Sicco Kooiker

Sicco Kooiker

Econometrics/ML PhD candidate at the Vrije Universieit Amsterdam. Research focuses on adaptive machine learning methods with applications in macroeconomics.

Nonparametric Time-Varying Granger Causality

Published in Working Paper, 2025

Author: Sicco Kooiker

Abstract:

Where parametric methods require heavy assumptions about the type of Granger causality, and static methods tend to overreject and lack power in dynamically changing time series environments, nonparametric time-varying Granger causality (NPTVGC) testing overcomes both problems. It is a procedure that integrates an exponentially weighted moving average (EWMA) version of the Diks-Panchenko (DP) test with EWMA local density estimators to assess Granger causality in dynamic environments. Two simulation studies demonstrate the importance of correct hyperparameters and the validity of the test. While the DP test rejects Granger noncausality for S&P 500 returns and volumes in both directions, the NPTVGC testing procedure shows that on a more fine-grained scale, there are only brief periods where Granger noncausality can be rejected. This gives a deeper understanding of the true causal relationship between the returns and volumes.

Presentations:

  • IAAE Conference in Applied Econometrics, Turin, Italy (2025)
  • CFE-CMS Conference in Financial Econometrics and Statistics, London, UK (2024)

Recommended citation: Kooiker, S. (2025). Nonparametric Time-Varying Granger Causality. Working Paper.