The Project "Econometrics of Systemic Risk: Inference, Model Comparison and Network Dependencies" is a project undertaken by the faculty of Business, Economics and Statistics at the University of Vienna and funded by the OeNB.

The project is decomposed into three main tasks. The first task is to develop statistical inference for popular systemic risk measures, like MES, SES, SRISK or Delta CoVaR. We will derive corresponding asymptotic distributions semi-parametric-based inference in order to equip these measures with confidence intervals. This will allow us to make statements of the statistical reliability of systemic risk estimates. The second task is to propose a statistical framework for comparing systemic risk models. We will propose a class of loss functions which enables us to jointly elicit MES and VaR. This allows us to rank MES estimates arising from alternative econometrics models and to pursue rigorous backtesting of competing forecasts. The third task will bring together network-based systemic risk measures with multivariate dynamic models for quantiles. We will propose a unifying approach to model causal as well as simultaneous dependencies in possibly high-dimensional financial networks.

 The overall aim of this research project is to reconsider the current position adopted by academics on how to justify the use of certain systemic risk measures. In contrast to the widely accepted proceeding, where researchers compare realizations of systemic risk measures to capital injections from the Federal Reserve or to the probability that a bank suffers high financial losses, we aim at analyzing the statistical precision of systemic risk measures and induced estimation and forecasting errors more deeply. Indeed, there is convincing evidence that signals produced by systemic risk measures are not necessarily reliable and are affected by statistical imprecision and measurement errors (e.g. Danielsson et al., 2011). This means that the statistical properties of market-based systemic risk measures, like MES or SRISK, are not very well understood. We expect that the outcomes of this research project will significantly contribute to the international literature on financial econometrics and financial stability.