Extracting the Sovereigns’ CDS Market Hierarchy: A Correlation‐Filtering Approach

Abstract

Since correlation may be interpreted as a measure of the influence across time‐series, it may be conveniently mapped into a distance and into a weighted adjacency matrix. Based on such matrix, network theory has attempted to filter out the noise in correlation matrices by extracting the dominant hierarchy (i.e. the strongest linear‐dependence signals) within time‐series.

The aim of this brief paper is to find the current hierarchy in the sovereigns’ CDS market after the structural shift caused by the failure of Lehman Brothers. Thus, based on two different correlation‐into‐distance mapping techniques and a minimal spanning tree‐based correlation‐filtering methodology on 36 sovereign CDS spread time‐series, the target is to identify which sovereigns are providing the strongest – less noisy – and most informative signals.

The resulting sovereigns’ CDS market hierarchy agrees with prior findings of Gilmore et al. (2010) regarding sovereigns’ bonds market, such as the importance of geographical clustering and the idiosyncratic nature of Japan and United States. Additionally, results (i) confirm that a small set of common factors affect the entire system; (ii) identify the relevance of credit rating clustering; (iii) identify Russia, Turkey and Brazil as regional benchmarks; (iv) suggest that lower medium grade rated sovereigns are the most influential, but also the most prone to contagion; and (v) suggest the existence of a Latin American common factor.”

Keywords: correlation, minimal spanning tree, correlation‐filtering, sovereign, credit default swap

JEL Classification: C32, G10, G11