Mapping of stock exchanges: contagion from a new perspective

ABSTRACT This paper addresses the contagion problem between stock exchanges and uses self-organizing maps (SOMs) to develop a two-dimensional map based on five market variables from 48 stock exchanges around the world for the 2000–2012 period. The study addresses markets worldwide that have different levels of economic development. Using five market variables, we group stock exchanges into distinctive clusters and analyze their similarities, differences, and dynamics. The technique applied, artificial neural networks (ANNs), is a non-parametric visualization tool to study the dynamics of the markets worldwide following a non-traditional perspective based on the specific formation of groups and their evolution. Each group is defined by the Euclidean distances between markets. There are certain common features, both economic and geographic, within the migrant and static exchanges, which offer new insights. By applying the standard contagion methodology to the resulting groups, we study the spread of the subprime crisis and find evidence of contagion, as well as a high interdependence between markets for the entire period under consideration.