How reliable are systemic risk measures? Model risk estimates of MES and

<ns3:p>The model risk of two systemic risk measures (SRMs) was quantified for a set of systemically important European banks, using the dispersion of SRM estimates as a proxy. A high model risk was observed, with dispersions of above 65% of the average value, associated with the parametrization error of the Monte Carlo algorithm alone, which has profound implications in the context of systemic risk. Ranking individual banks based on the SRM values was observed to become less dependable due to the high model risk of the SRMs, thus making it difficult for regulators to implement proper policies. Underestimation of the systemic risk of a bank increases the stress within the network, while overestimation of the systemic risk of a bank might lead to undue penalties levied upon the bank. The model risk metric we used additionally allowed us to rank the parameter contributions to the observed model risk.</ns3:p>