5. Outlier Robust Specification of Multiplicative Time-Varying Volatility Models
from EEG Investiga Hosted by School of Economics, Management and Political Science
Amado, C. (2025). Outlier Robust Specification of Multiplicative Time-Varying Volatility Models. Computational Economics, 66(5), 4107–4135. https://doi.org/10.1007/s10614-024-10838-4 This article proposes a robust specification procedure for time-varying multiplicative volatility models that is resilient to the presence of outliers. It addresses a key problem in financial time series, namely the difficulty of distinguishing smooth changes in unconditional variance from isolated extreme observations, which often generate spurious non-stationarity and inflated volatility persistence in standard GARCH models. The study builds on the MTV-GARCH framework, which decomposes volatility into a short-run stochastic GARCH component and a smoothly evolving deterministic component. To reduce the influence of additive outliers, the author introduces bounded M-estimators for the GARCH dynamics and develops a robust Lagrange Multiplier test for detecting variance changes. Monte Carlo simulations show that the proposed procedure substantially improves size and power relative to conventional tests under outlier contamination. Applications to daily commodity returns (corn and sugar) illustrate its practical relevance, revealing cases of false variance shifts under standard methods. The approach offers important implications for risk management and portfolio allocation.
Embed this episode
No transcript yet.