Quantification of Model Rrisk
Abstract: The awareness of model risk has increased due to the increased use of models to valuate financial instruments and their increasing complexity and regulators now require financial institutions to manage it. Despite this, there is still no industry or market standard when it comes to the quantification of model risk. The objective with this project is to find and implement a method that may be used to quantify model risk and evaluate it based on accuracy, efficiency and generalizability. Several approaches to model risk in the literature are explored in this thesis and it is concluded that existing methods are generally not efficient, accurate or generalizable. However, by combining two of the existing methods in the literature and using data on counterparty valuations, another method to quantify model risk can be constructed. This method is implemented and backtested and it is found to be accurate, general and more efficient than alternative approaches. Furthermore, this method may also serve in model validation as a mean to judge the quality of valuations and compare valuation models to each other. One limitation of the method is that if there are few counterparties for a valuation model, say 1 or 2, the method used in this thesis is not suitable.
AT THIS PAGE YOU CAN DOWNLOAD THE WHOLE ESSAY. (follow the link to the next page)