| The three essays of this thesis research model selection and estimation issues in financial econometrics. Special attention is given to comparing various approaches used previously in the literature and attempting to compare their out-of-sample performance.; Chapter two explores the value of modelling the deterministic seasonal component of volatility and alternate conditional distributions for the soybean futures market. The simultaneous modelling of deterministic seasonal volatility and conditional heteroscedasticity provides for much more precise predictions of observed options prices. Likewise, for the purposes of volatility forecasting, flexible conditional distribution modelling does not seem to benefit volatility forecasting.; Chapter three compares the performance of continuous-time and discrete-time models of time-varying volatility, specifically stochastic volatility and GARCH processes. These models are again applied to the soybean futures markets. The results conclude that significant tradeoffs must be made in using continuous-time models in option price prediction. The flexibility of GARCH processes when applied to discretely-sampled data allows more accurate and precise prediction and simulation of volatility processes.; Chapter four compares the performance of three alternative models for stochastic discount factor estimation. In doing so, previously overlooked weaknesses of the approaches are revealed. Contrary to previous research, this chapter finds less support for two-factor and nonlinear specifications. Most significantly, in out-of-sample pricing, the non-negativity restriction previously tested in Bansal and Viswanathan (1993) was found to possibly offer some benefit, as economic theory would suggest, but contrary to the findings in Bansal and Viswanathan. |
