Financial Risk Aggregation: A Simulative Study - Banks are exposed to many different risk types due to their business activities, such as credit risk, market risk and operational risk. The task of the risk management division is to measure all these risks and to determine the necessary amount of economic capital which is needed as a buffer to absorb unexpected loss associated with each of these risks. In this paper, four approaches are compared with respect to their ability to measure the total banking capital correctly. We find that the traditional approach variance-covariance (N-VaR) significantly underestimates economic capital. The additive approach (Add-VaR) overestimates total risk when risk correlations are less than one. The hybrid method (H-VaR), which combines marginal risks using a formula, is more accurate and tracks the advanced model based on Monte Carlo simulation (MCS) and copula quite well, especially when the risks exhibit very high correlations. The top-down approach based on MCS and Gaussian copula (MCS-copula) is adequate to form a joint distribution from specified marginals in an internally consistent and realistic manner while preserving important properties about the individual risks (asymmetry and fat tails). This comparative study has been realized utilizing simulative data about to credit, market and operational losses. With refer to risk correlations, we have used both simulative values and mean "empirical" values deducted from international accredited studies.
Jel Code: G28, G32, C16, C63