Sampling Strategies of ILSSI Baseline Surveys and Key Modules

SAMPLING STRATEGIES OF ILSSI BASELINE SURVEYS AND KEY MODULES
Data were collected from 439 households in Ethiopia, 450 households in Tanzania, and 902 households
in Ghana using the following modules: 1) crop and livestock inputs, production and practices; 2)
household and women’s dietary diversity; 3) child health, diet, feeding and anthropometry; 4)
household shocks, assets and credit; and 5) the Women’s Empowerment in Agriculture Index (WEAI). In
Ghana additional modules were added for credit, food consumption and farmer networks.
ILSSI sampling strategy in Ethiopia
The baseline survey in Ethiopia was implemented by the Association of Ethiopian Microfinance
Associations (AEMFI) during November 14th - December 26th 2014, covering the previous production
year.
The sample for the household survey is drawn from the 4 woredas in which IWMI/ILRI interventions on
small-scale irrigation are taking place: Dangila and Bahir-Dar Zuria in Amhara, Lemo in SNNPR, and
Adami Tulu in Oromia. For each of these woredas, we obtained a list of kebeles with data on population
size (using CSA data) and suitability for irrigation based on the ex-ante suitability analysis conducted for
the AWM Solutions project. The methodology for small-scale irrigation analysis is described in Xie et al.
(2014); this article describes how the average suitability scores were calculated for all rural kebeles in
these 4 woredas (see Figure 1).
We selected 15 kebeles from within the chosen woredas. In order to determine the number of kebeles
to select per woreda, we first calculated the total population size (number of households) across all the
kebeles with a suitability score of 30 or greater in each woreda. We then determined the number of
kebeles to select from each woreda based on the population contained in areas with a high suitability
score. This resulted in the following breakdown:
Dangila
Bahir Dar Zuria
Adami Tulu
Lemo
2
7
3
3
In other words, the number of kebeles selected per woreda depended on the size of the population
living in high potential areas within that woreda. Within the woreda, the intervention kebeles, (which
were not randomly selected) were included the sample. The remaining kebeles were drawn randomly
with probability proportional to size. This resulted in the following sample of kebeles:
Kebeles
Adami Tulu (4)
1 Edo Gojola
Mean
No. of
suitability score households
67.48
603
2 Bochesa*
Bahir Dar Zuria (9)
1 Yegoma Huletu
2 Robit*
3 Wejir Welda Menta
4 Tis Abay Town
5 Meshenti Town
6 Wegeligo
7 Gomibat Aba Gerima
Dangila (3)
1 Gumbri Abela Akana
2 Ligaba
3 Dangishta*
Lemo (3)
1 Ajo Teasa
2 Digba
3 Upper Gana*
58.62
474
58.89
52.66
50.94
49.82
47.17
44.88
40.00
1,997
1,916
1,166
4,369
1,101
970
1,235
44.77
43.12
?
1,236
667
?
42.09
36.63
34.01
565
738
609
*intervention site
Within each of the chosen kebeles, we obtained a list of households from the local extension office with
an indication of whether the household uses irrigation or not. We randomly selected 10 households
from the list of irrigators and 10 households from the list of non-irrigators for a total of 20 hhs per
kebele. Given that we are sampling 15 kebeles, this gives a sample size of 300 households. We included
in the sample all households participating in the intervention within the intervention kebeles (142
households). This brings the total number of households to 442.
NOTES: The kebele selected for the IWMI/ILRI intervention in Dangila did not appear on our GIS shape
file or in the population data provided by CSA so we do not have these data for this site. Some of the
intervention households originally identified by IWMI and partners changed after the baseline survey
was conducted. Therefore, some households originally sampled as intervention households ended up
not participating in the program.
Figure 1: Location of the Ethiopia ILSSI baseline survey
ILSSI sampling strategy in Tanzania
The baseline survey in Ethiopia was implemented by Sokoine University during June 24th – July 11th,
2015, covering the previous production year.
The sample for the household survey is drawn from the 2 districts in which IWMI interventions on smallscale irrigation take place: Kilosa and Mvomero districts in Morogoro region. For each of these districts,
we obtained a list of villages with data on population size (using the 2012 Population census data from
Tanzania National Bureau of Statistics) and suitability for irrigation (at the ward level) based on the exante suitability analysis conducted for the AWM Solutions project (Xie et al. 2014) and later refined for
Cervigni and Morris (2015).
