The document describes an agent-based model of school choice and admissions policies. The model simulates families ranking and applying to schools based on attributes like location and school performance. The model reproduces empirical patterns at both the system and individual agent levels. Various policy scenarios are explored, like closing the lowest performing school or improving school value-added. The results provide insights into what factors drive observed patterns and outcomes at both levels.
2. School Choice and Admissions
Distance-based admissions policies
hierarchies of school popularity
lead to the reproduction of social inequality
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
8. Parental Aspiration
http://landscapemodelling.net
“I don’t want my kids to
work in the supermarket.
I want my sons to
become a doctor, or a
pilot or something like
that.” [p.102]
“I just want them to be
happy and enjoy their life
and obviously, yes, do
well at school but so long
as they’re happy” [p.104]
IGU 2013 | Leeds, UK | 01/06/13
9. Parental Aspiration
“I think [education is] reasonably important. I
wouldn’t put it up there as really top ranked just
because, you know, I think there’s more
important things in a child’s life. As long as your
child is well adjusted and doesn’t suffer in
school think, you know, the school is
appropriate” [p.105]
“Oh, the single most important thing that a
parent can give to their children in life is a first-
class education” [p.98]
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
10. Parental Aspiration
“My gift to my children…you could say
inheritance, what I pass on to them, is that
they’ve all achieved a university education. If
they’ve done that then I’ll feel I’ve done my bit
as a parent” [p.101]
“I’d like [my daughter] to go to university but if
she didn’t it wouldn’t be the end of the world.”
[p.104]
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
11. Model Structure
http://landscapemodelling.net
Family Attributes
Aspiration
Child Age
Location
Attainment
Strategy
Rankings
Catchment (schools)
Satisfactory (schools)
Avoided (schools)
School Attributes
Value-added
Location
GCSE score
Available Places
List of applicants
List of allocated pupils
(i.e., families)
IGU 2013 | Leeds, UK | 01/06/13
12. Model Procedures
Families rank schools for moving & applying
Families can move once prior to application
Move anywhere or only into areas with lower or
equal aspiration (location constraints)
Schools allocate places to applicants based
on distance alone
School value-added influences pupil
attainment
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
13. School Ranking Strategies
Criteria
Catchment Satisfactory Avoided
Ranking
Catchment
Satisfactory
Avoided
Other Schools
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
14. School Ranking Strategies
Criteria
Catchment Satisfactory Avoided
[not empty] [empty] [not empty]
Ranking
Catchment
Satisfactory
Avoided
Other Schools
http://landscapemodelling.net
1st by GCSE descending
N/A
Remove from ranking
2nd by distance ascending
IGU 2013 | Leeds, UK | 01/06/13
15. Model Analysis
Reproduction of system-level patterns
Identify necessary structures and rules
Identify agent-level patterns
Which agents move?
Which agents fail to secure preferred school?
Explore implications of policy scenarios
Closing poor schools
Improving school value-added
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
16. System-Level Analysis
What conditions needed to reproduce
empirical system-level patterns?
Variation in school value-added?
Constraints on family movement?
Examine combinations of rules
No value-added (nVA), no location constraints (nLC)
Value-added (VA), no location constraints (nLC)
No value-added (nVA), location constraints (LC)
Value-added (VA), movement constraints (LC)
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
20. Parent-Level Analysis
What are consequences of system-level
patterns at agent-level?
ABM without agent-level analysis under-utilises
the approach
Identification of agent-level patterns to
search for empirically
Not predicting from past to future
Predicting from ‘known to unknown’
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
23. Scenario Analysis
Explore implications of policy scenarios
Examine consequent patterns at system-level and
agent-level
Closing poorest school
Re-allocate pupils to ‘best’ school
Re-allocate pupils to all schools
Improving all schools’ value-added
Implications for parent allocation strategies and
success
http://landscapemodelling.netIGU 2013 | Leeds, UK | 01/06/13
24. A:Pratio
GCSE score20 20 2080 80 80
0
10
Scenario: remove ‘poorest’ school
http://landscapemodelling.net
Poorest to Best Poorest to AllNo Removal
IGU 2013 | Leeds, UK | 01/06/13
27. Future directions
Simulating reproduction of social inequality?
Ethnicity?
Religion?
Class?
Wealth?
Definition of a ‘good’ school?
GCSE score used here – but what other factors?
‘Gaming the system’?
And other options (e.g., leaving state schooling)
http://landscapemodelling.net
Aspiration
IGU 2013 | Leeds, UK | 01/06/13
28. Summary
ABM of school allocation policy
Results support arguments about drivers of
empirical (system-level) patterns
Agent-level patterns identified to explore
empirically
http://landscapemodelling.net
james.millington@kcl.ac.uk
@jamesmillington
IGU 2013 | Leeds, UK | 01/06/13