Assessing the Impact of School Resources on Pupil Attainment

1
Assessing the Impact of School Resources on Pupil
Attainment

• Fiona Steele
• University of Bristol
• Anna Vignoles and Andrew Jenkins
• Institute of Education, University of London

2
Research questions

• Do pupils in better funded schools have better
  outcomes?
• Do pupils in schools with lower pupil teacher
  ratios have better outcomes?
• And, more ambitiously, can we establish a causal
  relationship?

3
Methodological Issues

• Need to account for hierarchical structure
• 3 levels pupils (1), schools (2), LEAs (3)
• Note that attainment is a pupil-level outcome,
  while resource variables are measured at the
  school level
• Funds are not allocated to schools at random
• Makes it difficult to establish whether any
  relationship between attainment and resources is
  causal

4
Data

• Analyse large cohort of pupils who completed KS2
  in 2000 (age 11) and KS3 in 2003 (age 14)
• PLASC data on 464,783 pupils and 3,011 secondary
  schools in England
• Dependent variables KS3 maths, science, English
• Resource information Expenditure per pupil and
  staffing ratios
• Includes range of control variables
• Good school variables
• Crude indicators of family background (e.g. fsm
  indicator)
• Linked to 2001 Census data

5
Example of Ignoring Hierarchical
StructureEffects of Expenditure per Pupil on
KS3 Maths
Effects adjusted for a large number of control
variables.
6
Establishing Causality Resource Allocation in
the UK

• Funds are not allocated to schools at random
• Funding of schools is largely administered by
  LEAs
• Central government gives funding to LEAs for
  schools and other services
• More disadvantaged LEAs get more funding
• LEAs have some autonomy in education spending

7
Selection Bias

• Schools with higher proportions of less able
  children, or children from disadvantaged
  backgrounds, are allocated higher levels of
  funding
• These children will tend to do worse in school
  anyway, so attainment in better-resourced schools
  will be low
• If we do not fully account for compensatory
  funding, we will underestimate a true positive
  effect of school resources on attainment

8
Selection Bias
9
Selection Effects at the School and LEA Level

• LEA level due to the way central government
  allocates resources to LEAs
• Fewer resources are given to LEAs with high mean
  attainment
• School level due to the way resources are
  allocated among schools within an LEA
• LEA gives fewer resources to schools with high
  mean attainment

10
Other Examples of Selection Bias

• Effect of premarital cohabitation on risk of
  divorce
• Effect of a training scheme on unemployment rate
  when participation in scheme is voluntary
• Effect of a health intervention on child
  mortality rate when intervention is targeted
  towards areas of greatest need

11
Our Methodological Approach

• Model attainment jointly with resource allocation
• Multilevel simultaneous equations model
• Allows for possibility that there may be
  unobserved variables at school or LEA level that
  affect both attainment and resource allocation
• Multilevel model allows for selection bias
  operating at the LEA or school level (or both)

12
Effects of Expenditure per Pupil on Attainment
13
Effects of Pupil-Teacher Ratio on Attainment
14
Effect Sizes

• For additional 1000 per pupil expenditure,
  expect increase in attainment of 0.07 of a level
  in maths and 0.2 of a level in science
• For reduction in the pupil-teacher ratio of one,
  expect increases of 0.12-0.13 of a level in maths
  and science
• (Expect improvements of 0.5 of a level per year)

15
Conclusions

• Positive effects of additional resources on maths
  and science
• Nonsignificant or negative effects on English
• Need to allow for non-random nature of resource
  allocation strong evidence of selection at
  school and LEA levels
• Need to allow for variation in resource (and
  selection) effects across subjects