Can school competition improve standards? The case of faith schools in England

1
Can school competition improve standards? The
case of faith schools in England

• Rebecca Allen and Anna Vignoles
• Institute of Education, University of London
• Presentation to PLASC/NPD User Group
• 18th November 2008

2
Motivation for paper

• Faith schools, choice, competition improve
  standards or sorting?
• Endogenity problem
• Inconclusive UK competition literature
• Claims of the effectiveness of faith school
  competition in US and Canada

3
Outline of talk

1. Institutional background
2. Choice and competition between the faith and
   non-faith sectors
3. Estimation strategy and data
4. Results competition and pupil achievement
5. Faith schools and pupil sorting

4
Institutional background

• Growth in faith schools I churches were the main
  providers of 19th century education
• Growth in faith schools II 1902 Education Act
  set up the Dual System of state schooling, with
  mass building of schools by churches in 15 years
  following the Act to prevent state displacing
  church
• Growth in faith schools III expansion in RC
  schools in 1950s/60s, taking advantage of
  government loan scheme
• Today, religious schools in the state-maintained
  system educate about 15 per cent of secondary
  aged children (two-thirds are Roman Catholic,
  most of rest are CofE)
• Almost all are Voluntary-Aided (VA) rather than
  Voluntary-Controlled (VC)

5
Who can feasibly choose a faith school?

• Typical admissions policy at a faith school
  prioritises
• own denomination
• related denominations
• other religions
• non-religious families based on proximity
• Clear identification of Catholic, Anglican and
  non-religious families is not possible
• Religiosity requirements vary by school
• Religious self-identification of the family may
  not be strong
• The proportion of families who can feasibly
  choose a faith school is not directly related to
  the size of the underlying religious population
• Families can adjust church-going behaviour to
  satisfy requirements
• Demonstration of religious adherence may not be
  onerous or necessary
• Nature of competition related to size of
  religious population

6
Is active choice happening?

1. If 25 per cent of pupils in an area are at VA
   faith schools, this is associated with a 20
   percentage point increase in the proportion of
   pupils not at their nearest school
2. Transitions between primary and secondary school

7
School responses to active choice

• Schools are incentivised to respond to faith
  school competition for pupils by improving
  perceived school quality where
• Parents who are considering the religious sector
  have children who are seen as desirable to teach.
• Parents who are considering the religious sector
  would be responsive to a change in the schools
  perceived quality because they value academic
  results highly, relative to other characteristics
  such as religious ethos.
• Competing schools are closely matched in terms of
  pupil achievement in exams and therefore league
  table position.

8
Strategies to improve perceived quality

1. Effort focused on raising pupil achievement at
   GCSE
2. Effort focused on altering the social and ability
   characteristics of the pupil intake
   (cream-skimming)

9
Area-wide effect of religious schools on
achievement

• Education production function
• Effects of religious schools
• Direct effect of differences in religious vs
  non-religious school effectiveness
• Competition effects
• Changes in relative sizes of differentially
  effective schools
• Changes in peer composition that impact on
  achievement

10
Secular school quality
Proportion of Catholic school places in the area
Pupil achievement at GCSE
Historical Catholic population
Catholic families
11
HOXBY (1994) - EFFECT OF US (PRIVATE) CATHOLIC
SCHOOLS ON AREA-WIDE ACHIEVEMENT

• Finds 10 percentage point increase in Catholic
  school enrollment produces 0.9 additional years
  worth of educational achievement and 6 higher
  wages
• Supply of Catholic schools is instrumented using
  the current size of the Catholic population in
  the area
• Has pupil-level control variable of religion from
  National Longitudinal Survey of Youth
• Identifying assumption Catholic families who
  live in predominantly Catholic areas are no
  different from Catholic families who live in
  areas with few other Catholic families

12
Card et al. (2008) Effect of Canadian
state-funded Catholic schools on area achievement

• Find small positive effects from competition in
  the area of Ontario between grades 3 and 6
• Use school fixed effects specification of test
  score growth (with repeated cross-sectional
  pupil-level data for 5 cohorts)
• Identification strategy comparisons between
  areas with different fractions of Catholic
  families and different rates of growth of housing
  stock

13
Secular school quality
Proportion of Catholic school places in the area
Pupil achievement at GCSE
Historical Catholic population
Catholic families
14
Pupil-level achievement models

• Estimated using large ancient counties (39), so
  little sorting across areas based on unobserved
  characteristics
• Wide range of area-wide controls, including
  religious composition of area from Census and
  Church surveys

15
Instrumenting catholic school supply

• Taking specification 1, but treating RCsch as
  endogenous to modern-day demand for Catholic
  schooling
• First stage uses Catholic populations in ancient
  counties in 1931, which predicts RCsch (F-value
  20.81)
• Local Average Treatment Effect (LATE) identifies
  the effects of variation in supply of Catholic
  schools resulting from historical differences in
  the size of the Catholic population, holding
  constant modern-day Catholic church-going in the
  county.

16
Data

• National pupil database school leavers (age 16)
  in 2005, matched to KS4, KS3, KS2
• Outcome variables
• GCSE capped to best 8 subjects
• English, maths and science GCSE (best score in
  each subject)
• Pupil control variables
• NPD indicators (FSM, EAL, SEN etc)
• KS2 marks data separately for English, maths and
  science
• Deprivation indicators (IMD and 57 ACORN dummies)
• Ancient county control variables
• Pupil-level characteristics aggregated up to
  county
• English Church Census 2005
• Religious proportions from 2001 Census of
  Population

17
Results competition from faith schools
18
(No Transcript)
19
Results competition from catholic schools
20
(No Transcript)
21
Faith schools in the local competition space

• Each school has a unique competition space
  nearest 9 secondary schools by distance (single
  sex school adjustment)
• For each school, where does it sit in the local
  schooling hierarchy in terms of FSM and top
  ability intake composition?
• For each school, how stratified is the local
  schooling hierarchy and is this related to the
  number of faith schools in the competition space?

22
10 school competition space by FSM composition of
pupil intake
23
10 school competition space by top ability
composition of pupil intake
24
Relationship between number of faith schools in
local schooling hierarchy and intake
stratification
25
Association between faith schools in ancient
county and intake stratification
D (FSM) D (Top ability)
faith schools 0.25 0.26
Catholic schools 0.31 0.52
26
Association between faith schools in ancient
county and increase in dispersion of test scores
from KS2 to KS4
27
Conclusions

• No evidence that faith schools improve (or
  damage) area-wide academic achievement by
  encouraging competition for pupils
• Faith schools are associated with more stratified
  local schooling markets
• Evidence of cream-skimming or parental choice
  strategies?
• Stratification lowers incentives to compete based
  on effort
• Apparent effectiveness of faith schools in
  regressions likely due to within-area sorting
  based on unobservables