Diplom- und Master-Arbeiten (eigene und betreute):
"An Empirical Analysis of Interest Rate Models";
Betreuer/in(nen): W. Scherrer;
Institut für Wirtschaftsmathematik,
Valuing and hedging fixed income securities and derivatives is of crucial importance to institutional investors. To price derivatives and quantify their risk, estimates and forecasts for the interest rate term structure are required. Future interest rate term structures are determined by the evolution of the risk-free short-term interest rate. Commonly, short term interest rates are modeled using diffusion processes. The discussion in this thesis is limited to time-homogenous one-factor models. Different specifications for the drift and the diffusion of the process lead to three classes of models: parametric, semi-parametric and nonparametric models. First, models from each class are presented and a brief introduction to the theoretical background is given. Then, the models are assessed from an econometrics as well as finance point of view. To this end, the models are applied to the GBP 1-month LIBOR. Tests on the estimates and the residuals are carried out to investigate the fit to the data. To validate the pricing quality, forward and swap rates are calculated and compared to observed market prices. The results show that the semi-parametric approach fails to produce real-valued drift and diffusion estimates for the GBP 1-month LIBOR. The econometric analysis for the nonparametric and the parametric approach yield that none of the models is able to capture the volatility clustering inherent in the GBP 1-month LIBOR time series. With respect to the mean squared error, the nonparametric model is preferred over the parametric approaches. The results for the pricing performance are not satisfactory for any model. For the semi-parametric approach no prices can be derived due to the failed parameter estimation. The mean squared pricing error for forward and swap rates based on the nonparametric model is bigger than for the parametric models. Calibrating the parametric models to the term structure improves the accuracy of the prices but the derived rates are not close enough to the market prices to be of any practical importance. We conclude that nonparametric and semi-parametric models do not offer an improvement over the considered parametric models.
Erstellt aus der Publikationsdatenbank der Technischen Universitšt Wien.