Multiplex Network Financial Models

Multiplex Financial Models

Background

This post will outline the background and first pass at the scope of examining global financial markets (using country data –not involving private corporations) as dynamic multiplex netowrks. I’ll start mostly by taking inspiration and techniques from the Maria del Rio-Chanona et al. paper that can be found at this link. This paper looks at the 2008 global financial crises by modeling the global financial system as a multiplex network (a graph with nodes representing countries, edges encoding cross-country financial assets, and layers as asset types).

The paper seeks to identify “systemacially important countries”, in the hopes of understandign the structure (or thresholds) of critically important nodes (countries)in the global financial system. I hope to understand their methodologies to apply a similar analysis on a different form of multiplex networks. At the outset, my hope is to represent the global financial network – ideally involving a more granular node set incorporating private sector firms and corporations – where the layers correspond to the posture of an entities portfolio according to Hyman Minsky’s income-debt relation classification of hedge, speculative, and Ponzi finance positions. This framing can be found in his 1992 paper “The Financial Instability Hypothesis” published by the Jerome Levy Economics Institute Working Paper No. 74 at this link.

Dataset and beginnings of paper/model

The paper uses “annual data from the IMF Coordinated Portfolio Investment Survey on countries’ cross-border equity and debt holdings”, as well as “quarterly data from the BIS International Banking Statistics on countries’ cross-border bank loans and deposits” (Maria del Rio-Chanona et al. 5). From there they construct a directed and weighted multiplex network consisting of three layers and 131 countries. Before we can begin constructing a similar model for our purposes, we must obtain, import, and clean the data. It’s time to do some plumbing and wrangling.

BIS data

Following the reference at the end of the BIS data quote from above we are taken to a BIS statistics page, a page containing international banking statistics. If you are interested in other BIS datasets, they have a consolidated link found here. From the latter page I downloaded the locational banking statistics and extracted the zip to a local datasets directory. Time to open up the CSV and see if there are any formatting peccadilloes to work through before importing into Rstudio. Scanning through, it seems well formatted and straightforward. Note there are both regions and countries, and multiple different asset types as expected.