2. Presentation outline
1
• Quick recap on GAGE and analysis approach
2
•Findings for Amhara
3
•Findings for Oromia
4
• Findings for Afar
5
• Next steps in terms of research
5. Research Locations
Zones selected based on:
• high prevalence of child marriage
(as proxy for conservative gender
norms) and then
• districts that are designated food
insecure (as proxy for econ
vulnerability)
6. Context diversity is critical in understanding programming
starting points
School enrolment varies tremendously—and in different ways:
• In South Gondar—65% of girls and 58% of boys would like to attend university
• In East Hararghe—51% of girls and 65% of boys would like to attend university
Adolescents’ aspirations are high but also diverge across regions by gender:
• Only 63% of teens are enrolled
• Young adolescents have missed 24% of school days in the last two weeks
• Average grade attainment is only 2.3 years
Zone 5, Afar educational disadvantage is stark:
In East Hararghe
• 85% of boys and 69% of girls are
enrolled
• Boys have completed a half year
“more” school than girls” 5 vs 4.5
In South Gondar:
• 96% of girls and 89% of boys are enrolled
in school
• Girls have completed a half year “more”
school than boys: 4.8 versus 4.3
Figures
from GAGE
baseline
7. Gender norms and practices related to CM and FGM
are also highly divergent across regions
Child marriage is shifting differently
across regions:
• In Amhara, the age of marriage
climbed from 14.4 to 16.2 between
2005 and 2016
• In Oromia, it was comparatively
unchanged, 17.1 to 17.4
• In Afar, it dropped from 16.7 to 16. 4
FGM/C varies in incidence and type
across regions:
• Nearly all girls under 15 are cut in Afar—
70% are infibulated
• “Only” half of girls are cut in Amhara,
3% are infibulated
• Infibulation is also rarer in Oromia (7%),
where girls are cut later
GAGE’s baseline highlights why:
• In Amhara, child marriages are mostly arranged, and families are committing to school as
farming becomes less tenable.
• In Oromia, marriages are increasingly initiated by adolescents themselves, and the age
of marriage is dropping in many places.
• In Afar, the absuma marriage system dictates to whom and when girls must marry.
Figures from
DHS
9. Quantitative sample
4,518 VYAs in 173 kebeles
Kebeles within woredas selected
based on mapping of access to
markets, infrastructure and
services (high and low)
Amhara: 1,113 girls and 819 boys
(74 Kebeles)
Oromia: 1,190 girls and 878 boys
(79 kebeles)
Afar: 298 girls and 220 boys
(20 kebeles)
Aged 10-12 during GAGE
baseline data collection
(late 2017; midline late 2019
8 months after programming
start)
Randomly selected from a door-to-door
household listing in communities
Randomly assigned kebeles to different
programming arms (and control) with
implementation starting in early 2019
10. Qualitative sample
1 kebele per intervention arm
and 1 control per region
(i.e. 5 kebeles in Amhara and
Oromia and 2 in Afar)
275 adolescents (179 HS or AWH
participants)
Inclusion of socially disadvantaged
adolescents based on disability
(10%) and marital status (10%) but
largely non-programme participants
Amhara:
• 112 adolescents (77 HS or
AWH participants)
• 73 parents
• 10 mentors
• 34 key informants
Oromia:
• 112 adolescents
• (77 HS or AWH participants)
• 87 parents
• 13 mentors
• 33 key informants
Afar:
• 51 adolescents (25 AWH
participants)
• 48 parents
• 7 mentors
• 10 key informants
14. Analysis approach for today (1)
• Her Spaces vs. Control (Arm 1 vs. Control)
• Combined treatment of AWH-E (curriculum) and AWH-E (curriculum + systems) vs.
control. (Arm 2+3 vs. control) Will call this AWH-E in this presentation.
• The text notes key outcomes where impacts of Her Spaces and AWH-E are different.
Briefs focuses on two main treatment control comparisons:
• Asset transfer really changed the intervention (in both good ways and bad) so
we look at it in comparison to the other two AWH-E arms separately in the text.
• Community level social norms and systems strengthening activities were 25%
complete at the time of this survey; and quantitative findings show limited
difference in impact on any outcomes between Arm 2 and Arm 3.
Why this approach?
15. Analysis approach for today (2)
• All adolescents and parents in our study who live in an AWH community are
considered treated, whether they were part of the program or not.
• Generally considered the policy relevant outcome (as uptake of these programs
is always less than 100%, and often much lower)
These are intent to treat estimates:
• Interest in the program was high across regions (much higher than generally
seen with these programs)
• Still manually entering attendance data to give exact numbers on enrollment
and attendance (time consuming!!)
Future quasi-experimental analysis will look at treatment on the treated
(e.g. are there larger impacts for those who actually participated in AWH).
All quantitative regressions control for a set of baseline covariates, use
appropriate sample weights, and cluster standard errors at kebele level.
16. Analysis approach for today (3)
Qualitative data
was thematically
coded in MAXQDA
and disaggregated by
region, gender and
by participant status.
