The gold stock nexus: Assessing the causality dynamics based on advanced multiscale approaches

Abstract

This paper relights the discourse surrounding the gold-stock nexus, delving into intrinsic cross-correlations and causalities through a trio of multiscale-based methodologies. Leveraging the adaptive characteristics of wavelet analysis, empirical mode decomposition, and swarm filtering, we construct bivariate bidirectional causality curves. Our analysis scrutinizes the dynamic lead-lag relationships by investigating intrinsic multiscale cross-correlations between gold-S&P 500 and gold-DJIA pairs. This comprehensive examination sheds light on the underlying mechanisms governing the gold and stock markets, especially during the turbulent 2020–2023 period, marked by the influence of the COVID-19 pandemic and the Russian war. The study also extends its purview to the relatively tranquil periods, encompassing over 15 years of daily observations in various sub-samples. Notably, our primary contribution lies in providing a profound analysis of the gold-stock nexus, utilizing advanced multiscale Granger causality estimators. Our findings from time-frequency cross-correlations reveal the presence of unidirectional and bidirectional causality, characterized by varying degrees of strength contingent on scales and pairs. The S&P 500 and DJIA indices emerge as leading indicators, reciprocally influencing gold prices, and vice versa. These initial outcomes are substantiated by results obtained through three Granger causality tests, affirming bidirectional causal relationships and highlighting the significance of heterogeneous interactions among the considered pairs. This study enhances our understanding of gold's role as a haven or hedge, especially during periods of financial turbulence, empowering market participants to formulate strategies founded on credible expectations. The incorporation of multiscale analyses, with their flexible properties, ensures a more comprehensive assessment, unveiling hidden information across time scales.