Utility maximization with convex constraints and partial information (Q996765)

For a finite time horizon the authors consider a market model consisting of one money market and \(n\) stocks whose return process has a diffusion dynamics and the drift process is not, in general, adapted to the filtration generated by a \(n\)-dimensional Wiener process. They assign the utility \(U(x)\) to the portfolio value \(x\) and suppose that the investor wishes to maximize the expected utility of his terminal wealth obtained following some self-financing trading strategy. The optimal decision depends on the expected dynamics of the drift process but the investor can only observe the prices, i.e. he has only partial information. Using filtering techniques, the model is transformed to a market model with full information and the Brownian motion to an innovation process. Convex constraints are introduced. In the unconstrained case the existence of optimal trading strategies is established by simple martingale representation arguments after a suitable change of the measure. Then a verification result is provided and it is shown how results of optimization under convex constraints can be used directly for a continuous time Markov chain model for the drift. For logarithmic utility the performance of unconstrained and constrained strategies on the market data is compared. A version of the stochastic volatility model is considered and the strategy is computed in the hidden Markov model for the market price of risk for a general utility.