We examine how the repo market operates during liquidity stress by applying network analysis to novel transaction-level data of the overnight gilt repo market including the COVID-19 crisis. We find that during this crisis the repo network becomes more connected, with most institutions relying on previously used counterparties. There are however important changes in the repo volumes and spreads during the stress relative to normal times. There is a significant increase in volumes traded with the central counterparties (CCPs) sector. At the same time non-banks, except hedge funds, decrease borrowing and face higher spreads in the bilateral segment. Overall, this evidence reflects a preference for dealers and banks to transact in the centrally cleared rather than the bilateral segment. Our results can inform the policy debate around the behaviour of banks and non-banks in recent liquidity stress and on widening participation in CCPs by non-banks.

One of the most important elements of the post-crisis financial reforms was to establish credible resolution frameworks allowing for creditor bail-in when banks are entering resolution. Despite the obvious benefits of shifting the burden of resolution from taxpayers to bank creditors, bank bail-ins may also give rise to costs as the financial consequences for bank shareholders and creditors being bailed in could endanger their own financial situation and potentially entail systemic implications. Using granular data on the securities cross-holdings among the largest euro area banking groups, we construct a multi-layered network model where each layer represents bail-inable securities of a specific seniority layer of the creditor hierarchy. The model can be a useful tool for resolution authorities to inform the policy discussion on both the composition and level of loss absorbing capacity of banks and the direct contagion risk following a bail-in. We find that due to low levels of securities cross-holdings in the interbank network there is no direct contagion in terms of creditor banks failing as a result of another bank being bailed in. At the same time, we document that a bail-in will have pecuniary consequences on the different liability holders and our framework allows for an exact quantification of those effects. In addition, we show that recapitalisations following a bail-in will change the network structure and hence the embedded contagion risk.