In addition to the 2 intervention villages (one in each district), we randomly selected 12 additional
villages from within the 2 districts in the following way: We first determined the number of wards in the
district with an average irrigation suitability score of 30 or greater. We then calculated the total
population size living within high irrigation potential wards in each district. We determined the number
of villages to sample from each district based on the proportion of households living in areas with a high
suitability score. This resulted in the following breakdown:
District
Kilosa
Mvomero
Total
No. of villages
randomly selected
7
5
12
Total number of
villages
8
6
14
In other words, the number of villages selected per district depended on the size of the population living
in high potential areas within that district. Within the district, the selected number of villages was drawn
randomly with probability proportional to population size, with the exception of villages in which the
IWMI/ILRI interventions take place. The intervention sites were not selected at random but based on
scoping work done by IWMI/ILRI. Adding the two intervention sites brings the total number of villages in
the sample to 14. This resulted in the following sample of villages:
Districts/villages
Kilosa (8)
Chanzulu
Idete
Kimamba 'A'
Kibaoni
Kondoa
Madudumizi
Suitability Number of
score
households
35.87
35.87
67.25
52.27
52.27
47.24
3,617
1,735
6,076
2,218
1,564
3,882
Comment [RC(1]: Was ETH also refined
through the WB drylands study? If yes, we
need to mention this here.
-And in the end, only one method could have
been used, either AWM OR WB drylands..,
unlikely that a mix of two were used.
Comment [BE(2]: I sent a separate
attachment describing the methodology
Zombolumbo
Rudewa Mbuyuni*
Mvomero (6)
Magali
Mangae
Changarawe
Lubungo 'B'
Tangeni
Mkindo*
*intervention site
47.24
12.49
3,068
3,627
41.50
41.50
33.85
33.85
33.85
27.06
2,263
2,543
5,271
1,969
5,386
6448
Within each of these villages, we obtained a list of households from the local extension office with an
indication of whether the household uses irrigation or not. We randomly selected 14 households from
the list of irrigators and 14 households from the list of non-irrigators for a total of 28 hhs per village. In
each of the 2 intervention villages we have a considerable number of households identified (selection
not random but pre-determined) to participate in the program. In these villages all intervention
households are identified as the irrigators and we interviewed an additional 15 non irrigating
households (selected randomly) per village.
This results in the following breakdown of households:
District
Ward
Village
Kilosa
Chanzuru
Chanzulu
Kilosa
Chanzuru
Idete
Kilosa
Kimamba 'A'
Kimamba 'A'
Kilosa
Mabwerebwere Kibaoni
Kilosa
Mabwerebwere Kondoa
Kilosa
Zombo
Madudumizi
Kilosa
Zombo
Zombolumbo
Kilosa
Rudewa
Mvomero Melela
Rudewa Mbuyuni*
Magali
Number
of hhs
Breakdown of Households
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
14 irrigators, 14 non-irrigators, all
28 randomly selected
43 intervention households (see list), 15
58 non-irrigators (randomly selected)
28 14 irrigators, 14 non-irrigators, all
Mvomero Melela
Mangae
28
Mvomero Mzumbe
Changarawe
28
Mvomero Mzumbe
Lubungo 'B'
28
Mvomero Mzumbe
Tangeni
28
Mvomero Hembeti
Mkindo*
56
450
Figure 2: Location of the Tanzania ILSSI baseline survey
randomly selected
14 irrigators, 14 non-irrigators, all
randomly selected
14 irrigators, 14 non-irrigators, all
randomly selected
14 irrigators, 14 non-irrigators, all
randomly selected
14 irrigators, 14 non-irrigators, all
randomly selected
41 intervention households (see list), 15
non-irrigators (randomly selected)
ILSSI sampling strategy in Ghana
Comment [RC(3]: The sampling strategy is
not clear.
The baseline survey in northern Ghana was conducted by the University for Development Studies (UDS),
Tamale, Ghana from early November of 2015 to early February, 2016. UDS researchers and enumerators
worked closely with an IFPRI representative and survey consultant, Mr. Jacob Thompson in the field data
collection exercise in November 2015 to implement the household baseline survey including
questionnaires for household survey, women empowerment in agriculture index (WEAI), time allocation
and community questionnaire in 12 communities in Savelugu Nanton District in the Northern Region,
Kassena Nankana East, Garu Tempane and Nabdam Districts in the Upper East Region. The total number
of respondents interviewed in 12 communities was 902.
The ILSSI baseline survey was carried out in four districts in the Northern part of Ghana. Out of these
four districts, communities with irrigation projects (dams) were the main focus of the research, with
three districts (Savelugu Nanton, Kassena Nankana East, Garu Tempane and Nabdam) being ILSSI
communities whereas Garu-Tempane District was an iDE intervention area. The ILSSI intervention
communities are Bihinayelli in the Savelugu District, Zanlerigu in the Nabdam District, and
Dimbasinia/Nyangua in the Kasena Nankana East District. Nine additional communities were surveyed
from the Garu Tempane District: Gbenterago Alemgbek, Akara, Bugri Natinga, Binpiala, Denegu, Zule,
Mognoori, Yidigu and Asikiri.
