Issue: 2026/Vol.36/No.2, Pages 85-107
ECONOMETRIC MODELING OF UNEMPLOYMENT RATE IN THE UNITED KINGDOM: FROM CLASSICAL ARIMA TO EXOGENOUS AND MACHINE LEARNING APPROACHES
Van Hai Trieu Tran
, Tinh Minh Tu Tran
Cite as: V. H. T. Tran, T. M. T. Tran. Econometric modeling of unemployment rate in the United Kingdom: From classical ARIMA to exogenous and machine learning approaches. Operations Research and Decisions 2026: 36(2), 85-107. DOI 10.37190/ord214108
Abstract
Unemployment is a key macroeconomic indicator for the labour market’s health. Economic shocks, political changes, and structural shifts in the UK have shaped its dynamics. Using 326 monthly observations from 1997–2024 (UK Office for National Statistics), this study forecasts unemployment via ARIMA(1,1,1), ARIMAX, Random Forest, and XGBoost. Especially, ARIMA works for short-term predictions but misses structural breaks and non-linearities. ARIMAX, with gross value added as an exogenous variable, offers slight gains yet suffers from heteroskedasticity. XGBoost delivers the best performance by capturing nonlinear relationships, but direct interpretability is limited. The structural stability test was inconclusive, constraining regime-switching or rolling forecasts. Future research should address these limitations and integrate SHAP-based interpretability with feature significance analysis to better understand model behaviour and the drivers of unemployment.
Keywords: forecasting, econometrics, ARIMA model, unemployment rate, machine learning, United Kingdom
Received: 24 April 2025 Accepted: 10 November 2025
Published online: 10 November 2025