Conditional Davis pricing (Q784731)

The article deals with the problem of pricing derivative securities in incomplete models of financial markets. Usually in such models there is no unique price, but it is possible to determine an interval of prices. One of the solutions to this problem is Davis pricing, based on maximizing some utility function. In many classical models this approach allows to calculate a unique price. The authors considers this problem in very general setting and consider an augmented model with random endowment. The main goal of the article is to analyze the following problem of stochastic optimization: \[ \sup_{q\in R}\sup_{\pi \in A} E\left[ U \left( q(\varphi - p) + B +\int_{0}^T \pi_ dS_t \right) \right], \] where \(A\) is the set of possible trading strategies, \(U\) is utility function, the stochastic process \(S\) describes stock prices, \(\phi\) represent the random payoffs of the derivative instrument, which one wants to price, \(B\) is random endowment, \(p\) is the price of the instrument \(\phi\). The authors considers the properties of this optimization problem and the dual problem. The main results of the article concerns the uniqueness of its solution, restriction on possible prices \(p\) and differentially of the value function. The main results of the article are: 1. Characterization of the solution to the optimization problem in terms of the martingale measures in the model (Theorem 3.5). 2. Providing examples with non-unique solutions to the problem in some standards models of market (like Black-Sholes model). 3. Providing condition for the existence of derivatives of the value function and characterizing the derivatives as values of a linear stochastic control problem. 4. Providing the methods for calculating endpoints of interval of possible Davis prices.