Table 1: Project communities and number of respondents surveyed
COMMUNITY NAME
COMMUNITY ID
GBENTERAGO ALEMGBEK
AKARA
2
1
NUMBER
OF
INTERVIEWED
145
157
BUGRI NATINGA
BINPIALA
DENEGU
4
3
7
35
21
39
ZULE
MOGNOORI
YIDIGU
ASIKIRI
ZANLERIGU
DIMBASINIA/NYANGUA
BIHINAYELLI
TOTAL
5
6
8
13
11
9
10
-
20
280
52
57
39
40
17
902
RESPONDENTS
The sampling frame in Ghana is somewhat different than in Ethiopia and Tanzania because of the
inclusion of an additional intervention involving a collaboration between IFPRI and iDE. We therefore
surveyed all ILSSI intervention households as well as households participating in an IFPRI-iDE
intervention and controls. All ILSSI intervention households that IWMI, ILRI, and NCA&T are working
with Savelugu Nanton District in the Northern Region, Kassena Nankana East, Nabdam Districts in the
Upper East Region are included in the sample irrespective of how these households were selected to be
part of the intervention. These are 102 households across the three districts.
In Garu Tempane district, IFPRI collaborated with iDE-Ghana to introduce water extraction technologies
(motor pumps) in 9 selected villages in Northern Ghana identified by iDE-Ghana. The sampling frame
involves an experimental design in order to identify the impacts of the introduction of motor on
nutrition and health. The intervention is phased in over time so that 4 villages received the intervention
in 2015 (early treatment villages) and 5 villages will receive the intervention in 2017 (late treatment
villages). All villages selected meet the following criteria: 1) they are located within the Feed the Future
zone of influence, 2) they are near iDE areas of ongoing operation, and 3) they show significant potential
for irrigation based on the irrigation suitability score.
Following an introduction of the program to village leaders and then the community at large by iDE,
farmers were asked to self-organize into groups for the purpose of receiving a group loan from a microfinance institute. Farmers were first organized into “confidence” groups of 20 farmers in order to
receive group loans from a micro-finance institute. The confidence groups are further broken down into
smaller “trust” groups of approximately 5 farmers each. In total 158 trust groups formed which included
a total of 800 farmers. These groups received training from iDE on group dynamics and micro-finance.
iDE then linked farmers with a local micro-finance institute (MFI) or credit union to provide credit for the
purchase agricultural inputs, such as fertilizer and seeds. IFPRI (through iDE) offered an additional loan
to purchase a motor pump at a favorable interest rate to a random sub-sample of groups.
The identification of early and late treatment villages was not known at the outset of the project when
iDE first contacted the communities to form groups. Because the size of the villages and number of
groups formed in each village were not uniform across all 9 communities, IFPRI randomly selected
communities using probability proportional to size until the number of farmer groups were roughly
equal between early and late treatment villages.
Within the early treatment villages, a sub-sample of trust groups received an additional preferential
credit incentive to purchase a motor pump through a random lottery. Given the liquidity constraints that
micro-finance institutions in Ghana face, IFPRI provided guaranteed loan access using supplementary
funding provided by the Water, Land, and Ecosystems Research Program of the CGIAR (WLE), and the
loans were managed by iDE. These loans were to be used for the purchase a motor pump and were
conditional on the understanding that the pump would be shared amongst members of the trust group
in a predetermined schedule that is agreeable by all the group members. Groups that randomly receive
the loan offer through the lottery could decide not to purchase a pump and still apply for a loan for
other inputs through the micro-finance institute. On the other hand, the control group of farmers who
didn’t get the loan offer could still purchase pumps from the market either using their own money or
getting the loan from the MFI or any other financial institution.
Farmer groups are expected to repay the full cost of the pump to iDE at a preferential interest rate of
20%. The repaid loans will be used to extent the loan offer to: (i) groups that did not win the lottery in
Comment [BE(4]: Maybe again link to
methodology
round 1 in the early treatment villages and (ii) a random subset of groups in the late treatment villages
in 2017. Groups that did not win the lottery in round 2 in the late treatment villages would receive the
loan offer in 2018 at the end of the project.
This design allows a comparison between those in the early treatment villages who won the lottery and
those in late treatment villages, as well as between those who are in the early treatment villages but
didn’t win the lottery and those who are in the late treatment villages. The first comparison allows us to
capture the direct effect of improved access to irrigation technology on those who got the technology,
while the second comparison enables identification of any spill-over effects of living in a community
where some people have access to irrigation technology.
