Education, Energy Access and Labour Market Development in OPEC Economies: A Panel Econometric and Machine Learning Perspective (1995 2025)

Abstract

This study undertakes a comprehensive panel econometric investigation into the determinants of unemployment across twelve OPEC member states over the period 1995 to 2025, generating a balanced panel of 372 country-year observations. Recognising the distinctive structural characteristics of resource-rich developing economies, the analysis introduces energy access variables  specifically electricity access (ACCE) and clean cooking fuel access (ACCG)  as primary regressors, alongside gross capital formation (GCF), population growth (POPGR), and consumer price inflation (INF). The analytical framework is built around a multi-layered methodological architecture encompassing Pooled OLS with robust standard errors, country fixed-effects within estimation, the Between estimator, and a Mundlak-augmented Hausman test for model selection. Unit root tests reveal that ACCE and ACCG exhibit persistent upward trends consistent with near-unit root behaviour, while GCF, POPGR, and INF are stationary. Panel residual based cointegration tests yield strong evidence of a long-run equilibrium relationship between clean cooking fuel access and unemployment (panel t-statistic = −6.25, well below the −1.96 threshold). The Breusch-Pagan test confirms the presence of heteroskedasticity (LM = 31.36, p < 0.001), necessitating robust inference, while the Durbin-Watson statistic of 0.23 flags positive serial autocorrelation. The Mundlak-Hausman test strongly rejects the null hypothesis of random-effects exogeneity (F = 34.56, p < 0.001), confirming country fixed effects as the preferred estimator. Pooled OLS identifies access to clean cooking fuels as the only statistically significant determinant (β = −0.074, t = −4.55, p < 0.001), implying that each percentage-point improvement in clean cooking access is associated with a reduction in unemployment of 0.074 percentage points, with the economic effect spanning 1.58 percentage points across the interquartile range of ACCG. The Random Forest model achieves a cross-validated R² of 0.659, substantially exceeding OLS (0.032), confirming non-linear dynamics and complex variable interactions. Feature importance analysis consistently ranks ACCG (48.1%) and ACCE (40.3%) as dominant predictors. The study concludes that energy access constitutes a structurally significant labour market determinant in OPEC economies and advocates for the integration of clean energy access expansion into post-oil diversification strategies.