NEET forecasting brief: four quarter forward-looking signals about likely movements in the UK NEET rate

Overview

We commissioned the National Centre for Social Research to develop a short-term forecasting tool that gives four-quarter-ahead signals about the likely direction of the UK NEET rate.

This short report presents a summary of how the tool was built, what it currently forecasts, and how it should be used, focusing on headline results and implications for decision-making.

For a more in-depth details on the data sources, modelling approach, diagnostics, and limitations as well as the series of analysis scripts for the statistical software R that enable the full forecasting process to be repeated in the future, see the technical annexe report.

The research shows that short-term NEET forecasting is achievable and useful for policy. The tool can be updated every quarter, giving decision-makers an early, evidence-based view of likely NEET trends.

Methodology

The tool uses a time-series forecasting model (SARIMAX). It learns from past patterns in NEET rates, including seasonal ones, and adjusts its projections using economic and labour market indicators.

Core data: The model is built on the Labour Force Survey (LFS), the most widely used and officially recognised source of UK NEET statistics. It can be re-run each quarter as new LFS data are released.

Triangulation: To test the robustness of findings and guide which factors to include, models were also built using Understanding Society (USoc) survey data and Longitudinal Educational Outcomes (LEO) administrative data. Factors that improved accuracy consistently across datasets were prioritised.

Factors in the final model: the unemployment rate, GDP, the disability rate, and the gap between USoc and LFS NEET estimates. The last of these helps the model correct for periods when LFS estimates may be less reliable, given the survey’s recent data quality issues.

Development: Exogenous variables were proposed by NatCen and Youth Futures Foundation, and peer-reviewed by two independent experts, before being tested in more than 500 model combinations. Their selection was shaped by consultations held throughout the project, including model design workshops and midpoint calibration meetings.