Dynamic Bayesian networks and risk assessment in the over-the-counter derivatives market in the European Union

Objectives

The first objective of the paper is to critically appraise the possibility of using the European Markets Infrastructure Regulation (EMIR) database in research projects and to discuss the drawbacks of the methodology used in the European Systemic Risk Board (ESRB) Occasional Paper from 11/2016. In addition, the goal is to discuss a new methodology that can be used for risk assessment in the Over-the-counter (OTC) derivatives markets: the dynamic Bayesian networks approach.

Background

Supervisors were forced to make critical policy decisions after the outbreak of the financial crisis in 2008.  Decisions were made regarding exposures of counterparts and the protection sold in the event of default, and were implemented in a context of high uncertainty. Indeed, in 2009 there was still no database about transactions in the European Union (EU) OTC derivatives markets. In 2012, EMIR regulation introduced an obligation to report transactions on the OTC derivatives markets to authorized trade repositories. Since then, data on the transactions have been collected and analyzed by ESRB experts. In September 2016, ESRB published the report "First insights from the new EU-wide dataset OTC derivatives". It analyzed three main derivatives markets —interest rate swaps (IRS), credit default swaps (CDS) and foreign exchange (FX)— using descriptive statistics and static network analysis (based on Depository Trust & Clearing Corporation ( DTCC) data only for 2/11/2015).

Data/Methods

The methodology used by ESRB does not allow us to answer two crucial questions: how the network changes over time, and whether relations in the market are stable (whether the links between nodes are stable). Both questions are important from the perspective of the macroprudential policy in the European Union. In this paper we present a new methodology: the dynamic network approach. We use simulated data and show how this methodology could be used if access to EMIR data was provided.

Results/Expected Results

In this report, the visualization of the network of gross national links between counterparts was obtained based on a subset of the IRS market (on plain-vanilla 6M Euribor interest swaps with an original maturity of 10 years).  Therefore, conclusions are only valid for that segment. We study how the structure of the network is modified when tenor or maturity changes. In addition, we carry out further research on comparability of different derivatives networks.