Leila Van Keirsbilck (UCLouvain)

Title: Mapping and sampling impacts on centrality measures ability to explain financial institutions losses in crisis periods.

Abstract: 

Since the 2008 crisis has triggered a huge interest in the financial networks literature, researchers found that node centrality measures computed from a network representation of the financial system before a crisis period could explain partly losses incurred during the crisis. But how do the choices made in order to obtain a specific network representation affect this ability of node centrality measures to explain losses? This article tests whether changes in the network mapping specification and changes in the sampling affect the econometric relation between losses and the systemic vulnerability of institutions determined by a centrality measure. We find that the significance of the classic shortest path betweenness centrality is highly impacted by a decrease in the network size. We also perform a best subset selection exercise showing that alternative betweenness measures, more robust to changes in the network size, perform as good as other classic centralities in explaining losses.

This seminar will take place in Room S08 at the Faculty of Sciences.