Various types of vulnerability are not visible in standard
quantitative approaches
• Including assets helps to improve assessments
• Studying individuals and households over long periods
adds to the conceptual and methodological
complications of poverty measurement
• With new challenges to understand the impact of global
changes on the chronically poor, we need reliable
mixed-methods approaches to poverty dynamics
26. Some conclusions
• Movements out of poverty are usually slow -
declines can be fast and irreversible
• People move out of poverty
– by building up assets (land, livestock etc.)
business, agriculture, educated children
working, employment and remittances
• People moving out of poverty are still
vulnerable
– food prices, loss of income, illness, dowry
• Better understanding of the crises and
opportunities poor people face assists in
prioritising and rationalising anti-poverty
interventions and enhancing social protection