Writing Sample-Title: Pioneering Urban Transformation: The Collective Power o...
Analysing the Dynamics of Change: Using longitudinal, panel and cross sectional studies in development research.
1. John Thompson - j.thompson@ids.ac.uk
Future Agricultures Consortium / APRA CEO
Senior Research Fellow, Institute of Development Studies, UK
Seminar for Institute for Poverty, Land and Agrarian Studies, University of the Western Cape, South Africa
30 July 2019
Analysing the Dynamics of Change:
Using longitudinal, panel and cross-sectional
studies to investigate complex social,
environmental and technological processes
Lessons from Future Agricultures and the APRA Programme
2. Focus of presentation
• Introduce 3 types of studies that examine dynamic social, environmental and
technological change over time at individual, household and community level:
1. Longitudinal
2. Panel
3. Cross-sectional studies
• Case studies from Future Agriculture’s APRA Programme
– collaborative, multi-country, mixed-methods (quant-qual) studies
– tracing change over time and highlighting the differential outcomes on
local livelihoods
• Compare the basic elements of each approach and highlight their strengths
and weaknesses
3. What are longitudinal studies?
Observational studies that follow the same subjects (‘study
participants’) repeatedly over a period of time
• Some longitudinal studies (LS) track the same people throughout
their lives – e.g. from birth to old age
• Some follow individuals, while others follow households or social
groups, or even organisations
• Some will also compare more than one case / location / context
(multi-locational or multi-sited studies)
4. Why conduct longitudinal studies?
To understand…
• how key life transitions change the course of a person’s life (e.g.
marriage or divorce, or entering or retiring from the labour market)
• how earlier life circumstances affect later life (e.g. exposure to
disease, education, wealth)
• how patterns of behaviour change as people grow older
• how different areas of lives are linked (e.g. health, wealth, family,
education, employment and social attitudes)
• how different aspects of life vary for different people
• how human-environment interactions change over space and time
5. Common types of longitudinal studies
1. Cohort studies: chart the lives of groups of individuals who
experience the same life events in a given period
2. Panel studies: follow the same households or individuals over
time – e.g. the World Bank’s Living Standards Measurement
Study (LSMS), USAID’s Demographic and Health Surveys (DHS),
UNICEF’s Multiple Indicator Cluster Surveys
3. Tracker studies: follow the same households or individuals over
time, but track the ‘movers’ to other locations to reduce attrition
rates (‘drop outs’)
6. Cohort studies
• Cohort studies chart the lives of groups of individuals who experience
the same life events within a given time period
• Most are birth cohort studies which follow groups of people born
within a specific time period
Richter, et al. (2007) Cohort Profile: Mandela's children: The 1990 birth to twenty study in South Africa. International Journal of Epidemiology 36(3):
504–511 https://doi.org/10.1093/ije/dym016
• Largest cohort study in Africa is the ‘Birth to
Twenty Study’ (BT20) – began in 1990 and tracked
a cohort of over 3,000 children born in the weeks
following Nelson Mandela's release from prison
(‘Mandela’s children’)
7. Panel studies
• Panel studies follow the same individuals
over time – over a number of ‘waves’ – and
vary considerably in scale and scope
• ‘Snapshot’ vs. ‘Film strip’ – showing how
individuals or households changed over time
• In a household panel study, one or more
members of the household participate in
the study over two or more waves
• Can also be used to study changes in
organisations and institutions
8. Tracker studies
• People who move from their home areas are unlikely to be a
random subset of the baseline respondents
• They are likely to have certain characteristics that differ from
those who remain
• Tracker studies follow these ‘movers’ – as well as the ‘remainers’
• This can help to minimise the ‘attrition’ rate and address potential
selectivity biases
9. Tracker studies
Tracking: following respondents wherever they move…
Wave 1
• Josiah (male head of household)
• Gloria (female head of household)
• Mary (daughter 1)
• Thomas (son 1)
• Hannington (son 2)
• Eliud (male elder)
• Josiah (male head of household)
• Gloria (female head of household)
• Mary (daughter 1) – left household
• Thomas (son 1) – left household
• Hannington (son 2)
• Eliud (male elder) – deceased
• Andrew (male head of household)
• Mary (female head of household)
• Charity (daughter 1)
• Japhet (son 1)
Wave 2
• Thomas (male head of household)
HH 1, Village 1
HH2,
Village 2
HH1,
Village 1
HH3,
Town 1
10. Cross-sectional studies
• Cross-sectional studies – can compare different samples at a
specific point in time – thus they provide a ‘snapshot’
• Multiple variables can be studied – e.g. age, gender, ethnicity,
income, education, etc.
