Consistent Backtesting Systemic Risk Measures

Abstract This article offers two novel backtests to evaluate the adequacy of well-known systemic risk measures such as CoVaR, MES, SES, and SRISK. Both the new backtests are robust to estimation risk (i.e., their null distributions remain invariant in the presence of estimation risk). While existing backtest is consistent against divergence from the null hypothesis up to a finite order, the article shows that the new backtests are fully consistent. The real-world implications brought by the new backtests are economically significant as they reveal significantly more cases of inadequate systemic risk modeling among the major financial institutions.