Facelifting in utility maximization (Q261918)

In the paper, the term ``facelift'' pertains to the discontinuity of the value function in an optimization problem at the terminal time. Before developing a theory in a general semimartingale framework, it is shown that a facelift, together with all of its repercussions such as nonattainment in the class of countably additive martingale measures, appears in the very simple models and is not a ``cooked-up'' consequence of a pathological choice of model. Then the asymptotic behavior of the value function of the dual utility maximization problem with random endowment is analyzed as the time horizon shrinks to zero. It is established that the limiting value of the value function exists under minimal conditions on the inputs. This limiting value is computed explicitly. In addition to considering the Markovian environment, it is shown that the dual optimizer cannot be found in the set of countably additive martingale measures in a wide class of situations.