Network uncertainty in financial systems

Abstract

This thesis extends the literature of systemic risk in financial networks induced by interbank loans in two directions.

First, we propose a linear model to estimate the unobservable interbank liabilities between banks in the financial network. A simulation study shows that the proposed model yields an accurate set of the ordinary least squares (OLS) estimators for the interbank liabilities in both complete and ring financial networks. Further, we conduct an empirical analysis of the network induced by the banking sectors in ten central European countries, which also shows that the error in the OLS estimators is small. Moreover, the estimation error in the systemic loss generated from the financial network induced by the estimated interbank liabilities is small as well. All these results suggest that despite the simplicity, the proposed linear model has high practical relevance in assessing the systemic risk of financial networks.

Second, we introduce a financial network model that accounts for uncertainty in the fraction of liabilities owed by a bank to its creditors. We use the difference between the maximum loss and minimum loss incurred in the financial system to measure the uncertainty of systemic loss and provide an explicit representation for the interbank liabilities matrix under which these losses are generated when clearing happens. This measure allows us to study the impact of the intermediation level on the systemic loss uncertainty in the financial network. Our findings indicate the existence of a threshold above which higher financial intermediation reduces both systemic losses and the resulting uncertainty. When the intermediation level falls below this threshold, further increases reduce the minimum and maximum systemic loss, but this comes at the expense of higher uncertainty in the loss outcome.