Auri sacra fames : Interest Rates -- Prediction, Jumps and the Market Price of Risk

Sammanfattning: This thesis consists of three essays investigating different aspects of interest rates."Prediction of Future Risk-Neutral Short-Term Interest Rate Densities: Can the Black, Derman and Toy Model Assist?" (Co-authored with David Vestin.) This essay evaluates two different approaches to inferring expectations of future interest rates from asset prices. One is based on bond data and builds on the Black, Derman and Toy model, the other is based on option prices. We compare the outcome with a specified assumed benchmark data generating process. The main conclusion is that the option based model works well, whereas the bond based model has difficulties in capturing aspects of the true distribution."Natura non facit saltum – Or Are Jumps an Inherent Feature in European Interest Rate Markets?" A jump-enhanced diffusion model for the instantaneous interest rate is estimated on the EURIBOR, LIBOR and STIBOR one-week interest rates via the characteristic function and a Fourier transform to recover the density function. This is compared with an estimated non-jump diffusion model. Both continuous-time and discrete-time versions of the model are estimated. For all three interest rate series, likelihood ratio tests favor the jump-enhanced model at a statistically significant level. This result holds for both the continuous-time and the discrete-time versions of the model."Estimating the Market Price of Risk in European Interest Rate Markets Using Spectral GMM." The market price of risk in European interest rate markets, constituted by the market price of diffusion risk and the market price of jump risk, is estimated for a jump diffusion model, using the characteristic function and the Generalized Method of Moments. Utilized data is the EURIBOR twelve-month interest rate during 1999-2003. The results are in line with earlier studies on US interest rate markets. The estimation technique appears promising in its technical simplicity, but entails practical estimation difficulties such as start-value sensitivity and lack of efficiency.