• They are typically quicker and cheaper to conduct than multi-
wave longitudinal panels
• Limits of CS studies – harder to pin down clear cause-and-
effect relationships
11. Prospective vs. retrospective studies
• Prospective studies – follow individuals over time (i.e. into
the future) and data about them is collected as their
characteristics or circumstances change
• Panel studies and birth cohort studies are good examples
start
12. Prospective vs. retrospective studies
• Retrospective studies – select from a sample of individuals
and information is collected about their past
• Recall studies have significant limitations and therefore
require extra rigour (cross-checking/triangulation) to confirm
statements by informants
start
13. Prospective vs. retrospective studies
• In reality, many studies use both prospective and retrospective
methods
• In birth cohort studies, participants are often asked to
retrospectively provide information on their lives since the
previous interview
• In household panel and cross-sectional studies often collect an
array of retrospective information about past events
start
14. Combining elements over time
6 Year Olds
8 Year Olds 12 Year Olds10 Year Olds
10 Year Olds8 Year Olds
Yearofbirth
2008
2006
2014 2016 2018
Cohort
Comparisons
Cross-Sectional
Comparisons
Longitudinal
Comparisons
15. Funded by UK aid from the UK Government
www.future-agricultures.org/apra
A Brief Overview of the Programme’s Aims and Approach
APRA:
Agricultural Policy Research in Africa:
Analysing Pathways to Commercialisation
16. Pathways to ‘inclusive’ commercialisation
• Led by Future Agricultures Consortium (FAC)
• Five-year grant from UK DFID to establish APRA –
Agricultural Policy Research for Africa
• Focus: analysing the differential impacts of agricultural
commercialisation on local livelihoods and rural
economies – in 8 countries
• APRA involves large network of existing FAC partners,
including PLAAS and IDS – plus new collaborators in Africa,
UK, Sweden and USA (100+ contracted researchers)
17. A typology of agricultural commercialisation
1. Large-Scale Estate 2. Medium-Scale Commercial
4. Small-Scale Independent3. Outgrower/Contract
18. 1. ‘Stepping in’ – returning / moving into commercial agriculture
from a non-farm base; rise of medium-scale ‘investor’ farmers
2. ‘Stepping out’ – accumulating, diversifying and creating
alternative, non-farm economic activities
3. ‘Stepping up’ – improving and investing in existing
agricultural activities and engaging in the market
4. ‘Hanging in’ – maintaining subsistence level
5. ‘Dropping out’ – moving out of productive agriculture or
slipping into destitution due to shocks and stresses
Analysing 5 livelihood trajectories through
agricultural commercialisation
19. Funded by UK aid from the UK Government
www.future-agricultures.org/apra
Early Insights from the First Round of a
Multi-Country Panel Study of Agricultural
Commercialisation in Sub-Saharan Africa
Amrita Saha, Rachel Sabates-Wheeler and John Thompson
Agricultural Policy Research in Africa (APRA) Programme
Future Agricultures Consortium
20. APRA panel studies
Panel Studies of different commercialisation
pathways and people’s ‘selection choices’ over
two waves
• Combining quantitative and qualitative
elements
• Ghana – oil palm commercialisation models
• Tanzania – rice and maize commercialisation
• Nigeria – role of medium scale commercial
producers and interactions with smallholders
• Zimbabwe – tobacco and maize
commercialisation channels
21. APRA panel studies
• Mixed methods research carried out in Ghana,
Nigeria, Tanzania and Zimbabwe, each
comprising representative data on highly
commercialised areas (‘hot spots’)
• Panel survey – first wave conducted in 2017-18
• A total of 3,993 farmer households interviewed
• Farmers in each country grow a variety of
different crops, e.g. maize, oil palm, rice,
tobacco, mostly for commercial sales, and some
home consumption
• A second wave is planned for 2019-2020
42%22%
12% 17%
22. APRA livelihood outcome indicators
Analysing livelihood outcomes indicators resulting from
engagement with the different commercialisation types in 5 key
areas:
1. Agricultural commercialisation choices
2. Women’s empowerment
3. Labour and employment
4. Food and nutrition security
5. Poverty and inequality
23. APRA panel research process
Purposively selected study areas where farmers were increasingly
engaging with markets
• Used Computer-assisted personal interviewing (CAPI) interviewing
technique – WB’s ‘Survey Solutions’
• Structured core questionnaire + country-specific modules for randomly
selected households on: plots cultivated, crops grown and sold, and incomes
calculated ‘household commercialisation index’ (HCI)
• Plus data on the other APRA livelihood outcome indicators
• Combined with secondary data and in-depth qualitative research
Some emerging insights from the analysis of the poverty and
inequality indicator data – but only preliminary findings
24. Agricultural commercialisation & income poverty
Ghana
Note: Based on adult equivalised total net cash income from crops and non-farm income.