Qualitative data
used to triangulate
findings from quant
survey, to provide
context specificity to
findings, and to
disentangle potential
pathways to explain
positive, negative or
non-effects.
Key informant
interviews
provided additional
understanding of
community, district
and zonal political,
economic and social
dynamics.
Quotes
are used to illustrate
findings; where they
represent the voices
of outliers this is
noted.
18. Amhara: context snapshot
•South Gondar communities include both highland and lowland communities
•(different livelihood structures and migration opportunities)
Baseline but especially midline saw rising levels of unrest, translating into
escalating risks of violence and sexual violence in study communities
Overall, relatively stronger presence of Women’s Development Army, history of
girls’ education, girls’ clubs in schools
Gender norms: child marriage is declining but where it happens tends to be arranged
by parents; FGM is practiced in infancy, Type 1
20. Baseline characteristics (control)
Overall
HH size 5.653
HH head literate 0.375
HH currently receives PSNP benefits 0.330
Girls Boys
Age 10.968 10.981
Enrolled in school during most recent
session
0.975 0.899
Reported having control over money in past
12 months
0.092 0.082
Has savings 0.039 0.039
Has not experienced or witnessed HH
violence in last 12 months
0.343 0.277
21. Girls’ Primary Indices (subset)
Violence
Health and
Nutrition
SRH
Voice and
Agency
Economic
Empowerment
Gender
Equitable
Attitudes
Knowledge
Her Spaces 0.081 0.028 -0.257 0.197** 0.179 -0.062 0.283***
(0.078) (0.086) (0.344) (0.094) (0.127) (0.090) (0.101)
AWH (no assets) -0.004 0.012 0.141 0.291*** 0.280** 0.193** 0.311***
(0.078) (0.078) (0.394) (0.076) (0.108) (0.076) (0.080)
P-value: B1 /= B2 [0.285] [0.862] [0.196] [0.193] [0.404] [0.002] [0.774]
Control Mean -0.040 0.177 0.353 0.080 -0.006 0.246 0.270
Observations 826 844 65 814 795 834 821
22. Girls’ secondary outcomes (knowledge, subset)
Knowledge
Index
girls reach
puberty first
menstruation
frequency
menarche
allows
pregnancy
legal age
marriage
girls
legal age
marriage
boys
FGMC has
risks
where to
get help for
violence
where to
keep money
negotiation
skills
Her Spaces 0.283*** 0.024 0.178*** -0.024 0.081** 0.036 0.079 0.164*** 0.017 0.030
(0.101) (0.038) (0.047) (0.036) (0.031) (0.022) (0.068) (0.060) (0.037) (0.036)
AWH (no assets) 0.311*** 0.014 0.158*** -0.054* 0.110*** 0.028 0.136*** 0.142*** 0.026 0.065**
(0.080) (0.036) (0.039) (0.028) (0.028) (0.018) (0.049) (0.045) (0.027) (0.029)
P-value: B1 /= B2 [0.774] [0.801] [0.628] [0.345] [0.409] [0.721] [0.345] [0.686] [0.802] [0.311]
Control Mean 0.270 0.287 0.584 0.810 0.210 0.050 0.510 0.269 0.887 0.162
Observations 821 845 844 845 844 845 832 841 841 845
23.
24.
25. 21% Less likely to agree that
‘gender roles can be changed
38% Less likely to disagree that
‘girls are expected to be humble’
No change in identifying
negotiation skills
No change in knowing that
menarche allows for pregnancy
8% more likely to
indicate that they
did not perpetrate
peer violence in
the past 12
months
7% More likely to
agree that ‘boys
should be able to
show feelings’
Positive effects: Negative effects:
No effects:
27. Discussion points
• Do the SG findings resonate with implementer experiences/ perceptions? What
was surprising if anything – positive or negative?
• Given there is always a gap between changes in individual knowledge and
attitudes, what more from a programming perspective are you planning to do to
bring about transformation?
• E.g. major challenge of sexual violence risks in SG – how is this being factored
into programming going forward?
29. Oromia: context snapshot
Ongoing insecurity in East Hararghe esp at baseline, but also midline high proportion
of very vulnerable IDPs in communities and active presence of qeerroo youth
movement in some communities.
Khat commercial crop farming dominates lives and because it is relatively
lucrative often truncates adolescent aspirations
Under-investment in zone historically is reflected in poorer quality services and
infrastructure.
Gender norms: child marriage tends to be adolescent-initiated (often via Shegoyeh
traditional dance and/or brokers); FGM is widespread and often carried out in early
adolescence, Type 2 and 3
30. Oromia: a bit more on the research sample
Quantitative: 1,190 girls and 878 boys (79 kebeles)
Qualitative:
• 112 adolescents,
• 87 parents,
• 13 mentors,
• 33 key informants
• Implementers able to contact 91% of GAGE sample; 98% interested in programming.
