Climate conditions, financial uncertainty and market resilience in African equity markets: an event-driven probabilistic deep learning approach

Purpose This paper examines stock-market volatility in Egypt, Morocco, Nigeria, South Africa and Tunisia under geopolitical and financial uncertainty and physical climate conditions. African evidence rarely integrates event-time shock analysis with full-distribution forecasting while distinguishing physical climate signals from climate-related financial risk. The study develops an event-driven probabilistic framework grounded in a Climate–Market Conditioning Framework (CMCF). Design/methodology/approach The study combines cumulative abnormal volatility (CAV) analysis with Mixture Density Network–Long Short-Term Memory (MDN-LSTM) volatility and return-density models. Daily market data from January 2015 to April 2025 are combined with Geopolitical Risk, Financial Stress Index and National Aeronautics and Space Administration Prediction of Worldwide Energy Resources climate indicators. Forecasts use a strict chronological holdout and are compared with Generalised Autoregressive Conditional Heteroskedasticity (GARCH)(1,1)-Normal, GARCH(1,1)-Student-t, Exponential GARCH (EGARCH), Glosten–Jagannathan–Runkle-GARCH (GJR-GARCH), Random Forest and Extreme Gradient Boosting. Distributional performance is additionally benchmarked against probabilistic GARCH using the weighted interval score (WIS). Findings Shock responses are heterogeneous across markets. The four fixed global events provide descriptive economic differences but weak small-sample inference, whereas larger factor-specific event samples yield selected market-dependent effects. GARCH-family models remain stronger for point forecasts and GARCH-Normal also records lower WIS, while MDN-LSTM often produces lower-tail hit rates closer to the nominal 5% level. Distributional and Value-at-Risk diagnostics nevertheless show country-specific calibration limitations, especially in South Africa and Nigeria. Practical implications The findings provide useful insights for investors, regulators, and policymakers by improving volatility monitoring, supporting tail-risk management, and informing macroprudential risk assessment in vulnerable financial markets. The evidence also supports integrating climate-sensitive volatility indicators into scenario analysis and sustainable-investment risk monitoring, while avoiding causal policy claims that are not identified by the forecasting design. Originality/value The paper integrates event-based volatility analysis with probabilistic forecasting in African equity markets and develops a Climate–Market Conditioning Framework. The framework distinguishes physical climate conditions from climate-related financial risk and proposes that financial consequences are conditioned by market structure, liquidity, international integration and adaptive investor behaviour rather than being uniform or mechanically causal.