Diversity, choice and the quasi-market: an empirical
analysis of England’s secondary education policy, 19922005
Steve Bradley and Jim Taylor
Department of Economics
Lancaster University Management School
How has education policy changed?
What have been the consequences of the policy reforms?
How can the impact on outcomes be estimated?
Pre-1990
o Local Education Authorities (LEAs) determined the distribution
and use of school funding
o LEAs determined allocation of pupils (except for church schools
and grammar schools)
o LEAs appointed and employed teaching staff
o Limited role for head of school
o Limited role for parents and governors
Early 1990s: the creation of a quasi-market in secondary
education
o Motivation: general dissatisfaction with educational outcomes
o Aim: to improve educational outcomes
o Method: creation of quasi-market + targeting of ‘disadvantaged’ pupils
Current policy
Three main strands:
1. Establishment of a quasi-market: competition between schools
2. Specialist schools programme: diversity to improve pupil-school ‘match’
3. Urban education policy: Education Action Zones for ‘disadvantaged’
The quasi-market reforms: post-1990
Pre-conditions for quasi-markets
Policy reforms
Decentralised
decisionmaking
Local management of schools
+
Opting-out of government
control
+
Parents on governing body
+
Funding based on enrolments
+
Choice
Incentives
Information
+
+
+
Parental choice of school
+
Specialist schools
+
Attainment Tables + OFSTED
Voice
+
+
Purpose of the quasi-market
o Improve performance through greater competition for pupils
(diversity + choice + local management of schools)
o Increase transparency and accountability
o Improve efficiency through direct funding
- schools now responsible for 90% of recurrent expenditure
- more efficient allocation of resources - increase in total educational
product
o Induce private funding into state education
- private funders can contribute to creation of new schools
(academies) or take over ‘failing’ schools to raise performance
But will the quasi-market improve educational outcomes for
all pupils?
o Choice may lead to more sorting/segregation:
- ‘poorly educated’ parents less able to utilise information flows
- better-off parents move to live within a ‘good’ school’s
catchment area (allocation - lottery?)
- also better-off parents can afford travel costs leading to
cream-skimming by popular schools
o Why is sorting harmful?
- may lead to loss of peer effects for lower ability pupils; efficiency
losses if peer effects are non-linear
- long term - reinforces persistence of income disparities
Constraints on the quasi-market
o ‘Comprehensive’ schools cannot (ostensibly) choose pupils
o Entry and exit severely limited
o Excess demand for places in popular schools
o Accurate information needed for choice
(5-yearly inspection reports, annual assessment tables, open-days,
annual school reports). But information can be misleading (e.g. raw
scores and value added)
o Choice severely limited in many school districts
(non-metropolitan areas (20% of districts have 4 schools or less)
Diversity: the Specialist Schools Programme
Number of specialist and non-specialist secondary schools in England
3,500
3,000
2,500
Non-specialist schools
2,000
1,500
1,000
Specialist schools
500
0
1992
1994
1996
1998
2000
2002
2006: 80% of schools now specialist
2004
Specialisms
Year first
introduced
Total in
2006
Technology
1994
585
Languages
1995
221
Arts
1997
421
Sport
1997
350
Business
2002
229
Engineering
2002
57
Maths
2002
225
Science
2002
303
Humanities
2004
72
Music
2004
27
Total
-
2490
Urban Education Programme
extra funding for schools in disadvantaged urban areas
(28% of all schools) - 1999/05
(Education Action Zones)
o Support for gifted and talented pupils
- learning mentors for individual pupils
o Support for the ‘hard to teach’
- learning support units (to improve attendance)
o Provision of high-tech equipment in poorly equipped schools
Estimating the impact of the educational reforms
o Have
educational reforms been effective?
(e.g. exam results, truancy)
o Have the reforms had any distributional consequences?
o Which policies have been the most effective?
