Statistical Methods in Finance 2025

Composite Entropic Measures for Short-Horizon Portfolio Volatility Forecasting

By:George-Rafael Domenikos
Nanyang Technological University

In portfolio risk assessment, entropy-based metrics offer an alternative lens to traditional variance, but relying solely on Shannon entropy can underrepresent distributional features such as heavy-tailed behaviour. In this talk, I introduce a composite entropic indicator that adaptively combines multiple entropy forms into a single stabilized measure of portfolio uncertainty. The approach constructs a composite entropy accounting for different features of the distributions with weights that are determined by relative divergence scores between observed return distributions and fixed reference models. This yields a dynamic weighting scheme that emphasizes the entropy most consistent with prevailing market conditions. I apply this framework to a range of equity and sector indices, including the S&P 500, NASDAQ, XLE (Energy), U.S. Real Estate, India’s NIFTY, and China’s CHI30, and demonstrate that the composite entropy consistently improves the prediction of top-decile 5-day volatility events across all tested portfolios. The results suggest that blended entropic measures can capture risk-sensitive structure missed by classical variance or single-entropy formulations, offering a flexible tool for short-horizon portfolio diagnostics.