[ARTICLE] The authors present a new method for forecasting interconnected data with long-term memory. It analyzes multiple data series together, outperforming existing methods like stock volatility forecasting, and suggests these long-term dependencies arise from complex system interactions.

by Guillaume Chevillon (ESSEC Business School), Sébastien Laurent, Luc Bauwens

Two recent contributions have found conditions for large dimensional networks or systems to generate long memory in their individual components. We build on these and provide a multivariate methodology for modeling and forecasting series displaying long range dependence. We model long memory properties within a vector autoregressive system of order 1 and consider Bayesian estimation or ridge regression. For these, we derive a theory-driven parametric setting that informs a prior distribution or a shrinkage target. Our proposal significantly outperforms univariate time series long-memory models when forecasting a daily volatility measure for 250 U.S. company stocks over twelve years. This provides an empirical validation of the theoretical results showing long memory can be sourced to marginalization within a large dimensional system.

[Please read the research paper here]