Abstract

This thesis examines the pricing of options under several models with market incompleteness. The theoretical approach relies on the absence of stochastically dominating portfolios containing the underlying asset, the option and the riskless bond. The stochastic dominance approach provides two bounds on the equilibrium pricing of options by risk-averse investors. The two bounds are discounted conditional expectations of the option payoff under two probability measures. This research generalizes the previous stochastic dominance pricing results in discrete time to non-i.i.d. underlying asset return processes and to contingent claims with non-convex payoffs. The new results are then used to examine the stochastic dominance pricing bounds for several discrete and continuous time processes of the underlying asset. The continuous time bounds are obtained by constructing a sequence of discrete approximations that converge weakly to a given continuous time process. The weak convergence property provides the convergence of the two option bounds, which are discounted expectations of the option payoff. In the case of a univariate diffusion process, the two option bounds converge to a common limit. The two bounds converge to distinct limits when the underlying asset follows a jump-diffusion mixture. The non-iid stochastic dominance pricing results are then applied to the pricing of options for a LARCH specification of the underlying asset returns. The two stochastic dominance bounds are obtained both for conditional normal and non-normal returns. The impact of the model estimation error is examined by generating a return sample from a known model and computing the stochastic dominance bounds implied by several estimated models.