schools' ses and university academic performance
Post on 14-Jul-2015
40 Views
Preview:
TRANSCRIPT
The University of Western Australia
Acknowledgements
National Centre for Student Equity in Higher Education
for funding support and data provision
Useful comments on the paper from John Phillimore and
participants of the Honouring Paul Miller event,
November 2014
This paper is dedicated to the memory of Paul Miller,
who conceived the original research question but who
passed on before the project commenced
All mistakes remains those of the authors.
2
The University of Western Australia
Motivation for the research
Higher education reform
Bradley review 2008
Target of 40% of Australians aged 25-34 with degree by
2025
Equity target of 20% of higher education enrolments
from low SES backgrounds
Move to demand driven system in 2012
17.4% low SES enrolments in 2013 (Department of
Education 2014)
3
The University of Western Australia
Data
De-identified student record data from an anonymous university
• Commencing undergraduate degree in 2011 to 2013
• Admitted on basis of completing Year 12
• Information on school where student graduate from
• 8,417 observations
Linked to the ABS Socio-economic Index for Areas
Linked to data from MySchool (ACARA)
• Contains data on schools’ characteristics
• 183 schools in the sample
4
The University of Western Australia 5
Student records
• Age
• Gender
• English-speaking
background
• Residential postcode
– Index of Economic
Resources
– Index of Education and
Occupation
• Primary field of study
• ATAR
• WAM in first year
MySchool
• Sector (Catholic,
Independent, Government)
• Rural/urban
• Co-ed, boys or girls school
• Funding per student (all
sources)
• Teacher/student ratio
• Non-teaching staff/student
ratio
• Index of Community
Socioeconomic Advantage
(ICSEA)
The University of Western Australia
ICSEA – A measure of Schools’ SES
Measure of students’ socio-educational similarity
Student level measures
• Parental education
• Parental occupation
• Geographical remoteness
School level measures
• Indigenous student enrolment
• NESB student enrolment
• Aggregated socio-educational measures
National mean of 1,000
• Advantaged if above 1,000, disadvantaged if below 1,000
6
The University of Western Australia
Selected descriptive statistics
Mean ATAR = 82.3, mean WAM = 63.7 (8,417 obs)
7
Variable Govt Cath Indp
ATAR 81.7 82.6 82.7
Weighted Average Mark 64.3 63.1 63.3
ICSEA (school SES) 1,041 1,065 1,117
Income per student 14,602.8 14,880.0 18,360.3
Teacher/student ratio 0.076 0.075 0.084
Non-teaching/student ratio 0.026 0.033 0.044
No. of schools 94 34 55
No. of students 3,478 2,580 2,359
The University of Western Australia
Methodology
Education production function
𝐴𝑃𝑖 = 𝑓 𝐵𝐶𝑖 , 𝑆𝑖 , 𝑃𝐴𝐴𝑖 , i = 1,…,n (1)
Where AP = academic performance
BC = background characteristics
S = school characteristics
PAA = prior academic achievement
8
The University of Western Australia
Methodology – multilevel models
Hierarchical structure – students clustered within schools
Random intercept
𝐴𝑃𝑖 = 𝛼0𝑗 + 𝛼1𝐵𝐶𝑖 + 𝛼2𝑃𝐴𝐴 + 𝜀𝑖 (2)
𝑖 = 1, … 𝑛.
j = 1,…,k.
Random coefficients
𝐴𝑃𝑖 = 𝛼0 + 𝛼1𝑗𝐵𝐶𝑖 + 𝛼2𝑗𝑃𝐴𝐴 + 𝜀𝑖 (3)
𝛼1𝑗 = 𝑓 𝑆𝑖
𝛼2𝑗 = 𝑓 𝑆𝑖
𝑖 = 1, … 𝑛.
j = 1,…,k.
9
The University of Western Australia
Standardisation of variables
For ease of interpretation, continuous variables of interest were
standardised
Standardisation has been done using population or grand means
• Comparison of between school effects
Standardisation (for ATAR and student SES) has also been done
using within school means in two models estimated (presented last)
• Comparison of within school effects
10
The University of Western Australia
Is Schools’ SES associated with
WAM?
Is the impact of Schools’ SES on
WAM associated/affected by other
variables?
