Vector autoregression in cryptocurrency markets: unraveling complex causal networks Article Swipe
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.1093/comnet/cnaf014
· OA: W4412383766
Methodologies to infer financial networks from speculative asset price series generally involve predictive modeling to reveal causal and correlational structures. The required model complexity relates to market efficiency, with highly efficient markets displaying fewer simple relationships. This has led to research on complex nonlinear models for developed markets. However, it’s unclear if simple models can provide meaningful insights into the rapidly developed cryptocurrency market. We show that multivariate linear models can create informative cryptocurrency networks reflecting economic intuition and highlighting high-influence nodes. The resulting network confirms that node degree, a measure of influence, significantly correlates with market capitalization ($ \rho=0.193 $). However, some nodes exhibit influence beyond their market capitalization. Simple linear models reveal inherent complexity in the data, supporting their use to prevent surrogate effects and achieve accurate causal representation. A reductive experiment shows most network structure is contained within a small portion, consistent with the Pareto principle, where a fraction of inputs generates most effects. Our results demonstrate that simple multivariate models provide nontrivial information about cryptocurrency market dynamics, largely dependent on a few key high-influence coins.
