Conditional Integration of AI Equities with Traditional and Alternative Assets: Evidence from Wavelet Coherence and Time-Varying Connectedness
Year: 2026
Type: article
Abstract: This study examines whether the equities of leading artificial intelligence (AI) firms were integrated with Bitcoin, gold, West Texas Intermediate (WTI) crude oil and clean energy from January 2018 to July 2026. Using an equally weighted index of NVIDIA, Apple, Alphabet and Microsoft, we combine wavelet coherence, time-varying parameter vector autoregressive (TVP-VAR) connectedness and time-varying Granger causality, complemented by windowed scalogram difference and variance decomposition analyses. Dependence is time-varying and horizon-dependent: significant co-movements concentrate around the COVID-19 crash, the 2022 inflation and monetary tightening shock, and the AI-driven boom. Average total connectedness was low (20.23%) but exceeded 50% in March 2020. Clean energy and the AI index are net transmitters of shocks, whereas Bitcoin, gold and oil are net receivers; however, much of the transmitting role of AI equities reflects their exposure to the broad equity market. Lagged predictability is episodic, and AI equities are almost never predicted by the other assets. The results are robust to ex ante AI exchange-traded funds, alternative model specifications, weekly data and Brent oil. AI equities therefore appear only conditionally integrated with other asset classes, and diversification benefits observed in calm periods should not be assumed to persist during stress episodes.
DOI: 10.3390/jrfm19110907
Authors: Heni Boubaker, Nadia Farjallah
Institution: University of Sousse
Language: English
Subfield: Economics and Econometrics
Domain: Social Sciences
Open Access status: hybrid
Estimated APC paid: $1,689
OpenAlex ID: W7225640804