500
10001500200025003000
lowessEquiv.incomeHCI
20 40 60 80 100
HCI
120013001400150016001700
lowessEquiv.incomeHCI
0 20 40 60 80 100
HCI
Zimbabwe
0
200000400000600000800000
lowessEquiv.IncomeHCI
0 20 40 60 80 100
HCI
Nigeria
400000600000800000
10000001200000
lowessEquiv.IncomeHCI
0 20 40 60 80 100
HCI
Tanzania
• Nigeria & Tanzania: increasing
relationship between HH
equivalised income (agri+non-
agri) and HCI
• Ghana: Initially decreasing, then
increasing up to a certain
threshold
• Zimbabwe: Initially increasing at
a decreasing rate, then
increasing at an increasing rate
up to a threshold
Lowess: Running locally weighted least squares regression (no control variables)
increasing
increasing
increasing
Increasing at
decreasing
rate
Increasing at
increasing
rate
Threshold?
Threshold?
Poverty & Household Commercialisation Index (HCI)
decreasing
25. Agricultural commercialisation & income poverty
Ghana
Note: Based on adult equivalised total net cash income from crops and non-farm income.
ZimbabweNigeria
Tanzania
• Tanzania & Nigeria: increasing
relationship between HH
equivalised income (agri+non-
agri) and HCI
• Ghana: Initially decreasing, then
increasing up to a certain
threshold
• Zimbabwe: Initially increasing at a
decreasing rate, then increasing
at an increasing rate up to a
threshold
Lowess: Running locally weighted least squares regression (no control variables)
0
500
100015002000
0 20 40 60 80 100
HCI
adult equivalent net income in 2011 PPP$ lowess aenyppp HCI
0
200000400000600000800000
0 20 40 60 80 100
HCI
pp_income_equiv lowess pp_income_equiv HCI
0
10002000300040005000
20 40 60 80 100
HCI
pp_income_equiv lowess pp_income_equiv HCI
0
500000
10000001500000
0 20 40 60 80 100
HCI
HH Equiv. Income lowess HH Equiv. Income HCI
Scatterplot for HH Equiv. Income
26. Agricultural commercialisation & inequality
0
.2.4.6.8
1
cumulativeoutcomeproportion
0 20 40 60 80 100
population percentage
1st quartile (Gini = .604) 2nd quartile (Gini = .5)
3rd quartile (Gini = .444) 4th quartile (Gini = .347)
Ghana=0.52
0
.2.4.6.8
1
cumulativeoutcomeproportion
0 20 40 60 80 100
population percentage
1st quartile (Gini = .485) 2nd quartile (Gini = .336)
3rd quartile (Gini = .3) 4th quartile (Gini = .477)
Nigeria=0.62
0
.2.4.6.8
1
cumulativeoutcomeproportion
0 20 40 60 80 100
population percentage
1st quartile (Gini = .509) 2nd quartile (Gini = .317)
3rd quartile (Gini = .292) 4th quartile (Gini = .462)
Zimbabwe=0.54
0
.2.4.6.8
1
cumulativeoutcomeproportion
0 20 40 60 80 100
population percentage
1st quartile (Gini = .606) 2nd quartile (Gini = .504)
3rd quartile (Gini = .517) 4th quartile (Gini = .483)
Tanzania=0.63 Nigeria & Zimbabwe:
Most commercialised &
least commercialised
HHs with highest
inequality
Ghana & Tanzania:
Least commercialised
HHs are most unequal
Note: Lorenz curves on a regular-grid of 20 equally spaced points across the population (plus a point at the origin)
Inequality declines with
Rise in commercialisation
Inequality declines with
Rise in commercialisation
High inequality across
commercialisation
High inequality across
commercialisation
27. Exploring the relationship between commercialisation & assets
5678
Commercialisation(LogSales)
0 .2 .4 .6 .8 1
lpoly smoothing grid
95% CI lpoly smooth
Local polynomial for pooled data
Local polynomial smooth
5.5
6
6.5
7
7.5
Commercailisation(Logsales)
0 .2 .4 .6 .8 1
lpoly smoothing grid
95% CI lpoly smooth
local polynomial for pooled data
Local polynomial smooth
5678
Commercialisation(LogSales)
0 .2 .4 .6 .8 1
lpoly smoothing grid
95% CI lpoly smooth
Local polynomial for pooled data
Local polynomial smooth
5.5
6
6.5
7
7.5
LogCommercialisation
0 .2 .4 .6 .8 1
lpoly smoothing grid
95% CI lpoly smooth
local polynomial for pooled data
Local polynomial smooth
Pooling the four country samples and running a local
polynomial non-parametric regression (without any
control variables) a slightly upward gradient with
commercialization (sales) emerges for the production
equipment & consumer durables assets.
