| dc.description.abstract | This thesis investigates the predictive properties of the Markov-switching jump diffusion LIBOR market model. A numerical scheme obtaining forward LIBOR rate forecasts is specified by letting the parameters in a jump diffusion LMM, specified under a marked Poisson point process, be dependent on an underlying continuous time Markov chain and then asserting an Euler scheme. The final objective is to see whether adding a Markov-switch component will significantly affect the predictions regarding counterparty risk and kurtosis. Further, an ordinary log-normal LIBOR market model is implemented in addition to a jump diffusion LMM such that the results can be properly compared. Procedures for calibrating the MSJD to market data is described in detail. Finally, crude forms for backtesting is done for the risk measure potential future exposure, in addition to considerations on kurtosis. | |
