1A_3_A geodemographic classification of london primary schoolsPresentation Transcript
A geodemographic classification of London primary schools Anne Gibbs, John Stillwell & Linda See 14th April 2010
Structure of the presentation Aim of the research Research questions Methodology The Classification What it reveals: the performance of different types of schools Further research
Why Classify? There is a complex set of relationships between schools, neighbourhoods and performance. Pupil populations are particularly diverse in London. Classifying schools by their pupil populations: creates some order out of the chaos; enhances understanding by highlighting similarities and differences between schools; has potential to be used as a benchmarking tool by policy makers and managers; provides a framework for further research from both geographical and educational perspectives.
The Database Spring Census 2007 All pupils in maintained primary schools in London Reception to Year 6 Derive new variables e.g. % of mobile pupils in Year 6 English & Maths Levels
matching to Lower Layer Super Output Areas (LSOAs)
The Database Key Stage 2 Results Edubase IDACI / LSOA Area Classification
Ethnicity of London Primary Pupils Base population = total number of pupils in the school in the statutory years of Reception to Year 6
Socio-economic variables 1 % eligible for Free School Meals (FSM) % with English as an Additional Language (EAL) Base population = total number of pupils in the school in the statutory years of Reception to Year 6 (except for mobility rate, where base is Year 6 only)
% FSM Pupils (quartiles)
% EAL Pupils (quartiles)
Socio-economic variables 2 % eligible for Free School Meals (FSM) % with English as an Additional Language (EAL) % with Special Educational Needs (SEN) % of 2006/7 Year 6 pupils who entered their school after the beginning of Year 5 (MOBILE) Base population = total number of pupils in the school in the statutory years of Reception to Year 6 (except for mobility rate, where base is Year 6 only)
Correlations between variables All correlations are significant at the 1% level of confidence
Deriving the Schools Classification k-means algorithm in SPSS; distribution of many variables skewed, so the data was range-standardised; ran clustering routine for n=4 to n=22 clusters giving 19 alternative classifications; alternatives assessed for homogeneity and evenness of cluster size, using standardised data; short listing of classifications with 7,10,14 & 16 clusters.
Final Selection More detailed assessment of ‘short listed’ solutions using 3 measures of cluster validity: Selection of 14-cluster solution
A 5 9 2 B 13 C 14 1 8 3 10 6 11 12 7 D 4 The Schools Classification: Visualisation White British Mixed ethnicity Non-white / EAL Well-off / stable Mobile Needy Source: After Harris et al. (2005) Figure 6.3, p.170.
Cluster profiles: Group D
Cluster profiles: Group A
Cluster profiles: Group B
Cluster profiles: Group C
Super Groups by LEA
Contextual Performance Ranking of Clusters(indexed to global average = 100)
Possible Future Research Classification as a framework for more detailed analysis of relationship between schools and their neighbourhoods. Classification as a framework for more work on mobility and its impact on educational attainment. Potential for use as a benchmarking tool or as a sampling frame for qualitative research. Feasibility of an online system. Updating the Classification.