View
954
Download
0
Category
Tags:
Preview:
Citation preview
Building Panel Survey Systemsin Sub-Saharan Africa:the LSMS Experience
GERO CARLETTO
Lead Economist
Manager, Living Standards Measurement Study
Development Research Group
The World Bank
Florence, October 13-15, 2014
Development Research
Outline
• LSMS: a primer
• Panel surveys in SSA: LSMS-ISA
• Experience with Attrition and other practical considerations
• Some final thoughts
… will not talk about CAPI and its value added!
LSMS: a primer
• Established in 1980…– More than 100 surveys, plus TA in many more countries …
• Original impetus: Measure poverty but also analyze correlates, study household behavior & interactions w/ policies.
• Multi-topic design: One survey on a range of topics more powerful tool than a series of single purpose surveys– Consumption exp., Income, Education, Health, Labor, Nutrition, inter alia.
• Representative at national and sub-national level– But with option to “zoom in” via imputation (poverty and nutrition maps)
• Sampling unit: Household (consuming & producing unit)
• Demand driven; policy needs of each country key to survey design, yet with an eye to standardization
• Historically, emphasis on …– Capacity building, work with national statistical offices
– Open data access policy
LSMS-Integrated Surveys on Agriculture
• Objectives– Role of ag in poverty reduction
– “Rotating” platform for data and methodological imporvements
• 8 countries in SSA (with NSOs)
• LSMS cum …– Panel
– Agriculture+
– Individual/gender• Assets
• Farm and non-farm ownership
• Decision-making
– Geo-referenced• Dwellings
• Plots (including area)
Survey Schedule
Country Baseline Follow Up
Tanzania 2008/09 2010/112012/13
(Oct 2014)2014/15 2016/17
Uganda 2009/10 2010/11 2011/122013/14
(Dec 2014)2015 …
Malawi 20102013
(Oct 2014)2016 2018 2020
Nigeria 2010/11 2012/13 2015/16 2017/18
Ethiopia 2011/122013/14
(Dec 2014)2015/16 2017/18
Niger 2011 2014
Mali 2014/15 2016/17
Burkina Faso 2014/15 2015/16 2017/18
Tracking in the TZ NPS
• The NPS tracks every eligible household member.
• Eligibility is defined as being 15 years and above, excluding live-in servants.
• Mobile field teams are assigned target households and provided information on the location of the households and its members.
• Three separate forms are used to compile information about the location of the household /individuals from informed respondents.
• Contact information is collected for several informed respondents, relatives, facilities.
Tracking in the TZ NPS (cont’d)• Tracking is divided into two classifications; Local tracking and
Distance tracking.
• Households /individuals within the assigned areas of the mobile team are local tracking cases and are completed by the same team.
• Households /individuals residing outside of the assigned area are classified as distance tracking. Information on distance tracking cases is rapidly sent electronically to HQ for processing.
• Distance tracking cases are either assigned to a mobile team operating within the area or assigned to a designated tracking team.
• Periodically a “tracking round” is scheduled and the tracking team will circle the country seeking to locate clustered distance tracking cases not covered through standard protocol.
Tracking in the TZ NPS
NPSY1 to NPSY292% individual, 97% household
NPSY2 to NPSY392% individual, 97% household
Sample Size:NPSY1: 3265 households, 16,709 members
NPSY2: 3924 households, 20,559 members
NPSY3: 5010 households, 25,412 members
Movements NPSY1-NPSY2
Malawi Tracking: Protocol and Attrition• Same protocol used in Tanzania
• Tracked all individuals greater than 12 years of age at the time of follow-up excluding live-in servants, those that moved out of the country and those living in institutions
Households Individuals
Baseline Tracking Eligible
Sample3,246 10,769
Successfully Tracked 3,104 9,866
Died 20 229
Overall Attrition Rate 3.78% 6.39%
* 3.17% of baseline individuals migrated out of the country, moved to army barracks or police compounds so were NOT tracked. Excluding these individuals from the calculation results in an attrition rate of 3.26%
Malawi Tracking: Split-offs
Distribution of Baseline Households
By # of Splits Between 2010 & 2013Observation Percent
Contribution to
IHPS 2013 Sample
0 2,384 76.8 2,384
[2384*(1+0)]
1 574 18.49 1,148
[574*(1+1)]
2 123 3.96 369
[123*(1+2)]
3 17 0.55 68
[17*(1+3)]
4 5 0.16 25
[5*(1+4)]
5 1 0.03 6
[1*(1+5)]
TOTAL 3,104 100 4,000
Malawi Tracking: Fieldwork Organization
Field-based Mobile Teams
• 12 teams responsible for 12-17 Enumeration Areas
• Comprised of 1 supervisor and 4 enumerators
• Responsible for all local AND distance tracking cases falling within assigned districts
Tracking Team
• 1 tracking team
• Highly mobile team comprised of 1 supervisor and 4 enumerators
• Travel to assist teams with heaviest tracking workloads
• Spent majority of fieldwork in two largest cities
But many pending issues/trade offs …
• Ethiopia: resident enumerators, no split-off tracking (but not much movement).
• Nigeria– non-concurrent data entry of tracking
– large country, 36 states, some with major security-related issues (dropped a few EAs)
• Uganda– 2009/10 follow up of 2005/06 UNHS; 20% of
households only
– perverse incentives of Kampala-based teams and no separate tracking team (until recently!)
Some final thoughts …
• Heterogeneity in protocols and attrition
– Difference in results driven by protocol and incentives
– Difference in protocols driven by context and costs
• Use tracking manager to carefully monitor all tracking cases and make proper, rapid assignments to teams (and lower cost)
• Use separate tracking team to assist with more difficult cases in urban areas and help keeping equal workload
• Allow time for possible “tracking marathon” during fieldwork to focus solely on tracking cases and make contact with as many respondents as possible
Some final thoughts …
• Collect information from 3+ informants to best assist enumerators with tracking cases
• Enter tracking form together with data (no issue in CAPI, but also in CAFE)
• Verify phone numbers before leaving household, if respondent agrees– Provide incentive to call from respondent’s phone
• Equip teams with adequate airtime and fuel (SSA!)
Longitudinal studies within NSS: it’s not impossible!– Sustainability
– Cost-effectiveness (“unit” cost!)
– Seek integration and explore scope for re-purposing/add-ons• On-going collaborations with WHO, CGIAR, WFP, …
Selected resources
• Witoelar, Firman (2011) “Tracking in Longitudinal Household Surveys”, LSMS Working Paper
• Basic Information Documents
• www.worldbank.org/lsms
• www.worldbank.org/lsms-isa
• Solutions.worldbank.org
The End
Tracking households
John
JaneJessieJoshuaJennifer
12
111
403718168
12
333
11311
2
2
Complete a tracking form for the household
2
2
Tracking individuals
John
JaneJessieJoshuaJennifer
12
111
403718168
12
333
11311
2
2
Complete a tracking form for each member
Household Tracking Form
Household Tracking Form
Household Tracking Form
Individual Tracking Form
Individual Tracking Form
Individual Tracking Form
Recommended