Answers are grounded in this paper and cite their sources.
This study tests whether climate policy uncertainty (CPU) improves short-horizon out-of-sample S&P 500 return forecasts and how its value varies by market regime and model. Using daily data from 2000–2024 in a strictly time-ordered design, it combines factor-based predictors, liquidity measures, technical indicators, and CPU in three model classes: a linear ARIMAX benchmark, a nonlinear LSTM, and a Hybrid framework that embeds linear econometric structure in a nonlinear machine-learning model. Forecasts are evaluated with RMSE, MAE, out-of-sample R², Diebold–Mariano tests, and Clark–McCracken comparisons over the full sample and major crises. Short-horizon predictability is weak on average but strongly regime-dependent. ARIMAX performs poorly, while the LSTM delivers the best full-sample performance by capturing nonlinear, time-varying dynamics. During persistent structural instability—the Global Financial Crisis, Eurozone debt crisis, oil price collapse, inflation shock, and Russia–Ukraine war—the Hybrid framework produces the most robust and statistically superior forecasts. CPU adds little in tranquil markets but becomes much more informative in crises with elevated macro-financial and geopolitical uncertainty. The study is among the first to systematically compare linear, nonlinear, and Hybrid models to assess the regime-dependent predictive role of CPU in daily equity returns and focuses on real-time out-of-sample forecasts rather than pricing effects or spillovers. It concludes that CPU is both a financial risk and a conditional forecasting signal whose value depends on market regime and model choice, supporting state-contingent Hybrid forecasting systems for environments with structural instability and policy uncertainty.
No takes yet. Share an insight, caveat, or question.
Wafi et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: