A unified framework for utility maximization problems: An Orlicz space approach (Q930672)

The authors study the following maximization problem: \[ \sup E[u(x+(H\cdot S)_T]\text{ for some time horizon }0< T\leq\infty. \] Here, \(u\) is a utility function which is finite-valued over \((a,\infty)\) for some \(-\infty\leq a<\infty\), satisfying weak regularity assumptions; \(x> a\) is the initial endowment of the agent. \(S\) is an \(\mathbb{R}^d\)-valued càdlàg semimartingale (in the case \(T=\infty\), for all processes under consideration it is assumed that the limit \(Y_t\to Y_\infty\) \((t\to\infty)\) exists); \({\mathcal H}\) is a class of admissible \(\mathbb{R}^d\)-valued \(S\)-integrable predictable processes representing the allowed trading strategies. Finally, \(H\cdot S\) \((H\in{\mathcal H})\) denotes the stochastic integral, and \((H\cdot S)_T\) is the terminal gain achieved by using \(H\in{\mathcal H}\). Embedding the utility maximization problem in Orlicz spaces enables the authors to formulate the problem in a unified way for both cases \(a\in\mathbb{R}\) and \(a=-\infty\). Using duality methods, the authors prove the existence of solutions for the primal and dual problems. It turns out that a singular component in the pricing functional may also occur in the case of utility functions which are finite on \(\mathbb{R}\). (Utility maximization problems in the context of \(\mathbb{R}^d\)-valued semimartingales have been studied, e.g., by the authors [Finance Stoch. 9, No.~4, 493--517 (2005; Zbl 1088.60041)]; \textit{J. Cvitanić}, \textit{W. Schachermayer} and \textit{H. Wang} [Finance Stoch. 5, No.~2, 259--272 (2001; Zbl 0993.91018)]; \textit{D. Kramkov} and \textit{W. Schachermayer} [Ann. Appl. Probab. 9, No.~3, 904--950 (1999; Zbl 0967.91017)]; \textit{W. Schachermayer} [Ann. Appl. Probab. 11, No.~3, 694--734 (2001; Zbl 1049.91085)]).