https://doi.org/10.52903/wp2025335
DISAGGREGATING VIX
Stavros Degiannakis
Bank of Greece and Panteion University
Eleftheria Kafousaki
Panteion University
ABSTRACT
The present study highlights the economic profits of markets’ participants, accumulated from the disaggregated forecasts of the stock market’s implied volatility, generated from an ensemble modelling architecture. We incorporate six decomposition techniques, namely, the EMD, the EEMD, the SSA, the HVD, the EWT and the VMD and four different model frameworks that of AR, HAR, HW and LSTM, which are tested against a trading strategy. We diverge from quantifying forecast accuracy solely on statistical loss functions and report the cumulative returns of short or long exposure on roll adjusted VIX futures. The findings show that decomposing a time series into its intrinsic modes prior to modelling and forecasting, can result in generating forecast gains that are translated into improved profits for trading horizons of 1 to 22 days ahead. Important trading implications are drawn from these results.
Keywords: decomposition techniques, implied volatility forecasting, ensemble learning, objective-based evaluation criteria, VIX futures
JEL-classifications: C22, C53, C61, G11, G17
Disclaimer: : The views expressed in this paper are those of the authors and not necessarily those of either the Bank of Greece or the Eurosystem.
Correspondence:
Stavros Degiannakis
Economic Analysis and Research Department
Bank of Greece
El.Venizelos 21, 10250 Athens, Greece
email: sdegiannakis@bankofgreece.gr