a new approach to systemic credit risk Tomohiro Ota Markets, Sectors and Interlinkages Division Bank of England 21 st May 2013
Network analysis at the Bank of England RAMSI (Risk Assessment Model of Systemic Institutions) Simulating major banks profits and balance sheets Top down stress testing: how macroeconomic scenarios affect the UK banking system? Banks are independent in normal times, but interlinked during crisis If a bank goes bankrupt, the model calibrates domino defaults based on the interbank exposure network data
Network analysis at the Bank of England Prudential Regulation Authority s new regulatory dataset (2012) 180 banks (nearly all) exposures to 590 counterparties Broken down by instrument Equity, unsecured loans, repo, IR derivatives, FX derivatives, CDS, Broken down by maturity Exposure limits, collateral amounts, etc
Network analysis at the Bank of England Langfield, Liu and Ota (2013) Mapping the UK interbank system Exposure network Funding network
Does financial network matter? Scott (2012) Interconnectedness and Contagion direct impact of Lehman s collapse on these counterparties was not as problematic or destabilizing as many feared it would be had AIG not been bailed out, direct losses imposed upon its counterparties would not have been a major problem either A classic question: financial contagion, or financial correlation?
Does financial network matter? Scott (2012) Interconnectedness and Contagion A classic question: financial contagion, or financial correlation??
Financial network does matter Scott (2012) argues Lehman s failure was not problematic because the counterparty risk was well diversified and collateralised Actual losses by Lehman s failure was not the heart of the problem. (Subjective) Potential losses matter more which could trigger panics and others Scott (2012) mainly focuses on large counterparties. Langfield, Liu and Ota (2013) shows that small UK banks are significantly exposed to some large banks
Motivations From the cause of crisis to a part of crisis: Network contagion may not be big enough to trigger domino defaults, but could trigger panics etc leading to domino defaults From stress tests to comparative statics: We need to know if the new regulations (Dodd-Frank, BIS s derivative reforms etc) are appropriate compared with their costs From Physics to Economics: Network metrics are very useful measures to summarise complicated networks, but not necessarily backed by economic intuitions
Default contagion and marginal contagion Default Contagion
Default contagion and marginal contagion Default Contagion
Default contagion and marginal contagion PD:+50% +5% +20% +2% +20% +2% +14% +7% Default Contagion Marginal Contagion
Decomposing systemic credit risk Deriving the first order approximation of the calibration Network matrix Probability of Default Wholesale Funding ratio This is an extended Merton model internalising network externality (say, Chained Merton model)
Concluding remarks Sufficiently practical, but appropriately theoretical Can study marginal benefit (in containing systemic risk) of capital injection to a bank Allen and Gale (2000) say that complete network is always better in reducing systemic risk: but the model here tells that star-shape network is more robust if shocks are small Finding economic implications behind some major network metrics Left for future studies: liquidity contagion and behaviour Another important direction: from domestic to international network