A0262
Title: Empirical analysis of crude oil dynamics using affine vs. non-affine jump-diffusion models
Authors: Katja Ignatieva - University of New South Wales Sydney (Australia) [presenting]
Abstract: The dynamics of the US oil ETF (USO) are examined through stochastic volatility (SV) models across three classes: SV with jumps in both returns and volatility (SVCJ), SV with jumps in returns only (SVJ), and pure SV without jumps. Eighteen models, including affine and non-affine variations, are evaluated using particle Markov chain Monte Carlo methods. The analysis employs the deviance information criterion (DIC), Bayes factors, probability plots, and deviation measures against the crude oil ETF volatility index (OVX) and realized volatility (RV) benchmarks. Findings highlight SVCJ models, especially SVCJ-PLY-0.5 and SVCJ-PLY-1.0, as most effective in capturing USO dynamics, outperforming standard SV models. The SVCJ-PLY-0.5 model is notably superior based on DIC and Bayes factors, closely aligning estimated volatility with OVX and RV. Statistical criteria favor jump models, with affine models SVJ-LIN-0.5 and SVCJ-LIN-0.5 showing particular promise for finance theory, ranking high among tested frameworks. The predictive accuracy of the evaluated models is underscored for forecasting future volatility trends.
