Do some schools narrow the gap Differential school effectiveness by ethnicity, gender, poverty and p

1
Do some schools narrow the gap? Differential
school effectiveness by ethnicity, gender,
poverty and prior attainmentPaper presented to
the International Congress on School
Effectiveness and School Improvement, Vancouver,
Canada, 4-7 January 2009Dr. Steve
StrandAssociate ProfessorUniversity of
Warwicksteve.strand_at_warwick.ac.uk44 (0)24 7652
2197
2
Overview of the paper

• Definition of differential SE, why important,
  particular focus on ethnicity
• Briefly review previous literature
• Present empirical data from population study in
  England (530,000 pupils age 11 in over 14,000
  schools)
• Conclude no evidence of substantial
  differential school effectiveness by pupil
  groupings
• Consider implications for understanding
  differences in attainment between ethnic groups

3
Differential school effectiveness

• Well established that schools are differentially
  effective, in the sense that some are more
  effective than others
• But differential school effectiveness more
  particularly used to denote the effectiveness of
  a school over different subjects, over key
  stages, or with different pupil groupings (e.g.
  boys/girls, ethnic majority/minorities, low/high
  SES etc).
• Important for equity issues

4
Ethnic gaps in attainment

• US - Substantial Black-White gap in the across
  all age groups (1SD), continues in contemporary
  data (NAEP, 2005) (e.g. Age 9 reading proficiency
  41 White vs. 13 Black)
• UK more ethnic variation and smaller White
  British-Black Caribbean gap, but still seen at
  all ages in contemporary data (DfES, 2006)

5
Previous research

• School quality frequently proposed explanation
  for ethnic gaps, particularly growth in gaps over
  time, but evidence is mixed and contradictory
  (Fryer Levitt, 2004 Phillips et al, 1998)
• Most of the above use econometric methods focused
  on fixed school effects, very little research
  directly modelling within-school gaps
• Two studies suggested differential effects re
  ethnicity (Nuttall et. al. 1989, Thomas et al
  1997) three studies do not (Brandesma Knuver,
  1989 Sammons et al, 1993 Strand, 1999), one is
  ambiguous (Kyriakides, 2004)
• need for recent large scale population study

6
Dataset

• England national population year 6 (age 11) in
  summer 2004 national tests (reading, writing,
  spelling, mathematics, mental mathematics and
  science)
• All state-maintained mainstream primary schools,
  including age 7 prior score pupil background
  534,724 pupils in 14,289 schools
• DV is average age 11 test marks (range 0-280)
  normal score transformed. White British-Black
  Caribbean gap is 0.3 SD.

7
Explanatory variables

• Age in months (normalised)
• Age 7 average test score (normalised)
• Sex (boy / girl)
• Ethnic group (White British against 12 other
  groups)
• Poverty (entitled to a FSM)
• Special Educational Needs (SAP or statemented)
• Mobility (pupil joined school during the key
  stage)
• Interactions (EthnicFSM, EthnicSex etc.)
• School level aggregates e.g., school mean age 7
  score FSM, SEN, mobile etc.

8
Fixed Effects

• Reduces Black Caribbean gap by up to half but
  significant gap remains. Black Caribbean pupils
  make less progress age 711 (i.e. gap widens even
  further)
• Significant interactions between
• Ethnic gender
• Ethnic FSM
• Ethnic FSM age 7 score.
• White British-Black Caribbean gap in progress
  age 7-11 is greatest for boys not on FSM (and
  most marked for the more able at age 7).
• Black African - more progress than White British
  across all gender FSM combinations.

