Educational inequality in secondary schools in three developing countries
Rhiannon Moore & Bridget Azubuike
CEID Launch Symposium
UCL Institute of Education, 15 June 2017
2. EDUCATIONAL INEQUALITIES AT SECONDARY
Secondary education ‘critical’ to
breaking intergenerational
transmission of poverty (World
Bank, 2009)
Inequalities in access - Unequal
transition to secondary reinforces
exclusion of disadvantaged groups
in labour market (DFID, 2017)
Inequalities in learning - A high
degree of inequalities in test
scores at secondary level suggest a
high degree of wage inequality in
the future (Nickell, 2004, in Das &
Zajonc, 2010)
3. YOUNG LIVES & LINKED SCHOOL SURVEYS
Longitudinal survey of children,
their households, schools and
communities running for 15 years
in 4 countries
Young Lives school surveys:
introduced in 2010 with a sub-
sample of YL children and peers
2016-17 school surveys: school
effectiveness in Ethiopia, India and
Vietnam
• Ethiopia: upper primary (Grades 7-8)
• India: lower secondary (Grade 9)
• Vietnam: upper secondary (Grade 10)
School effectiveness design –
‘value-added’ of 1 year of school
5. CROSS-COUNTRY ANALYSIS: A COMMON SCALE0
.002.004.006
300 500 700 900
Maths Score
YL India YL Vietnam YL Ethiopia
W1 Maths Performance by YL Country First step –
putting country
data on a
common scale
with a mean of
500 and SD of 100
On this common
scale, the
countries rank in
the order we
would expect
But there is also a
lot more overlap
than we might
have anticipated –
similarities as
well as
differences
Ethiopia India Vietnam
Mean Maths Score
(start of school year)
418 489 586
6. CROSS-COUNTRY: WHO ARE THE HIGH ACHIEVERS?
214
508
196
568
289
703
319
554
347
618
432
760
341
756
376
796
461
834
BOTTOM TOP BOTTOM TOP BOTTOM TOP
ETHIOPIA INDIA VIETNAM
Maths score distribution in top and bottom
deciles
Min Mean Max
Big learning inequalities
within countries.
Children in the top
decile of the test score
distribution on average
have more educated and
literate parents.
Located in urban areas
in India, Addis Ababa in
Ethiopia and Hung Yen in
Vietnam.
More girls in the top
score deciles compared
to the bottom in
Vietnam.
8. CROSS-SECTIONAL DATA REVEALS INEQUALITIES
Large differences in
learning attainment in
the India sample at
the start of Grade 9.
E.g. by household
wealth
Reflects different
experiences prior to
Grade 9…
• E.g. type of school
attended, home
background, parental
education, access to
different educational
opportunities outside
school…
Q1
(poorest)
Q2 Q3 Q4 Q5 (least
poor)
Mean Maths
Score (start
of school
year)
439 460 484 506 540
0
.001.002.003.004
Density
300 500 700 900
Maths Score
Q1 (poorest) Q2 Q3 Q4 Q5 (least poor)
Maths Performance By Wealth Quintiles
9. LONGITUDINAL DATA SHOWS UNEQUAL PROGRESS
Gaps in test performance appear to be widening over the course of the
school year
E.g. the least poor students make an average of 43 points of progress,
while the poorest students make 22 points – falling further behind
0
.001.002.003.004.005
Density
300 500 700 900
Maths Score
Wave 1 - poorest quintile Wave 2 - poorest quintile
Wave 1 - least poor quintile Wave 2 - least poor quintile
Maths Performance By Wealth Quintiles
10. SCHOOL EFFECTIVENESS & VALUE-ADDED
Looking at the role which schools
play in reducing or exacerbating
learning inequalities.