Production Assets
Consumer Durables
Housing Services
Positive correlation between ownership of
consumer durables & production equipment with
household’s commercialisation
0.44
0.26
0.45
0.59
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
GHANA TANZANIA NIGERIA ZIMBABWE
All Assets Production Assets Consumer Durables Housing Services
28. Commercialisation, assets & subjective perceptions of poverty
0
.2.4.6
0
.2.4.6
0 5 10 15 0 5 10 15
Ghana Nigeria
Tanzania Zimbabwe
Richer Poorer
kdensitylog_comm
x
Graphs by country
HHs that perceive themselves as richer
Ghana: comparable sales but marginally
higher assets
Tanzania: have only marginally higher sales
and assets
Nigeria: have only marginally higher sales
and assets
Zimbabwe: marginally higher sales but much
higher assets
02460246
0 .5 1 0 .5 1
Ghana Nigeria
Tanzania Zimbabwe
Richer Poorer
kdensitynormalAsset_Index
x
Graphs by country
HHSalesHHAssetIndex
-Richer than Average (Amongst the richest in the village;
Richer than most households; The richest in the village)
-Poorer than Average (A little poorer than most
households; Amongst the poorest in the village; The
poorest in the village)
Only marginal differences in commercialisation with differences in
subjective welfare outcomes;
Differences in asset ownership by subjective welfare outcomes.
29. Observations
• First wave gave us first impression of commercialisation choices and
pathways, but not full picture of outcomes
• Qualitative research was essential – used to conduct initial reconnaissance
of sites and then to understand the changing commercialisation contexts
• Aiming to conduct analyses of first wave data for all APRA outcome areas
• While we are awaiting those results, we have commissioned a study of WB
Living Standards Measurement Survey (LSMS) data to analyse the trends
in commercialisation and its related outcomes in our APRA study countries
• Now preparing for second wave, which should allow us to get a better
picture of commercialisation choices and outcomes
30. Experiences conducting the first round panels
• Implementing a multi-themed survey with numerous modules was
technically challenging for the country teams
• Large investments were made in supporting the teams to design
the studies, collect the data and write-up their results – but still
found capacity gaps – major resources and commitment required
for backstopping!
• Lots of pre-testing was done, but still there were issues – e.g. with
confirming local units of measure, sales figures – test and test
again!
• Datasets were large and took months to clean, despite using CAPI
approach - there are no shortcuts!
