A large deviations approach to optimal long term investment (Q1424711)

The author develops an asymptotic continuous-time dynamic version of the outperformance management criterion. Such a problem corresponds to an ergodic objective of beating a given benchmark. It is emphasized that long term horizon problems are expected to be more tractable than finite horizon problems and should provide some good insight for management problems with long, but finite, time horizons. From a mathematical viewpoint, this leads to nonstandard large deviation probability control problem. One risky and one riskless asset are considered and their mean returns are effected linearly by one economic factor with positive correlated noise. A stochastic benchmark typically depends on the economic factor, the risky asset and/or a non traded asset. The asymptotic outperformance management criterion is formulated as a large deviation probability control problem. By using techniques of large deviations the value function and nearly optimal controls for the criterion are derived. This is stated under appropriate conditions on the value function of the associated risk-sensitive control dual problem. The appropriate conditions for the main theorem are stated by a careful study of the behaviour of the value function of the risk sensitive dual problem at the boundary of its domain. The results are illustrated with three examples where closed-form solutions can be derived: the Black-Scholes model, the Platen-Rebolledo model and the Vasicek interest rate model.