Exam results at age 16: % with 5+A*-C grades
60
55
50
45
40
35
30
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Truancy rate (%)
1.40
1.35
Days lost through unauthorised absence
1.30
1.25
1.20
1.15
1.10
1.05
1.00
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Proportion of pupils with ‘good’ exam results
(5 or more A*-C grades)
60.0
55.0
Specialist schools
50.0
45.0
Non-specialist schools
40.0
35.0
30.0
1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Gap widened from 7 (2001) to 14 (2005)
Metropolitan v non-metropolitan
schools
% 5 or more A*-C grades
60.0
55.0
50.0
%
Non-metropolitan
45.0
40.0
Metropolitan
35.0
30.0
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
Gap narrowed from 7 (2001) to 3 (2005)
2005
Truancy rate (%)
1.8
1.6
Metropolitan schools
1.4
1.2
Non-metropolitan schools
1.0
0.8
0.6
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Truancy rate = % of half days unauthorised absence
2002
2003
2004
2005
Estimating the effect of the policy reforms on educational
outcomes
Following Hanushek (1979, 1986), a school’s production function can be written as
follows:
Yst = f(PUPst, FAMst, SCH,t) + errorst
Y = outcome (e.g. exam results, attendance)
PUP = pupil characteristics (e.g. ability, gender, ethnicity)
FAM = family background variables (e.g. household income, parental education)
SCH = school inputs (e.g. school & teacher quality)
Extending this to include three separate measures of education policy:
Yst = f(PUPst, FAMst, SCHst, COMPst, SPECst, URBPROGst) + errorst
COMP
= competition from other schools in the same district
SPEC
= specialist schools policy
URBPROG = Education Action Zone policy (low income areas)
Endogeneity problems with the OLS production function
o Single equation production function likely to produce biased results:
- Error term includes unobservables (e.g. parental attitudes towards
education & innate ability of pupils)
- FAM and SCH are correlated (e.g. schools with a high proportion of rich
children find it easier to recruit ‘good’ teachers)
- SCH is endogeneous (e.g. schools with ‘good’ exam results find it easier
to recruit ‘good’ teachers)
o Hence:
- school quality variables (e.g. pup/teach): underestimated
- policy effects (SPEC and URBPROG): overestimated
An alternative approach: fixed effects model with panel
data
Endogeneity problems less severe - control for unobservables
Model to be estimated:
Yst = αs + λCOMPst + ηSPECst + δURBPROGst + Xstβ + Ttλ + εst
Y
COMP
SPEC
URBPROG
X
T
αs
= exam outcome
= exam outcome of other schools in district (lagged)
= a specialist school dummy (policy-off / policy-on)
= inner city schools policy
= time-varying controls (e.g. pup/teach, % poor)
= year dummies
= school fixed effects (time invariant)
- FE model estimates effect of policy variables on within-school
variation in Y over time
Fixed effects model: dependent variable = exam performance
Estimated coefficient
Competition
0.20***
Urban programme
1.8***
Specialist schools programme
0.9***
Pupil-teacher ratio
Part-time / full-time teachers
Number of pupils
Number of pupils squared
% eligible for free school meals
-0.001***
0.008
0.010***
0.000
-0.260***
Y94 (Some year dummies)
1.8
Y95
2.0
Y97
3.4
y00
5.5
y02
6.6
y04
8.3
y05
10.6
Constant
0.298
R-squared
0.41
Single-year OLS v fixed effects results
No. of schools in district
Competition
Urban
education
programme
Specialist
schools
programme
OLS model for 2005
0.13***
8.5***
6.5***
Fixed effects model for 1992-2005
0.20***
1.8***
0.9***
Controls = year dummies, pupil-teacher ratio, % pupils eligible for free school meals, etc.
Effect of including policy variables on time trend of exam performance
Explanatory variable
With policy
effects
Without policy
effects
Competition
0.20
-
Urban programme
1.8
-
Specialist schools programme
0.9
-
y94
1.8
2.1
y95
2.0
2.7
y96
3.3
4.0
y97
3.4
4.5
y98
4.1
5.6
y99
5.4
7.4
y00
5.5
8.4
y01
5.8
9.5
y02
6.6
11.1
y03
7.7
12.7
y04
8.3
13.9
y05
10.6
16.6
Note: Controls not shown
More detailed policy effects
Explanatory variables
Estimated coefficient
Competition
0.20***
Urban programme: phase 2000
2.3***
Urban programme: phase 2001
1.4***
Urban programme: phase 2002
1.1***
Art
1.1***
Business studies
2.5***
Engineering
-0.7
Languages
0.0
Maths
0.0
Science
0.7*
Sport
-0.2
Technology
1.6***
Humanities
-0.3
Music
0.8
Aggregate effect of education policies on exam results, 1992-2005
Main findings:
o 10pp improvement in competitor schools is associated with a 2pp
improvement for individual schools
– small (but significant) effect: overall effect around 3pp
o Specialist schools effect in arts, business studies, science and
technology: but only 1pp overall
o Urban programme raised exam score by 1.8pp
Total policy impact: 6pp of the 16pp improvement in exam results
(1993-2005) is ‘explained’ by the three policies.