31. Longitudinal Study of
Rice Commercialisation and Agrarian
Change on the Fogera Plain of Ethiopia
Dawit Alemu, Abebaw Asaye, Degu Addis, Tilahun Tadesse, Agajie Tesfaye, John Thompson,
Rachel Sabates-Wheeler and Shewaye Abera
Funded by UK aid from the UK Government
www.future-agricultures.org/apra
32. APRA longitudinal studies
Longitudinal Studies of pathways to
agricultural commercialisation
• 6 countries: Ethiopia, Ghana, Malawi, Nigeria,
Tanzania and Zimbabwe
• Analysing how different pathways of
commercialisation evolve over time from a
wider historical assessment of the dynamics
of agrarian change
• Examining how these dynamics influence the
different livelihood trajectories of rural men
and women in different contexts
• Focusing on processes of ‘stepping up’ and
‘stepping out’, in particular
33. Rice commercialisation as a driver of change
• Agricultural commercialisation is
generally viewed as an essential part of
structural transformation of the economy
• Introduction of rice in 1980s changed
Fogera Plain from one of the most food
insecure areas of the country to a thriving
commercialisation ‘hot spot’
• With increased commercialisation of rice,
there have been diverse agricultural and
socio-economic changes in the region –
some negative, many positive
35. Methodology
• Secondary data sources
• Primary data:
– 3 linked formal surveys
– 3 rounds of KIIs and FGDs using
participatory methods with stakeholders
• Retrospective and prospective analysis
36. Surveys and sampling
Survey Sample size Sampling method
Smallholder rice
producers
723
Stratified random
samples
Rural labourers 265
Stratified random
samples
Rice processors 123 Population census
37. Results and discussion
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
38. Farming system changes
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
• The introduction and expansion of rice in the Fogera plain has
brought considerable land use changes:
– Shift towards to more commercially viable commodities like
rice, pulses and vegetables
– Decline in other cereals and livestock
• Different trends in the two major rice agro-ecologies (lowland and
upland systems) – but increasing rice production in both
39. Farming system changes
Start of the fall of Derge
Regime (1990s)
Ten years ago
(2000s)
Current period
(2017)
Upland
Lowland
40. Changes in land tenure
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
• Increased rice commercialisation has been among the key factors
behind the emergence of different forms of land transfers
• Within the policy framework prevailing in the Amhara region, the
land markets followed three approaches:
1) Land sharing
2) Leasing
3) Public land allocation
• Prevalent land disputes over sharing and leasing have forced the
regional government to establish mandatory registration of any land
transfers at local level
• The directive stipulates that rural land is restricted to agricultural use
41. Changes in land tenure
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
• With increased landlessness esp. among rural youth, there is
administrative land allocation mainly distributing communal
lands (previously used for grazing)
• The number of youth beneficiaries and total land allocated in
Fogera shows a considerable increase over 2008-2018:
‒ 1,600 ha of communal land was distributed to about
2,950 youth in Fogera woreda alone
‒ Many have become rice farmers
42. Changes rural labour markets
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
• With increased rice production and high labour
requirements, there has been the emergence of dynamic
rural labour market
• Over 55% of rice farmers reporting hiring labourers
• The major farm operations for hired labour use were
ploughing, weeding, harvesting and/or threshing
• Majority of hired labour is from the local area, though some
are migrants from more distant semi-arid regions
43. Rural-urban linkages
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
• Increased rice commercialisation has led to the rapid
emergence of a growing processing industry in nearby towns
• This has created dynamic rural–urban interactions and
multipliers
• These are related with marketing related linkages for:
‒ Rice processing and marketing – value addition
‒ Transportation
‒ Asset accumulation and increased household consumption
‒ New services (banking, hotel, education, etc)
‒ Seasonal and annual labour with increased wage rates
44. Rural-urban linkages
1) Dynamics in farming systems
2) Land tenure changes
3) Emergence of labour market
4) Dynamism in rural – urban linkages
5) Changes in consumption patterns
Trends in the expansion of rice processing industry [number of rice
processors] in Fogera area
1 1 2 2 3
8 9 11 13 16
36 40 44
50
57
64
69
83
97
123
0
20
40
60
80
100
120
140
1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
Numerofoperationalprocessors
Year
45. Rural-urban linkages
Aspirations of rice processors in Fogera
Business type
% of
processors
Type of expansion businesses
% of
processors
Plan to expand
rice business
88.5
Purchase more processing machines 74
Utilise the full potential processing capacity of the same machines 74
Employ more labour 69.9
Import modern processing machine 57.7
Establish new processing facility in the same town 44.7
Establish new processing facility in other towns 32.5
Engage in direct rice production 19.5
Plan to expand
to non-rice
business
59.0
Trading of other products 35.8
Transport sector 30.1
Processing of other agricultural products 27.6
Building construction (town residence or business house) 22
Hotel industry 15.4