• Successfully surveyed 83.2% of the sample at midline
• Lower than Amhara; largely due to conflict limiting access to locations
• Not differential by treatment, nor based on baseline characteristics
31. Baseline characteristics (control)
Overall
HH size 6.986
HH head literate 0.263
HH currently receives PSNP benefits 0.362
Girls Boys
Age 10.835 10.995
Enrolled in school during most recent
session
0.710 0.885
Reported having control over money in
past 12 months
0.155 0.220
Has savings 0.004 0.011
Has not experienced or witnessed HH
violence in last 12 months
0.282 0.238
32.
33. 46% less likely to disagree
that a ‘woman should
obey her husband’
17% less likely to disagree
that boys who behave
like girls are weak
No change in knowing
that gender roles can be
changed
14% more likely to
know of a place to go
for help from violence
38% more likely to
know that FGM/C has
risks
However, 21% less
likely to believe that
boys are not
biologically smarter
Positive effects: Negative effects:
No effects:
34. Mentors
Some positive gains – especially in terms of changes in attitudes and practice re gender division of
labour in the hh
However, also high levels of dissatisfaction reflected in higher turnover than elsewhere,
uneven attendance, some disbanding of groups prematurely
Concerns expressed by mentors included limited support from supervisors, limited site visits
(due to both distance –tended to live in Harar and challenging to visit districts during unrest),
yet significant resistance and backlash from some communities esp. due to SRH content and
fears about religious conversion.
Communication around girl-only asset sparked considerable discontent among boys, parents,
community leaders – in sites with the asset transfer as well as neighbouring communities, and was
exacerbated by over-promising by some mentors
35. Discussion Points
• East Hararghe’s economic and gender norm dynamics require very tailored
approach – as programming expands how is this being addressed?
• District officials face very constrained budget environment and also capacity
limitations (esp. given re-shuffling of public sector personnel) – how is
Pathfinder/Care factoring in local political economy dynamics to the systems
strengthening work?
37. Afar: context snapshot
Zone 5 is politically marginalized
in the region which exacerbates
service deficits; some very
remote communities which are
seldom visited by district officials
and have no NGO presence.
During baseline and midline
serious clan conflicts as well as
conflict with the Argobba ethnic
group in a neighbouring district
make difficult terrain even less
accessible, and also risky for
adolescents.
Drought conditions
experience at baseline had
much improved, as had local
food security - but rapidly
rising rates of migration of
adolescent girls to Djibouti
and MENA.
Major change in midline in civil service cadre
– new generation, with greater competencies
and vision for change.
Gender norms:
• absuma (mandatory maternal cousin) marriage
system;
• Type 3 FGM and sometimes infibulation carried
out as young as infancy through to early
adolescence
38. Afar: sample
• Smaller sample size. This was always meant as a pilot from the research POV. Can we
successfully conduct longitudinal mixed methods research in remote pastoralist settings.
• Implementers able to contact 84% of GAGE sample; 95% interested in programming.
• Successfully surveyed 81.5% of the sample at midline
• Not differential by treatment, nor based on baseline characteristics
• Just a very hard to reach population!
Research sample:
• Quantitative: 298 girls and 220 boys (20 kebeles)
• Qualitative:
• 51 adolescents, 48 parents, 7 mentors, 10 key informants
39. Baseline characteristics (control)
Overall
HH size 6.320
HH head literate 0.053
HH currently receives PSNP benefits 0.398
Pastoralist HH 0.877
Girls Boys
Age 10.952 10.762
Enrolled in school during most recent
session
0.713 0.536
Did not miss a week of school in the
past year (among enrolled)
0.656 0.756
Has not experienced or witnessed HH
violence in last 12 months
0.575 0.56
Has an adult s(he) trusts 0.464 0.337
40.
41.
42. 45% less likely to disagree
that a ‘woman should obey
her husband’
20% less likely to agree that
‘boys and girls should share
HH tasks equally’
No change in knowing
where to keep money
No change in knowing that
menarche allows for
pregnancy
Positive effects: Negative effects:
No effects:
29% more likely to
know about the risk
of FGM/C
82% more likely to
know of a place to
go for help from
violence
46. Ways forward for research…..
1
Update briefs based on today and any other written feedback received by Monday COB.
• Regional dissemination starts next Friday (April 2)
2
Journal articles
• 1-2 articles on short run findings focused on specific domains. Main overall paper will wait for latest
round of data collection. Possible paper on COVID-19 (but ability to reach rural sample by phone
was limited)
3 Hoping to collect second midline data pre-election (i.e. now; but COVID is spiking)
4
Finalizing (with Pathfinder) proposed design for Older Cohort evaluation, which includes
urban programming in Debre Tabor (GAGE has increased its sample to support robust
evaluation in urban area
5
Exploring possibility of adding new VYA cohort to evaluate second round of AWH-E
programming with VYA.
47. Contact Us
WEBSITE
www.gage.odi.org
TWITTER
@GAGE_programme
FACEBOOK
GenderandAdolescence
About GAGE:
Gender and Adolescence: Global Evidence
(GAGE) is a nine-year (2015-2024) mixed-
methods longitudinal research programme
focused on what works to support
adolescent girls’ and boys’ capabilities in
the second decade of life and beyond.
We are following the lives of 20,000
adolescents in six focal countries in Africa,
Asia and the Middle East.