The role of contagion in financial crises: an uncertainty test on interbank patterns

The main lesson learned from the recent financial crisis is the crucial role of interconnectedness between banks as a factor that can push the effects of bank defaults to extreme levels. One bank in distress can compromise the ability to repay obligations of its creditor banks, thereby inducing a more general crisis that spreads from the banking system towards the real economy. Several empirical and theoretical studies have focused on the role of the interbank market in causing contagion in financial crises. In this regard, one frequent problem encountered in dealing with contagion risk in the banking system is that only data on interbank credits and debts aggregated at bank level are publicly available, whereas the whole matrix of interbank linkages would be needed in order to estimate systemic risk correctly. One common solution is to assume that banks maximise the dispersion of their interbank credits and debts, so that the interbank matrix can be approximated by its maximum entropy. This paper tests the influence of this hypothesis on simulations by verifying if variations in the structure of the interbank matrix lead to significant changes in the magnitude of contagion. In order to do this, an algorithm was developed that generates interbank matrices with higher concentration. Then a Monte Carlo simulation was run by making use of the SYMBOL model (SYstemic Model of Banking Originated Losses) jointly developed by the JRC, DG MARKT, and experts of banking regulation (see De Lisa et al., 2010). We than compared results obtained using the maximum entropy approximated matrix with those obtained from more concentrated matrices. Numerical experiments, performed on samples of banks from four European countries, highlight that concentration in interbank loans does affect results but that, when considering the probability distribution of losses, even significant changes in the interbank matrix do not deeply affect results