Asymmetric autocorrelation in the crude oil market at multiple scales based on a hybrid approach of variational mode decomposition and quantile autoregression

Heterogeneous dependence and memory effects are widely recognized in financial markets, including the crude oil future market. However, few studies have examined the correlation between heterogeneous dependence and memory effects. This association reveals differences in the different memory-trait components, yet the literature is lacking. Our study aims to uncover heterogeneous dependence and memory effects on crude oil future returns and their components at multiple scales and to explain the asymmetry of dependence patterns in the crude oil market through the perspective of irrational investor behavior induced by memory effects. The regressions in this study are based on West Texas Intermediate (WTI) crude oil future prices from 1983 to 2023. We propose a hybrid approach that combines variational mode decomposition (VMD) and quantile autoregression (QAR) to process the return and fluctuation series. Similar to the stock market, we find that the QAR coefficients vary across quantiles. The coefficients are positive for the long-term memory component and negative for the anti-persistent component, indicating the momentum and revert effects. The impacts of extreme lagged returns and negative lagged returns on the distribution of coefficients are evident not only in the return series but also in the two components. Lagged fluctuation and extreme lagged fluctuation accelerate the current fluctuation growth at higher quantiles due to rapid accumulation. Finally, the robustness test confirms that the VMD-QAR method is more resistant to noise and sampling disturbances compared to existing methods. Our study contributes to the analysis of the crude oil market in terms of theoretical and analytical methods in finance.