The Multiplex Dependency Structure of Financial Markets

Figure 1

The multiplex nature of dependence among financial assets. The plots report the network analysis of a multiplex whose four layers are Planar Maximally Filtered Graphs (PMFGs) obtained from four classical dependence measures, namely, Pearson, Kendall, Tail, and Partial correlation, computed on rolling time windows of 23 trading days between 1993 and 2015. Each of the four layers provides different information on the dependency structure of a market. Although market events and trends have a somehow similar effect on the average dependence between nodes at the different layers (panel (a)), each layer has a distinct local structure. This is made evident by the plots of the average edge overlap (panel (b)) and of the fraction of edges unique to each layer, which confirm that an edge exists on average on less than two layers, and up to of the edges of a layer are not present on any other layer. Moreover, the same node can have different degrees across the four layers, as indicated by the relatively low values of the pairwise interlayer degree correlation coefficient reported in panel (d) for three pairs of layers over the whole observation interval.

(a)

(b)

(c)

(d)