Farmers that did not receive preferential loans for the purchase of motor pumps in both the early and
late treatment communities involved in the iDE experiment can also be used as controls for the ILSSI
intervention farmers. Because these farmers already expressed interest in irrigation and improving
agricultural production practices in general, this minimizes the selection bias due to fact that ILSSI
households self-selected into the ILSSI intervention. All the communities sampled in Ghana are shown in
Figure 3, distinguished by their classification as an ILSSI intervention community, iDE early treatment
community, or iDE late treatment community.
Figure 3: Location of the Ghana ILSSI baseline survey
KEY MODULES FOR THE ILSSI BASELINE SURVEY
Intra-household surveys: Irrigation and women’s empowerment
ILSSI is using the Women’s Empowerment in Agriculture Index (WEAI) to measure the relationship
between women’s empowerment and irrigation and how women’s empowerment influences nutrition
outcomes. The WEAI is a survey-based tool, asked of both the main male and female decisionmakers in
a household used to determine inclusion of women in domains important to the agricultural sector. It
takes about 40 minutes to complete.
There are multiple domains of empowerment. Whereas previous measurements were only able to
measure individual domains, the WEAI has the advantage of measuring five domains that are important
in the agricultural sector. The five domains of empowerment in the agricultural sector measured in the
WEAI include:
•
Production: decisions about agricultural production, including sole or joint decisionmaking
power over food or cash-crop farming, livestock, and fisheries, as well as autonomy in
agricultural production
•
•
•
•
Resources: access to and decisionmaking power over productive resources, including ownership
of, access to, and decisionmaking power over productive resources such as land, livestock,
agricultural equipment, consumer durables, and credit
Income: sole or joint control over income and expenditures
Leadership: Leadership in the community, including membership in economic or social groups
and being comfortable with speaking in public
Time: allocation of time to productive and domestic tasks and satisfaction with the time
available for leisure activities
These are measured through 10 individual indicators in the survey based tool, and weighted using the
weighting scheme listed on the right in Table 1. In addition to the domains, the WEAI calculated score
also includes the Gender Parity Index. This reflects the percentage of women who are as empowered as
the men in their household. This component takes into account the male counterpart’s responses to the
10 indicators and calculates a how many women achieve parity with their husband, and for those who
do not, how great is the gap of inadequacy.
Specifically, ILSSI is using a modified WEAI to better capture linkages between irrigation and gender. In
some cases, questions were added to distinguish between irrigated and rainfed production, e.g. in the
case of decisionmaking roles and autonomy in decisionmaking. In addition, response codes or
categories were added to capture specifics about irrigation. For example, the module on productive
capital includes categories for irrigation equipment and water storage. Similarly, the time allocation
module adds a category for time spent irrigating or working with irrigation equipment. Other
modifications to the WEAI include additional modules or questions on credit, savings, and group
membership.
Anthropometric Measurements
In taking the anthropometric measurements, the specific eligibility criteria for respondents to be
measured were emphasized to enumerators both during the pre-survey training and in the field. The
criteria are restated follows:
Children less than five years old (i.e. up to 59 months) were eligible for height, weight, head
circumference and mid arm-upper circumference (MUAC) measurements. It also included Oedema
assessment.
Women within their reproductive age (between 18 and 50 years) were also eligible for weights and
heights measurements.
Community Interviews
The survey also included a community questionnaire which was conducted through interviews with key
informants comprising of village chief, village chairmen, community elders, opinion leaders and other
important personalities in the community. The community survey was carried out to provide a
comprehensive and in-depth understanding of information relative to community physical infrastructure
and endowments such as schools, hospitals, clinics, markets, lorry parks, means of transport, water
sources for domestic and agricultural uses, crop grown, animals reared, irrigation practices, diseases,
illnesses, as well as agricultural and health support services among others at the village level. One
community survey was administered in each community, thus, 12 community surveys for the 12 ILSSI
project communities.
REFERENCES
Cervigni, R. and M. Morris (eds.) 2015. Confronting Drought in Africa’s Drylands: Opportunities for
Enhancing Resilience. Africa Development Forum series. Washington, DC: World Bank. License: Creative
Commons Attribution CC BY 3.0 IGO.
Xie, H., L. You, B. Wielgosz and C. Ringler. 2014. Estimating the potential for expanding smallholder
irrigation in Sub-Saharan Africa. Agricultural Water Management. 10.1016/j.agwat.2013.08.011. 131(1):
183–193.