What about the other 10pp? Grade inflation?
Distributional consequences of the quasi-market reforms
Have the reforms benefited some groups more than others?
Three tests:
1. Effect on different ability groups
2. Effect on different income groups
3. Effect on different ethnic groups
Do policy effects vary over the ability range?
Exam score quintile
Competition Urban
Specialist
education
schools
programme programme
Schools with lowest exam scores
0.23***
0.9***
1.9***
Second quintile
0.24***
1.7***
1.5***
Third quintile
0.26***
2.9***
1.2***
Fourth quintile
0.18***
2.3***
0.1
0.04*
2.4***
0.5*
Schools with highest exam scores
Answer:
• competition: effect is very small at top end of ability range
• urban programme: effect is weakest at bottom end of ability range
• specialist schools programme: effect is greatest at bottom end of ability range
Do policy effects vary over the family income range?
Free school meals quintile
Competition
Urban
education
programme
Specialist
schools
programme
-0.1
-1.1*
0.2
Second quintile
0.13***
0.9
0.8*
Third quintile
0.25***
1.5***
1.1***
Fourth quintile
0.24***
1.4***
1.2***
Highest % eligible for free meals (poor kids)
0.23***
1.4***
2.9***
Lowest % eligible for free meals (rich kids)
Answer:
Schools with highest poverty levels have benefited the most from education
policy
Do policy effects vary according to a school’s ethnicity?
Ethnicity
Competition Urban
Specialist
education
schools
programme programme
Under 10% ethnic minority
pupils
0.16***
0.7***
0.9***
10% to 50% ethnic minority
pupils
0.15***
1.7***
0.5
Over 50% ethnic minority
pupils
0.27***
2.8***
2.4***
Answer:
Biggest policy effects for schools with high % of ethnic minority pupils
Distributional consequences of the specialist schools
programme: by specialism
% eligible for free school meals
(average 1992-2005)
Lowest
quintile
Middle
quintiles
Arts
0.1
1.3**
2.3***
Business studies
1.1
2.3**
6.0***
-2.7**
1.5
-3.9*
Languages
-0.5
0.0
5.6***
Mathematics
-0.6
0.9*
2.2
Science
0.1
1.6***
2.7***
Sport
0.0
-0.1
-0.2
1.1***
1.5***
4.3***
Controls included?
Yes
Yes
Yes
R-squared (within)
0.42
0.39
0.50
n
8091
24143
8017
Engineering
Technology
Highest
quintile
Metropolitan v non-metropolitan schools
Why might the policy effect differ between metropolitan and
non-metropolitan schools?
(i) Parental choice is greater in metropolitan areas
(ii) Greater competition for pupils in metropolitan areas
(iii) Extra resources for deprived urban areas since 1999
- Education Action Zones (virtually all schools
in metropolitan areas + some other deprived areas)
Impact of competition, urban programme and specialist
schools programme: metropolitan v non-metropolitan
Competition Urban
education
programme
Specialist schools
programme
Non-metropolitan
areas
0.11***
0.9**
0.5***
Metropolitan areas
0.39***
1.1***
1.7***
o Much
stronger policy effects in metropolitan areas
Impact of policy on truancy rate: metropolitan v non-metropolitan
Competition Urban
education
programme
Specialist schools
programme
Non-metropolitan areas
-0.42**
-0.13***
-0.05**
Metropolitan areas
-3.35***
-0.22***
-0.10**
Policy effects much stronger in metropolitan areas
Some conclusions
1. Effect of increased competition
- Only around 3pp of the increase of 20pp can be attributed to
the increased competition for pupils
- But impact bigger in metropolitan schools
2. Specialist schools programme
- accounted for only an extra 1pp in exam results
- but variation between specialisms (up to 3pp in business studies/
enterprise)
3. Inner cities programme has accounted for an extra 2pp in GCSE results
4. Hence only one-third of the total improvement is accounted for by
the three major policy initiatives
5. Estimated impact of policy has had important distributional benefits
(biggest effects for low ability and low income groups)
Descargar

Slide 1