‘Value-added’ used as a summary
measure of school quality
Value-added analysis: Learning
progress attributable to schools
and teachers after removing prior
attainment and background effects
Does not focus on the absolute
levels of attainment, but on how
much students have improved
during the school year, whatever
their initial learning levels were
11. DIFFERENCES IN THE VALUE ADDED BY SCHOOLS
-100
-50
0
50
100
0 50 100 150 200
School by Value-Added Rank
PA PUA SG TSW
Value-Added in Maths by School type (Conditional)
There are
notable
differences in
the value added
by schools in
the sample
Very large
differences in
VA between top
and bottom
performing
schools – more
than 200 points
12. WHO ATTENDS SCHOOLS THAT ADD MOST VALUE?05
101520
Never been to school Primary Secondary Upper Secondary Higher education
Mean School Value-Added (Maths) by Mother's Education
VA (unconditional) VA (conditional)
The least poor, boys, and those with
more educated parents more likely to
attend schools which add most value.
Suggests that the most advantaged
students are ‘sorted’ into ‘better’
schools, where they learn more.
This suggests that inequalities will
worsen over time
05
1015
Q 1 (poorest) Q 2 Q 3 Q 4 Q 5 (least poor)
Mean School Value-Added (Maths) by Wealth Quintiles
VA (unconditional) VA (conditional)
02468
Meanschoolvalue-added
Female Male
Mean School Value-Added (Maths) by Gender
VA (unconditional) VA (conditional)
13. DIFFERENCES IN LEARNING WITHIN THE CLASS?
VARIABLES Maths T1 Maths T2|T1
Age -4.8641 -1.7654
(-3.497)*** (-1.277)
Boy 18.2010 1.2039
(5.564)*** (0.361)
Schedule caste (other backward caste) -8.9434 -5.0366
(-3.543)*** (-2.395)**
Schedule tribe (other backward caste) -10.2707 -9.4542
(-2.098)** (-2.052)**
Mother’s education – secondary (none) 6.5140 -2.7507
(2.297)** (-1.018)
Mother’s education – above secondary (none) 7.9202 5.3102
(2.028)** (1.350)
Father’s education – secondary (none) -2.6847 4.1171
(-0.874) (1.532)
Father’s education – above secondary (none) 7.9542 9.5376
(2.260)** (3.139)***
Wealth index 2.4239 0.1321
(3.667)*** (0.219)
Education level expected 5.7996 2.8848
(9.959)*** (4.732)***
Ever dropped out -9.7908 -6.3534
(-3.077)*** (-1.964)*
Time spent on chores 3.7258 1.9222
(3.115)*** (1.749)*
Time spent on paid work -4.6575 -2.5270
(-2.967)*** (-1.410)
Constant 476.7230 555.2901
(21.992)*** (10.472)***
Observations 7,745 7,745
R-squared 0.045 0.267
Number of classes 307 307
r2_b 0.402 0.718
r2_w 0.0449 0.267
Class fixed effects
regression reports average
within-class effects across
307 classes in the survey.
Reveals that certain
student background
characteristics are
associated with learning
and progress over one year.
But many lose significance
once prior attainment is
controlled for.
FE provides further
evidence of ‘sorting’ into
classes - more variation
explained by differences
between classes/ schools
than within them.
14. DISCUSSION & IMPLICATIONS
Gaps between countries but also notable similarities in highest achievers
– despite the differences in education systems and country context.
Within the Indian sample, inequalities in learning outcomes are increasing
– serious implications for equality of opportunities after secondary school.
Evidence appears to suggest more advantaged children are ‘sorted’ into
‘better’ schools, even when the background of children is controlled for.
Most inequalities appear to be found between classes/schools, rather
than within them – suggests that improving the performance of lowest
performing schools is key to reducing inequalities.
On average, the poorest children start Grade 9 100 points behind the
least poor in maths – that’s equivalent to more than 3 average school
years. This gap widens further over the course of the school year.
The large gaps present at the beginning of Grade 9 suggests efforts to
equalise learning need to happen earlier – by secondary school,
inequalities are already heavily entrenched.