In this work we propose the use of a nonparametric procedure to investigate the relationship between the Regulator’s Global Systemically Important Banks (G-SIBs) classification and the equity-based systemic risk measures. The proposed procedure combines several permutation tests to investigate the equality of the multivariate distribution of two groups and assumes only the hypothesis of exchangeability of variables. In our novel approach, the weights used in the combination of tests are obtained using the Particle Swarm Optimization heuristic and quantify the informativeness about the selection. Finally, the p value of the combined test measures the reliability of the result. Empirical results about the selection of G-SIBs show how considering the systematic (\(\beta \)), stress (\(\varDelta \)CoVaR) and connectedness components (in–out connection) of systemic risk cover more than \(70\%\) of weight in all the considered years.

在这项工作中，我们建议使用非参数程序来调查监管机构的全球系统重要性银行（G-SIB）分类与基于权益的系统风险度量之间的关系。拟议的程序结合了几种置换检验来研究两组多元分布的相等性，并且仅假设变量可交换性的假设。在我们的新方法中，使用粒子群优化启发式方法获得测试组合中使用的权重，并量化有关选择的信息。最后，组合测试的p值可衡量结果的可靠性。有关选择G-SIB的经验结果表明，如何考虑系统性（\（\ beta \）），应力（在所有考虑的年份中，系统性风险的\（\ varDelta \）CoVaR）和连通性组件（进出连接）覆盖的重量超过了\（70 \％\）。