2016 has seen some big changes in our school roll projections service with the development of a new projection model which has been written in the statistical programming language R, allowing us to harness R’s power to make use of much bigger datasets and a more complex model.
Every year the GLA intelligence unit produces projections for the number of state school places that will be needed over the next 15 years in the London Boroughs. The Boroughs are responsible for ensuring that every one of their resident children is given a suitable school place – a huge challenge in many parts of the Capital, given London’s rapid growth, and the limited space and resources available. School roll projections are an integral part of the process of planning future provision.
With its dense population and good transport links, London has an unusually high rate of cross-borough school journeys and many children do not attend their nearest school. The image below taken from the Schools Atlas shows the example of London Nautical School (red pin) which draws pupils from more than four different boroughs (shown shaded in orange).
The new model explicitly accounts for these journeys, a feature made possible by the National pupil database (also used by the Schools Atlas above) which is collated by the Department for Education. It contains details of the pupils at each school including which part of London or surrounding areas they live in. This allows us to model the flow of pupils from their home wards to their schools and then link these flows to the GLA’s small-area population projections. Previous versions of the GLA’s models took a simpler approach of linking rolls to the population of the area surrounding the school.
This new flow model allows for transparency in the model to trace back patterns in the school rolls projections to the population projections in the specific areas that pupils are drawn from. The boroughs now have explicit information on both pupils coming into their schools from other boroughs and private school uptake, and the results better reflect possible impact of new housing development.
Our published Pan-London Demand Projections use the same underlying data and simplified version of this methodology. As part of our commitment to open data practices the results and methodology have been published complete with the code on the London Data Store.
It’s great having the increased transparency and more detailed outputs provided by the new model, but this comes with the challenge of communicating this more complex data to users. In order to aid with analysis of the new projections, we are taking advantage of interactive visualisation tools that R provides. The example shown below uses the R Shiny application to develop an interactive map for exploring the outputs of the school rolls projections and their relationships to population projections.
With this range of new tools we are excited about some even bigger developments that are in the pipeline for school rolls projections. We have recently been approved for access to a dataset on pupils’ order of preference of schools during the schools applications process. This will give us much better insight into the demand for each school, and allow for scenario testing with increased predictive power on the impact of adding, expanding or changing schools.