The educational gender gap, catch up
and labour market performance
Martyn Andrews (University of Manchester),
Steve Bradley, Dave Stott & Jim Taylor
(Lancaster University)
The educational gender gap

Issues




Performance of girls is superior to boys and getting
wider
Concern about low achieving boys
Girls do better in ‘language’ based subjects, boys do
better in Maths & Science
Even if girls outperform boys, does it matter if they are
discriminated against in the labour market?
The educational gender gap


Objectives
Use biannual YCS (1985-2001) & NPD (2002-03)

1. Define & measure the gender gap and document how it
changes through time

2. Explain how the gap changes when we control for


Observable effects – individual, family, school, neighbourhood
Unobservable effects


School-level (e.g. discipline, tiering, streaming
Individual-level (e.g. attitudes, motivation)

3. Repeat 1 & 2 for subject groups

4. Measure & explain how the gap changes during the educational
process

Age 11-16 (at KS2, KS3, KS4)
Previous research

Educational

Descriptive studies

School effectiveness

Qualitative / case studies






e.g. Gorard et al (1999)
e.g. Wong et al (2002)
e.g. OFSTED (2003)
Organisation, teaching & learning, curriculum & assessment
School organisation
Culture of laddishness
Idiosyncratic school effects
Home background
Economics
e.g. Dolton et al (1999), Burgess et al (2004)
Data & methodology

Estimate education production functions

Outcome = function of:







Are there correlations between girl and (observable & unobservable)
effects?






Girl (gap)
Individual characteristics
School characteristics
Neighbourhood characteristics
Unobserved individual-level effects
Unobserved school-level effects
Zero – gap is the published figure
Girl & personal (zero?)
Girl & school (sorting?)
Girl & unobserved individual effects (motivation)
Girl & unobserved school effects (sorting?)
Unobserved individual-level & unobserved school-level effects
Data

Pooled cross-section (YCS) data (1985-2001)



Observed variables





YCS2-3 – GCE/CSE
YCS4+ -- GCSE
Individual – gender, ethnicity, age
Family – parental occupation, single parent, housing
tenure
School – Pupil-teacher ratio, pupil composition, size,
competition
Neighbourhood – unemployment rate, occupational mix
YCS6-11 observe the same school up to 6 times –
school level unobservables
Data

NPD 2002 & 2003



Observe KS2, KS3 & GCSE results
Population
Advantages:
Control for (estimate?) unobserved individual effects
 41,000 pupils move schools
 Identify individual & school level unobservables


But … few individual-level covariates
Outcomes – measures of educational
performance

Pass/fail for each subject (grade C +)

Number A*-C GCSEs – all subjects

5 + A*-C GCSEs – headline figure

Points score – distribution (A*=7, etc.)
Absolute versus relative gaps

Debate




Educationalists label the
absolute gap as the
‘politicians error’
Absolute gap increases
as relative gap falls
Absolute gap is correct
Note the increase in
the gap from the
introduction of GCSE
Econometric findings - observables
What explains the gender gap (differential)?

Selective schools have a very large effect on attainment

Single sex schools have a large, but smaller, effect

Neither of these effects contribute much to the gender gap


Other observable differences between girls and boys (e.g. family
background, poverty) do not explain the gap
Are the findings genuine? Biased sample for YCS but we observe
similar effects for NPD (population)
The story so far


Observable differences between girls & boys do not explain
the gap
Girls must therefore behave differently prior to GCSEs

1. Choice of secondary school

2. Subject level gaps at GCSE

3. Differences in exam performance between KS2 & KS4
1. Choice of school

Control for school-level unobservables


Controlling for school level unobservables is important



level not trend
Discipline, tiering, streaming
Between 1991-2001 the gender gap is halved
-

YCS6-11 & NPD1-2 (panels)
E.g. YCS10 = 0.04 versus 0.10
Implication: Has the quasi-market (ERA, 1988) meant that girls are
marginally more attractive to better schools?

Un-testable because of lack of linked school data prior to 1991
2. Subject level gaps at GCSE
2. Subject level gaps at GCSE





Data shows that girls outperform boys in languages, English &
vocational subjects
‘One-off’ GCE-GCSE effect disadvantaging boys – languages, science,
maths
Since 1988 the gap has increased at the same rate – girls catch-up in
maths & science
Controlling for observable & unobservable differences lowers the gap
by one-tenth of a GCSE grade
Girls ahead in English, languages & vocational, level in humanities &
behind in Maths and Science
3. Differences in exam performance between
KS2 & KS4


Maths, English, Science at KS2, KS3 & KS4 (population)
See Table on KS2-4
Gaps at GCSE: English (0.63), Maths (0.03) and Science
(0.06)
 At KS2: Girls better in English (0.23), behind in Maths
(-0.07) & Science (-0.04)
 Girls improve between KS3 & KS4 in all subjects, but only in
English between KS2 & KS3

Differences in exam performance


Controlling for school & pupil-level unobservables
1. Correlation between ‘Girl’ & individual-level = 0!




But, disaggregating we find that girls are unobservably better in English
and worse in Maths & Science
Note that KS2 & KS3 do not test other ‘girl-good’ subjects – see YCS results
2. The correlation between unobserved-school level & unobserved
individual-level effects is greater than zero
 Unobservably good pupils go to unobservable good schools (i.e.
middle class parents, catchment areas)
3. The correlation between ‘Girl’ & unobserved school-level effects is
greater than zero (see YCS results)
 Girls go to unobservably better schools
 Girls are observably better at KS2 – schools therefore select them
Conclusions & implications for policy

1. Gender gap emerges once the GCSE system is introduced


2. Girls are better than boys






Learning & assessment methods favour girls
A) English
B) Selected into unobservably better schools
3. No effect of single sex schooling
4. Selective schools & poverty have a small effect on the gap
5. Gap is greatest in English & languages and has closed in
Maths & Science
6. Unobserved differences between schools (e.g. discipline,
tiering, streaming) are important – YCS only
Speculation


A) Introduction of GCSE system created the gap
B) Quasi-market exacerbated the gap



Cumulative & self-perpetuating


changed incentives facing schools
select the best – girls
Girls go to good schools
But the gap stabilises


Shocks A & B eventually burn out (equilibrium)
The introduction of KS2 helps boys (fewer ‘girl-good’
tests), which means they also sort into ‘good’ schools
Descargar

The educational gender gap, catch up and labour market