46. Changes in consumption patterns
• The increased production and marketing of rice has resulted in
major changes in household consumption patterns
• Food and nutrition security has improved in terms of
composition and frequency of consumption – from 1-2 to 2-3
meals per day; greater dietary diversity
• But some decline in the consumption of milk and dairy products
linked with the reduction of livestock in the area
• Injera (traditional Ethiopian flatbread) now made from mix of
rice and teff or rice and millet
47. Observations
• Rice commercialisation having a multiplier effect in Fogera
‒ Increased household incomes
‒ Improved food and nutrition security
‒ Creation of casual and permanent labour opportunities
‒ Dynamic land market – but with displacement of livestock
‒ Growing rural-urban linkages
‒ Enhanced investment in other economic sectors
• Changing livelihood trajectories – from ‘hanging-in’ to ‘stepping-up’
and ‘stepping-out’
• Observed changes were not induced by interventions from public
programmes, but mainly farmer-led innovations
48. A Tracker Study of Groundnut
Commercialisation and Livelihood
Trajectories in Malawi
Blessings Chinsinga, Loveness Msofi, Mirriam Matita, Jacob Mazalale,
Masautso Chimombo and Ephraim Chirwa
APRA Work Stream 2 Longitudinal Study
Funded by UK aid from the UK Government
www.future-agricultures.org/apra
50. Tracking change in Malawi
• Assessing long-term changes in livelihoods
associated with the differential engagement
with agricultural commercialisation
• Focus is on groundnut commercialisation in
Malawi’s central districts of Ntchisi and
Mchinji
• Goal is to understand the drivers of
groundnut commercialisation in these two
districts, from both a historical and
contemporary perspective, and their impacts
on livelihood trajectories
51. APRA Tracker research methods
The APRA Malawi Tracker is based on a survey that was conducted
in the 2006/2007 growing season, which focused on the Farm
Input Subsidy Programme (FISP)
The team revisited individuals from the original survey, including
any off-shoot households established in the last decade
• Tracked 216 original HHs out of total sample of 240 HHs
• Total sample 519 HHs interviewed
• 25 outside country (Zambia, Mozambique, SA)
• Qualitative data (28 FGD and 6 KII interviews completed)
53. Observations
1. Agricultural Extension (ADMARC) service non-existent – reduced to a
vehicle for FISP (providing vouchers for purchase of fertiliser and hybrid
seeds)
2. Private vendors are biggest buyers of farm produce
3. Less progress in HH farming enterprises; largely ‘hanging in’, few ‘stepping
out’ cases @ rural trading centres
4. Mchinji farmers more exposed to market forces – NGOs, export markets to
Mozambique and Zambia
• APRA Malawi team have not fully analysed their quantitative data; data
cleaning is underway; further qualitative work planned
• Several stakeholder engagements & meetings undertaken
54. Experiences tracking households
• Transfers/failure to locate HH, death of eligible members
• Some ‘ghost’ HH members in 2006/7 roster in anticipation of handouts
• Few refusal cases
• Excitement among respondent to be tracked after 10 years increase response rate
• Unexpected community events slowed work
• Long distance travel to conduct interviews of split HHs
• Some HH members who moved are still dependent on original HH heads
• 2 blogs specifically written on tracker expectations and experiences available on the
Future Agricultures APRA website:
1. https://www.future-agricultures.org/blog/malawi-tracker-study-experiences-from-the-field/
2. https://www.future-agricultures.org/blog/conducting-a-tracker-study-a-tough-nut-to-crack/
56. Strengths of longitudinal studies
• Detail over the life course: Value increases as each wave
builds on what is already known about the study participants
• Contain far more detailed information than could be collected
through a one-off study (but danger of data overload!!)
• Reducing recall bias: Prospective LS help reduce the impact of
recall error or bias, which occurs when people forget or
misremember events when asked about them later
57. Strengths of longitudinal studies
Many of the advantages of LS relate to the analytical
questions their data can help address – e.g.:
• Exploring patterns of change and the dynamics of
individual behaviour
• Understanding how people follow particular livelihood
trajectories (e.g. hanging in’, ‘stepping up’ and ‘stepping
out’…)
58. Strengths of longitudinal studies
LS can provide insights into causal mechanisms and processes:
• LS data cannot definitively ‘prove’ causality, but can give
more insights into the causal processes that might be
involved
• Can allow events to be ordered correctly in time
understanding the order in which events occur is important in
assessing causation
59. Weaknesses of longitudinal data
Attrition
• Over time, participants may cease to take part in a
longitudinal study. This ‘attrition’ can result from a range of
factors – e.g. migration, death, dropping out
Conditioning
• It is possible that participants may be ‘conditioned’ – i.e.
answer questions differently or even behaving differently as
a result of their participation in the study
60. Weaknesses of longitudinal data
• Limits of representivity: Because longitudinal cohort studies
focus on following a specific group (rather than a representative
sample of the population as a whole), they are only
representative of that group
• Complexity: Datasets can be more complex to manage and
analyse than the data for cross-sectional surveys (again, data
overload!)
