FV-84 | Systemic Risk Dynamics in Euro Area Sovereign Debt Markets

Prof. H. Zimmermann, Dr. K. Ters


Research Objectives
In this project, we will employ a network model based on the methodology of a VAR Variance decomposition in order to understand the dynamics of sovereign risk interconnectedness. Macroprudential regulation views systemic risk as dependent on collective behaviour (endogenous), therefore, a network methodology is essential. We will investigate the magnitude and dynamics within the network before the unconventional policy or regulatory measure (called “events”) and compare it with the dynamics estimated in a short time window after the measure was adapted. We are able to work with very short time windows around an event, as we operate with intraday data for bond markets and CDS markets. Also, having data for credit risk from both, the cash and derivative market, enables us to identify which market was leading in the contagion (systemic risk spillover) within the network. E.g. was the risk transmission taking place through the bond market or the derivative market from one country to another? 

Results/Completed steps
The credit risk data for the bond market (asset swap spreads) have been calculated on an intraday basis (30 minute intervals). The CDS data has been tested for intraday patterns (volatility smiles). All relevant events have been collected including their time stamps for our intraday analysis. The coding of the partial correlation is finalized. However, we have decided to add a network structure based on impulse responses from a Panel Vector Autoregressive model (PVAR). This coding is still in process. We have decided to change from the formerly proposed VAR structure to a PVAR due to its superior performance within our problem setup. PVARs have the same structure as VAR models, in the sense that all variables are assumed to be endogenous but with the difference that a cross-sectional dimension is added to the representation. The PVAR has several advantages over individual country VARs in a time series framework. By analysing a panel of countries, we can more accurately model contagion from one country to another since the panel approach captures country-level heterogeneity. The network representation will be based on the PVAR impulse responses to a one unit shock of each respective sovereign entity. The arrow representation within the network will follow the same logic as for the partial correlations.

Furthermore, we have also estimated certain events already such as the impact of different economic adjustment programs during the euro area sovereign debt crisis. Interestingly, we have found differing dynamics of risk transmissions. The first economic adjustment program for Greece had a much stronger impact in terms of contagion transmission onto other GIIPS countries in contrast to for example the second Greek economic adjustment program (EAP). We believe that this is due to the fact, that the first EAP was not executed under the European Financial Stability Facility. This gives some evidence, that the EFSF was able to decrease contagion risk amongst GIIPS countries.

Due to the problems and delays caused by the corona virus, our estimated completion of the project will be end-summer 2021.

Publications, Presentations, Conferences
As the research paper is not yet finalized, there has not been a submission made yet to an academic journal or conference.