Replication package for "Can Artificial Intelligence Improve Macroeconomic Forecasting under Climate Uncertainty? A Rigorous Out-of-Sample Evaluation for OECD Countries"

This package reproduces all tables, figures, and robustness checks in the article "Can Artificial Intelligence Improve Macroeconomic Forecasting under Climate Uncertainty? A Rigorous Out-of-Sample Evaluation for OECD Countries." It contains the genuine analysis panel for 38 OECD economies, 1995–2023 (oecd_panel_data_REAL_v2.csv), the data-construction pipeline that builds the climate variables from Copernicus/ERA5 and EM-DAT sources, the main rolling-origin forecasting and panel-robust Diebold–Mariano analysis code, the peer-review robustness scripts, and the four manuscript figures. Macroeconomic data derive from World Bank Open Data and the IEA; country-level temperature anomalies from the Copernicus Climate Change Service (ERA5, via Our World in Data, CC-BY); and the climate-disaster index from the EM-DAT International Disaster Database (CRED). See README.md for full instructions, data provenance, and software requirements.