9
Progress 7-11 by ethnic group, FSM gender
Effects are net of prior attainment at age 7,
age, FSM, gender, SEN, mobility, school FSM,
school mean age 7 score
10
School Effects

• Substantial school effect on pupil progress age
  7-11, 0.86 SD between schools at 5th and 95th
  percentile after control for intake
• Do Black Caribbean pupils attend lower quality
  schools?
• Identified all schools with 3 Caribbean pupils
  in the Y6 cohort
• 880 schools (6 of all primary schools)
• 43,376 pupils (72 of the Caribbean cohort)

11
High Black Caribbean vs. other schools
12
each ethnic group attending schools of
different quality
13
Directly modelling within-school gaps
How much variation between schools is there in
the size of the equity gaps (the within-school
gaps)? (n880 schools)
Note all variables allowed to vary
simultaneously at school level to allow for
possibility of variables being confounded.
Ethnicity reduced to three groups (White British,
Black Caribbean and Other)
14
Differential effects by prior attainment
While 187 schools (21) had intercepts sig. diff.
from zero, only five schools (0.6) had slopes
sig. diff. from sample mean.
15
Differential effectiveness for Black Caribbean
and White British pupils
R0.97
16
Gender and FSM

• For gender and FSM while the school level
  variation in the gaps was statistically
  significant the correlation between schools
  residuals for boys and girls was 0.98 and for
  FSM/No FSM pupils was 0.97.
• Schools that were the most effective for White
  British, for girls, and for pupils not on FSM
  were also the most effective for Black Caribbean
  pupils, for boys and for those on FSM.

17
Differential effects by FSM
R0.97
18
Covariance FSM gap vs Ethnic gap
R0.66
Those schools that narrow the Black Caribbean -
White British gap also narrow the FSM gap
19
Effectiveness against Black-White Gap
R-0.35
All pupils benefit from attending the more
effective schools, but White British do so to a
slightly greater degree
20
Conclusions school effects

• Black Caribbean pupils are concentrated in a
  small number of schools (72 in just 6 of all
  primaries) but no evidence from the
  cross-sectional analysis that these are low
  quality schools as measured by value-added.
• Could be argued that while overall school quality
  is not lower in the high Black Caribbean schools,
  some schools are particularly poor for their
  Black Caribbean pupils. However the within-school
  modelling shows that schools that do well for
  White British pupils also do well for their Black
  Caribbean pupils - key finding
• Growth in Black Caribbean gap age 7-11 is near
  universal, the White British - Black Caribbean
  gap does not vary significantly across schools,
  and no school appears to eliminate the gap.

21
Conclusions (Cont)

• Suggests more systemic factors are operating
• Ability grouping or tracking?
• Unequal distribution of novice teachers?
  (Clotfelter et al 2005)
• Widespread low expectations of Black pupils? (But
  Black African success?)
• Influences beyond school gates boys peer
  groups, street culture, acting white (Ogbu, 1986
  Sewell, 1997).
• Could be some or all of these (not mutually
  excusive).
• Further research needs to focus on within-school
  factors more than on between school differences.
• Equity-effectiveness trade off
• The most effective schools for White British
  pupils are also the most effective for Black
  Caribbean pupils BUT at the same time tend to
  increase the White British-Black Caribbean gap
  Raising all schools to the level of the best
  might actually increase the gap will require
  massive switch of resources to high minority or
  disadvantaged schools

22
End of presentation

• For a copy of the paper please e-mail me at
• steve.strand_at_warwick.ac.uk

23
Age 11 test score
24
Ethnic FSM age 7 interaction
25
Progress 7-11 by ethnic group, FSM gender
Model controls for prior attainment at age 7,
age, FSM, gender, SEN, mobility, school FSM,
school mean age 7 score
26
Direct modelling of size of within-school gaps
27
Conclusions pupil level

• Interaction effects are key White British and
  Black Caribbean pupils from disadvantaged
  families make equally poor progress, the big
  Black-White gap is for non-disadvantaged boys
• Black African pupils do not share this profile,
  they make better progress than White British in
  all gender FSM combinations
• Could be that SES controls (FSM, SEN, mobility
  and school FSM) fail to capture full extent of
  Black Caribbean social disadvantage, but other
  research suggests SES does not fully explain the
  gap (e.g., Philips et al, 1998, Strand 2008)