Mutual Information Based Copulas to Aggregate Banking Risks

This paper develops a methodology to aggregate the market, credit and operational risk and derive the economic capital. While the data is obtained by mapping the profit & loss items of income statement into risk types, the dependence structure between the risks is modeled through the combination of mutual information and copulas, which enables to capture both linear and non-linear dependence. The results show that the non-linear dependence could have influences on the risk measure such as Value-at-Risk (VaR), and that ignoring it leads to risk underestimation.