Research Briefing

How to Pull the Right Lever

School Attainment, Open Data Analytics and Policy — Lessons from Brighton and Hove

Adam Dennett · UCL Centre for Advanced Spatial Analysis, 2025
Read "How to Pull the Right Lever" report online Download PDF Full slides Short slides Launch Policy Simulator

Overview

School attainment matters and Policy Makers understandably want to improve it for all children to give them the best change of future success. But it is all to easy to pull the wrong policy lever when information and intelligence is incomplete. Recent experiences in Brighton and Hove have highlighted the gap between open data and actionable intelligence for Policy Makers, Schools and anyone wanting to actively contribute to statutory public consultations. Intelligence to inform local debates could be available from within freely available online data sources, but is largely hidden within mountains of data and a vast specialist literature.

Education professionals and researchers know what factors influence attainment and they know which are most important. But every school and every Local Education Authority is different and there is a gulf between general understanding and specific local intelligence that can target the most effective policy treatment for a particular location.

The How to Pull the Right Lever report desribes what can happen when good policy intentions can be derailed by an incomplete understanding, but shows how better understanding and better policy is within the grasp of any local education authority through the open data resources published by the Department for Education.

The Policy Simulator within this site is a first attempt bridge the open data -> actionable intelligence gap for local education policy in England. It has been designed to curate, process, visualise and analyse the extensive open data sources publised by the Department for Education in an accessible way. It allows users to explore what policy changes might have be biggest impact on school-level GCSE attainment for any state school in England, taking account of their own unique circumstances.

The simulator is underpinned by a robust statistical analysis of four years of DfE data detailed in the main report and supplementary materials found at the bottom of this page. The tool allows anyone to select a particular local authority or school and understand what the main attainment drivers for that place and thus what the most effective improvement policies might be.

Brighton and Hove is used as an illustrative case study example, the the tool and analysis are relevant for any Local Education Authority in England.

Key figures

13,419 school-year observations across four years
80% of variation in Attainment 8 explained by the model
#1 Overall absence is the strongest predictor of attainment
152 local authorities included with random effects

If policy makers fail to take account of all relevant factors influencing attainment and particular local contexts and variations, they risk pulling the wrong lever. We hope the tool and analysis in these pages will help local authorities and schools make better decisions to improve educational outcomes for all children.

Standardised model coefficients showing absence as the strongest predictor
Standardised coefficients from the multilevel model. Overall absence rate has the largest effect on Attainment 8 scores for all pupil groups, exceeding the impact of % FSM disadvantage.
Supplementary Materials

Analysis reports & data

Detailed technical reports underpinning the main paper.

Data Overview
Sources, panel construction, descriptive statistics, missing data patterns, imputation strategy, and variable definitions.
DataPanelDescriptive
Model Results
Multilevel models, diagnostics, observed vs predicted plots, residual geography, and the disadvantage gap.
ModelsDiagnosticsResults
Model Experiments
Human-readable lmer implementations, coefficient commentary, and variable importance analysis.
lmerCoefficients
Brighton & Hove Case Study
Interactive maps, non-linear effect interpretation, school-level residuals, and policy recommendations.
Case StudyMapsPolicy
Financial Data Overview
School financial returns (CFR & AAR): coverage, spending composition, trends, and relationships with attainment.
FinanceCFRAAR
Financial Model Experiments
Systematic test of whether school financial variables improve the baseline attainment models.
FinanceModels
LA Typology Report
Local authority clustering analysis: methodology, cluster profiles, pen portraits, and typology maps.
ClusteringTypologyLA
Journal Paper (PDF)
Condensed Journal Pre-Print — Chicago Author-Date citations.
JournalJEPPDF

CRediT authorship contribution statement

Adam Dennett: Conceptualisation, Methodology, Software, Formal analysis, Data curation, Writing – original draft, Writing – review & editing, Visualisation, Project administration.

Declaration of generative AI use

The authors used Anthropic's Claude Code (Claude Opus 4.6) to assist with the development of R data-processing pipelines, the implementation of multilevel model fitting and visualisation code, the formatting of tables and bibliography entries, and the rendering and deployment of Quarto outputs. Any analytical decisions and interpretations during the iterative analytical process were reviewed, verified and approved by the authors, who take full responsibility for the work.