Robustness Analysis when Estimating Economic Capital for Credit Risk

Credit risk modeling is an important part of the nancial protection used by banks during times of turbulence in the economy. More precisely, the modelling is about estimating how much economic capital a bank needs to hold in order to survive during an extreme loss. This thesis is about improving the robustness for the estimation of the economic capital, when it is updated as time passes. A decrease of the variations in the estimate of the economic capital would allow the bank to decrease the frequency of which the value of the economic capital is updated. This is the main aim for the thesis, as banks have a hard time explaining these variations based on any logical ground. The model used for estimating the economic capital is based on a multi-factor Merton model. The study will look at how the correlation matrix of the factors in the model can be updated in order to obtain as little variation as possible for the estimate of the economic capital. Four major approaches will be conducted to try and minimize this variation. The approaches use techniques such as weighting of consecutive correlation matrices, bootstrapping and standardization of the data using a multivariate CCC GARCH model