In this thesis we are concerned with valuing options on traded accounts using both continuous and discrete time models. An option on a traded account is a zero strike call on the balance of a trading account which consists of a position of size $\theta$ in a risky asset (which we refer to as a stock) and the remaining wealth in a risk-free account. The choice of trading positions throughout the life of the option are made by the buyer, subject to constraints specified in the contract at the time of purchase. The specification of these trading constraints gives rise to some of the more well known examples including passport options and vacation options. At maturity, the option buyer is entitled to any positive wealth accumulated in the trading account whilst any losses are covered by the option seller.
First, we examine the problem of valuing these options in continuous time. A review of some existing methods is presented, including a complete derivation of the pricing formula for the passport option and the option on a traded account following the methods proposed by Hyer et al. (1997) and Shreve and Vecer (2000), though we often use different techniques to those authors. We also present an alternative derivation for the value of a passport option using our own methodology which we believe is simpler than those currently available.
Secondly, we consider the valuation problem in a discrete time setting by looking at one specific discrete time model, the binomial tree. This is a new contribution to the literature as binomial models for these options have not been previously examined. Using this approach, the greatest difficulty is the determination of an optimal trading strategy which is required to price this class of option. We show that in general, binomial models and continuous time models do not have the same trading strategy, and in fact that the analytic determination of the trading strategy for an option on a traded account may in fact be impossible to obtain. We then turn to passport options, where we are able to derive an analytic optimal strategy which in this case is identical to that used in the continuous time models, thus the problem of valuing passport options is reduced to the same computational burdens as a binomial valuation without recombining branches.
Lastly, we examine some numerical methods which could be used to value options on traded accounts with binomial models. Our problem is shown to be an NP-hard convex maximisation which we convert into both an l1-norm convex maximisation and an indefinite quadratic program. Whilst we present algorithms which are guaranteed to obtain the optimal solution, they are also known to be inefficient and thus inappropriate for any likely application beyond a few time steps. We conclude by summarising our results and give directions for future research in this area.