• Cost: LS with a number of waves of data collection tend to be
more expensive than cross-sectional studies
61. Longitudinal vs. cross-sectional studies
Comparators Longitudinal Cross-Sectional
Time / sequence Several points in time – ‘film
strip’
One point in time – ‘snap
shot’
Sampling Same sample Different sample
Level of analysis Change at the individual / hh
level
Snapshot of a given point in
time
Case examples APRA tracker study
APRA WS1 two-wave panel
study
APRA WS1 first wave panel
study
62. Some sources of information on
longitudinal studies in
international development
63. Other examples of longitudinal studies
• Multi-wave, multi-country, panel study – ‘Afrint: Smallholder
Agriculture, Diversification and Gender in Africa’
http://fdslive.oup.com/www.oup.com/academic/pdf/openaccess
/9780198799290.pdf
• Retrospective longitudinal study – ‘More People, Less Erosion:
Environmental Recovery in Kenya’
https://www.odi.org/sites/odi.org.uk/files/odi-
assets/publications-opinion-files/4600.pdf
• Repeat, multi-country, cross-sectional study – ‘Drawers of
Water: 30 Years of Domestic Water Use and Environmental
Change in East Africa’ https://pubs.iied.org/pdfs/9049IIED.pdf
64. Other examples of longitudinal studies
• Repeat, prospective, longitudinal study – ‘EthiopiaWIDE:
Tracking Ethiopian rural communities since 1994’
http://ethiopiawide.net/
• Tracker study – ‘Migration and Economic Mobility in Kagera,
Tanzania’
http://documents.worldbank.org/curated/en/3207214687820970
29/pdf/WPS4798.pdf
• Repeat, cross-sectional study – ‘Evaluating 35 years of Green
Revolution Technology in Villages of Bulandshahr District,
Western UP, North India’
https://doi.org/10.1080/00220380601125180
70. Test Your Knowledge
Longitudinal Studies Quiz
1. True or false? A panel study contacts a fresh sample of people
each time a new wave or sweep is carried out.
Answer: False. Unlike cross-sectional studies, panel studies revisit the same group
of people repeatedly over time. Each wave collects new information about
participants’ lives, adding rich new data to what is already known about them.
71. Test Your Knowledge
Longitudinal Studies Quiz
2. Which of the following is a strength of longitudinal studies?
(select all that apply)
A. They establish the order in which events occur
B. They reduce the attrition of study participants
C. They are cheaper than cross-sectional studies
Answer: ‘A’ only. Longitudinal data collection allows researchers to build up a
reliably ordered account of the key events and experiences in the lives of study
participants.
72. Test Your Knowledge
Longitudinal Studies Quiz
3. Which of the following are characteristics of the household
panel studies described today? (select all that apply)
A.The sample changes over time as people leave households and join new
ones
B.They are representative of a particular generation of people, not the
population as a whole
C. They collect information from whole households at each wave
D.They don’t collect data through survey interviews; instead they rely on
information from secondary sources, such as administrative records
Answer: ‘A’ and ‘C’. Household panel studies follow whole households of people
over time. New participants join the study as households break up and new ones
form.
73. Test Your Knowledge
Longitudinal Studies Quiz
4. Common data collection methods in longitudinal studies include
which of the following? (select all that apply)
A.Structured questionnaire survey
B.Secondary administrative records – e.g. census records and benefit receipts
C. Clinical measurements – e.g. height, weight and blood pressure
D.Participant observation
E. Focus group discussions guided by semi-structured checklists
Answer: All but ‘D’. Participant observation is not normally a part of the
methodological repertoire of longitudinal studies.
74. Test Your Knowledge
Longitudinal Studies Quiz
5. What can longitudinal studies show us? (select all that apply)
A.How key life transitions change the course of a person’s life
B.How early circumstances affect later life
C. How different areas of our lives are linked and how those relationships
change over time
D.How different aspects of life vary for different people throughout their lives
E. How patterns of behaviour change as people get older
Answer: All of the above. The breadth of data collected over time allows researchers
to explore all of these areas.
75. Test Your Knowledge
Longitudinal Studies Quiz
6. How does a tracker study differ from a conventional longitudinal
study?
A.It follows the same households in the same location over a long period of
time (10 or more years)
B.People who move are unlikely to be a random subset of the baseline
respondents and therefore following the movers and interviewing them
minimises the attrition rate and addresses potential selectivity biases
C. The financial cost of tracking individuals can be considerably lower than in
standard household longitudinal studies
Answer: ‘B’ only. Trackers follow individuals over time, whether they remain in their original
locations/households or move elsewhere. They can be expensive, but add additional insights
that help improve our understanding of people’s livelihood trajectories.
