Clustering and Testing Financial Asset Returns Using the Spatial Dynamic Panel Data Model

In this work, we consider a spatio-temporal financial dataset, with \(p=84\) units (a sample of asset returns from the “G7” countries) and \(T=666\) daily observations, from May 2021 to December 2023. Assuming that the p units can be grouped into clusters, we apply a clusterized spatial dynamic panel data model and a multiple testing procedure, to investigate the best partition of clusters for the dataset. We show that, among the three candidate partitions considered in our analysis, the best partition is the one based on the asset’s economic sector.