11
Results
The University of Western Australia 12
Random intercept model results
Variable Model 1 Model 2
Age (at commencement) 0.408*** 0.392***
(0.081) (0.087)
Female 5.206*** 4.821***
(0.323) (0.326)
Foreign born 0.193 0.348
(0.417) (0.361)
NESB -0.323 -0.536
(0.578) (0.582)
IER+ 0.401** 0.414***
(0.174) (0.153)
IEO+ -0.140 -0.120
(0.205) (0.200)
ICSEA+ -0.637*** -0.729***
(0.238) (0.236)
FoS Not included Included
Prob > χ2 0.000 0.000
The University of Western Australia 13
Random intercept model resultsVariable Model 3 Model 4
Independent school 0.679 0.909
(0.634) (0.637)
Catholic school 0.098 -0.084
(0.568) (0.602)
Rural school 0.478 0.796
(0.596) (0.609)
Boy’s school -2.824*** -2.127**
(0.940) (1.064)
Girl’s school -1.607*** -1.106*
(0.555) (0.668)
School income per student+ -1.267**
(0.560)
Teaching staff per student+ 0.694*
(32.473)
Non-teaching staff per student+ -0.095
(24.945)
ICSEA+ -0.611** -0.426
(0.308) (0.310)
Demographics Included Included
FoS Included Included
The University of Western Australia
How does prior academic achievement
impact on university academic
performance?
How does prior academic achievement
impact on the relationship between
schools’ SES and university performance?
Do certain schools provide better
platforms for university study?
14
Results
The University of Western Australia 15
Random intercept model results
Variable Model 7 Variable Model 7
IER+ 0.202 Girl’s school -1.823***
(0.148) (0.703)
IEO+ -0.075 School income
per student+
-1.166**
(0.168) (0.491)
Independent
school
0.850 Teaching staff
per student+
16.649
(0.606) (32.030)
Catholic school -0.703 Non-teaching
staff per
student+
38.681
(0.532) (26.093)
Rural school 0.624 ATAR+ 5.944***
(0.599) (0.247)
Boy’s school -2.048** ICSEA+ -1.506***
(0.800) (0.277)
Demographics Included FoS Included
The University of Western Australia
Are there differences in the way within-
school variation in student characteristics
impact on the determinants of university
performance, particularly the role of
ATAR?
The following models standardise IER,
IEO and ATAR using means within
schools.
16
Results
The University of Western Australia 17
Random intercept model results
^standardised using means within schools
Variables Model 8 Variables Model 8
IER^ -0.042 Girl’s school -1.078*
(0.136) (0.641)
IEO^ 0.135 School income per student+ -1.148**
(0.128) (0.561)
Independent school 0.804 Teaching staff per student+ 40.673
(0.653) (33.140)
Catholic school -0.212 Non-teaching staff per
student+
2.617
(0.613) (21.834)
Rural school 0.987 ATAR^ 5.870***
(0.609) (0.171)
Boy’s school -2.598** ICSEA+ -0.370
(1.117) (0.313)
Demographics Included FoS Included
The University of Western Australia
Are there differences in the way schools translate prior academic ability into university performance?
Are there differences in the way schools with varying SES prepare their students for university?
Use of random coefficient model
Slope of ATAR and ICSEA allowed to vary
18
Results
The University of Western Australia 19
^standardised using means within schools
Random coefficient model results
Variables Model 9 Variables Model 9
IER^ -0.041 Girl’s school -1.381**
(0.135) (0.661)
IEO^ 0.141 School income per
student+
-1.166**
(0.129) (0.525)
Independent school 0.791 Teaching staff per
student+
44.400
(0.643) (32.059)
Catholic school -0.035 Non-teaching staff per
student+
5.591
(0.596) (21.808)
Rural school 0.675 ATAR^ 5.693***
(0.609) (0.176)
Boy’s school -2.635** ICSEA+ -0.386
(1.090) (0.309)
Demographics Included FoS Included
The University of Western Australia
Limitations
• Sample bias – students who have successfully gained
entry to university, despite SES background and/or
ATAR
• While the data covers 183 schools, only performance
at one university is examined
Key findings
• Students from lower SES schools perform marginally
better than peers from higher SES schools
• Individual SES background has no impact on
university performance
• School resourcing characteristics does not impact on
university performance 20
Concluding remarks
The University of Western Australia
Implications
• Admission regimes at university could take into
account relatively good performance of students
from low SES schools and advantage them in
gaining entry
• Resource allocation – is it a useful policy tool for
improving academic performance? Findings of the
present study suggest not – consistent with other
studies (Marks 2010)
• Suggestions that resource quality rather than
quantity matters
21
Concluding remarks
top related