76. APRA Contacts
Directorate/IDS
John Thompson j.thompson@ids.ac.uk
Oliver Burch o.burch@ids.ac.uk
Rachel Sabates-Wheeler r.sabates-wheeler@ids.ac.uk
Amrita Saha a.saha@ids.ac.uk
Lesley White l.white@ids.ac.uk
East Africa Regional Hub/CABE
Hannington Odame hsodame@gmail.com
Southern Africa Regional Hub/PLAAS
Cyriaque Hakizimana chakizimana@plaas.org.za
Ruth Hall rhall@plaas.org.za
West Africa Regional Hub/
Univ of Ghana
Joseph Yaro yarojoe@yahoo.com
www.future-agricultures.org/apra
APRA Webpages
www.future-agricultures.org/apra
Funded by UK aid from the UK Government
Editor's Notes
‘BT20’ – Mandela’s children cohort study – From its inception, BT20 was planned to be multidisciplinary, tracking the growth, health, well-being and educational progress of urban children across the first decade of their life. To do so, the research team had to innovate in a number of areas, such as establishing tracking systems under circumstances where few people had a street address or telephone, create a flexible dataset because names were inaccurately translated and transcribed from record to record, translating and adapting questionnaires and establishing norms for many of the measures used. Original planned for 10 years, the study ended up running from 1990-2010.
ITV/BBC ‘7-Up’ series - Begun in 1964, the BBC provided an amazing insight, not just into Britain through the decades, but also into the journey of the lives of those who took part. The children were selected to represent the range of socio-economic backgrounds in Britain at that time, with the explicit assumption that each child's social class predetermines their future. Every seven years, the director, Michael Apted, films material from those of the fourteen who choose to participate. The most recent installment, ‘63 Up’, premiered in the UK on ITV on 4 June 2019 over 3 nights.
In Lowess, the regression is weighted so that the central point (xi; yi) gets the highest weight and points that are farther away receive less weight. The estimated regression line is then used to predict the smoothed value of yi for yi only. The procedure is repeated to obtain the remaining smoothed values, which means that a separate weighted regression is performed for every point in the data.
Adding scatterplot to Lowess to examine concentration
The curve above shows the income distribution, compared to a straight diagonal representing perfect equality.
Illustrates the relationship between asset ownership and commercialisation
Illustrates differences for sales and asset ownership by subjective perceptions of poverty
Please prepare a brief presentation of your study design, methodology, preliminary findings and next steps.
The APRA Work Stream 2 studies aim at exploring the dynamic impacts of commercialisation on livelihoods over a longer period of time, understanding the pathways and drivers of different livelihood trajectories. In particular, WS2 is interested in providing evidence on the transition of households into different livelihoods as a result of commercialisation. These livelihood trajectories include dropping-out (moving out of agriculture due to destitution and other challenges), hanging-in (subsistence agriculture for survival), stepping-up (expanding their existing agricultural activities with the aim of increasing production and commercial activities) and stepping-out (accumulating wealth and diversifying into non-farm activities in both rural and urban areas).
Please prepare a brief presentation of your study design, methodology, preliminary findings and next steps.
The APRA Work Stream 2 studies aim at exploring the dynamic impacts of commercialisation on livelihoods over a longer period of time, understanding the pathways and drivers of different livelihood trajectories. In particular, WS2 is interested in providing evidence on the transition of households into different livelihoods as a result of commercialisation. These livelihood trajectories include dropping-out (moving out of agriculture due to destitution and other challenges), hanging-in (subsistence agriculture for survival), stepping-up (expanding their existing agricultural activities with the aim of increasing production and commercial activities) and stepping-out (accumulating wealth and diversifying into non-farm activities in both rural and urban areas).
24th October to 13th November--- round one 26 days; round two 19 days
FGD with male and female club and non-club members separately
25 Original HH not tracked
Total of 519 quantitative interviews
Initial estimate of 829 HH members above 18 years of age// 303 represent 37%
Engagements – Agribusiness conference; Meetings by CISANET, MAAFAS,
24th October to 13th November--- round one 26 days; round two 19 days
FGD with male and female club and non-club members separately
25 Original HH not tracked
Total of 519 quantitative interviews
Initial estimate of 829 HH members above 18 years of age// 303 represent 37%
Engagements – Agribusiness conference; Meetings by CISANET, MAAFAS,
‘ we can do nothing apart from farming’
Refusals – non response to phone call; keep changing location of interview; not available at agreed times; husband refusing wife to particpate