Back to search

Panteion University of Social and Political Sciences

Realized and implied volatility decomposition

Abstract

dc:description

This thesis consists an integrated work on analyzing, decomposing, modelling, and forecasting stock market’s implied and realized volatility, two intrinsic elements for policy agencies, governmental agencies, portfolio managers, market makers, researchers, and academics. By exploiting six diverse disaggregation techniques, namely the empirical mode decomposition (EMD), the embedded empirical mode decomposition (EEMD), the singular spectrum analysis (SSA), the Hilbert vibration decomposition (HVD), the empirical wavelet transform (EWT), and the variational mode decomposition (VMD), tailored exactly on the peculiar nature these measures have, thesis isolates their key characteristics in order to result in optimum modelling and profitable forecasts for those actively engaging in the markets. Modelling and forecasting were conducted via autoregressive processes (AR), the heterogeneous autoregressive model (HAR), the holt winters framework (HW) and the long-short term model (LSTM) of the deep learning environment. Through entire process no look ahead bias was committed, something innovative in decomposition techniques literature and a trading exercise on futures contracts was incorporated in order to highlight the substantial divergence there exists among forecast evaluation criteria. Thus, thesis targets on investigating the profits that can be accumulated by infusing the forecasting and trading practices with the proposed model architecture. Important implications are drawn out of these results that justify the choice to simultaneously test, for first time, six decomposition techniques under the same model specifications and under the actual way markets operate. The trading strategy we followed, generated significant economic profits for markets participants for both implied and realized volatility measures, as proxied by VIX volatility index and high frequency returns of S&P500 futures index, respectively, for at least two out of the six decomposition-based models, highlighting some critical facts of proposed methodological procedure. This thesis, also, was expanded to include another decomposition technique, the continuous wavelet transform (CWT), paired with wavelet coherency, but this time for conducting analysis in the time-frequency domain. The choice was inevitably guided by the need to further exploit the advantages decomposition methods, originating from other scientific fields can offer to the analysis of financial time series. Hence, we included four extra variables, namely the S&P500 stock index, the OVX crude oil volatility index, the high frequency returns of WTI futures index as proxy for the realized volatility of crude oil and the WTI crude oil price in order to analyze them in pairs and spot the strength and the direction of their co-movements. Findings tell for a strong implied-realized volatility relation and of a leading realized volatility against implied at different frequency scales revealing cycling patterns coinciding with the frequency investors reconstruct their positions. Findings also inform of a leading VIX against OVX and the stock market, again following altering time cycles. The most astonishing though finding was the absence of any significant daily and weekly cyclical co-movement for all pairs.

Degree

thesis:*
Grantor dc:publisher
Panteion University of Social and Political Sciences
Year dc:date
2024

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Kafousaki, Eleftheria
  • Καφουσάκη, Ελευθερία

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/56582

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
citation

Kafousaki, Eleftheria; Καφουσάκη, Ελευθερία. Realized and implied volatility decomposition. Panteion University of Social and Political Sciences, 2024. http://hdl.handle.net/10442/hedi/56582