Replication Data and Code for "When Does Climate Policy Uncertainty Improve Green Equity Volatility Forecasts? Evidence from Machine Learning across Horizons and Assets"

This repository contains the replication data and Python code for the study “When Does Climate Policy Uncertainty Improve Green Equity Volatility Forecasts? Evidence from Machine Learning across Horizons and Assets.” The dataset covers daily observations from July 2015 to May 2026 and includes clean-energy ETF returns, market, energy, and macro-financial variables, the U.S. Climate Policy Uncertainty (CPU) index, and constructed volatility measures. The repository provides the data and scripts required to reproduce the main out-of-sample forecasting results, forecast comparison tests, horizon analyses, individual ETF robustness analyses, and SHAP-based model interpretation. The forecasting framework compares benchmark and CPU-augmented specifications using Ridge regression, Random Forest, and XGBoost across 1-, 5-, and 10-day volatility horizons. Individual ETF analyses cover ICLN, TAN, and PBW. A README file and data dictionary are included to document the dataset, variable definitions, computational environment, and replication workflow.