Volatility Forecasting: A Fused Filter Algorithm

Abstract

ABSTRACT This paper proposes a fused filter algorithm (FFA) for volatility forecasting. The FFA extracts latent volatility by solving a penalized optimization problem, a formulation related to the HP filter and Whittaker–Henderson smoothing. To investigate the empirical performance of the FFA, we apply it to 10 major stock indices and compare its performance, both in‐sample and out‐of‐sample, against different benchmarks. The results show that FFA significantly dominates the benchmark models across different loss functions and horizons, and this dominance mainly concentrates in normal and low‐volatility regimes. A volatility timing strategy is used to evaluate the economic value of the FFA, and the results also confirm its competitiveness in comparison with the benchmark models. In summary, the FFA provides a computationally efficient alternative tool for market volatility filtering and forecasting.