Spatial Simulation for Education Policy Analysis in Ireland

Report
Spatial Simulation for
Education Policy
Analysis in Ireland
An Initial Exploration
Gillian Golden
University College Dublin
[email protected]
Overview
 Individual level modelling for policy
analysis in the education sector –
proof of concept exercise
 Exploiting statistical value of
available administrative data and
“Joined up data” – NSB Position
Papers December 2011
 Spatial component - important for
planning and efficient resource
provision.
Microsimulation
 Representing a system in terms of it’s individual units.
 Often generated synthetically using fitting techniquesCensus small areas and PUMS
 Model effect of a policy change on individuals and
aggregate the results
 Can provide a more insightful picture of a complex
social system
Example – Integration in
Washington DC
Spatial Microsimulation
Statistical table
Ethnic Group
%
White alone
42.90%
Black or African American alone
50.10%
American Indian and Alaska Native
alone
0.60%
Asian alone
Hawaiian and Other Pacific Islander
alone
3.80%
Two or More Races
2.50%
0.20%
Irish Education System
Overall budget of €9 billion
annually.
Primary sector – approximately
3200 schools with 520,000 pupils
Traditional macro analysis – Value
for Money reviews, 2009 Special
Group on Public Service
Expenditure
Can spatial microsimulation add
value?
Data Sources
Irish Census of Population 2011
POWSCAR file
Department of Education and
Skills school XY coordinates
Other school level data
combined from databases
held in the Department
County Mayo chosen as test
geographic area
Methodology
POWSCAR fuzzy northing
and easting
Primary school XY data
Spatial Join Operation
Result - Individual level data
with contextual info on
pupil’s home and school
Data Cleaning Issues
 Spatial Join – primary schools located next to each
other.
 Geographical information not “fine grained” enough.
 Alternative method to assign pupils to schools –
optimisation “bin packing” algorithm
 School Census returns 2010-2011 used as “bin volume”
 Pupils assigned to schools according to school size.
 Primary and post-primary school co-located. Remove
records at random.
Data Cleaning- Bin Packing
Algorithm
Matched dataset – Irish Student
Simulation Model (ISSIM)
Rich Dataset
Many
Possibilities
Comparison with Department of
Education census records
Simulating school
Amalgamations
 Can examine hypothetical scenarios
 Example analysis – Close all schools
with less than 50 pupils and reassign
pupils to other schools
 Distance calculated based on point
distance between school and
randomly generated point in small
area of residence
School Amalgamations
 Variations in distance between home and school,
indicator of active school choice.
 Proximity table of schools
 Pupils reassigned to school nearest the one attended
before amalgamation
 “Before and After” analyses of the effects of the
amalgamations
School Locations
Financial Effects
 Smaller schools have a higher unit cost
 Notional projected future cost of a teacher - €55,000
per annum
 Capitation grants for additional school level staff,
school running costs etc
 Computation of cost before and after simulating the
amalgamation
Financial Effects
Social Effects
 Socio-Economic “Equality” in schools
DEIS Schools
 ISSIM useful for targeting resources aimed at
alleviating educational disadvantage
 DEIS programme designated schools
Community Effects
 Add value to qualitative analysis also
 Individual case studies possible
 Local “catchment area” of school
 Community effects of closure of small schools
 ISSIM can add information to contextual analysis – to
what extent does the school serve the local
community?
Evaluation
 Comprehensive dataset
 Cost-effective insights compared to surveys
 Possibilities to convert from microsimulation to agentbased model by including records with uncoded
place of school and assigning records to specific
households
 From static to dynamic – enrolment and cost
projections
Evaluation
Data protection - Dataset
currently warehoused in
Department of Education
Strict access controls
Published material based on
POWSCAR must be approved
by CSO.
Next steps
 Expand the model to cover all of Ireland
 Develop standard “data cleaning” methodologies –
cities may present additional validation issues.
 Examine some of the policy issues explored here in
more detail – initial focus on policies affecting primary
schools

similar documents