Robust Asset Allocation Under Model Risk

In this chapter, we propose a robust asset allocation methodology, when there is some ambiguity concerning the distribution of asset returns. The investor considers several prior models for the assets distribution and displays an ambiguity aversion against them. We have developed a two-step ambiguity robust methodology that offers the advantage to be more tractable and easier to implement than the various approaches proposed in the literature. This methodology decomposes the ambiguity aversion into a model-specific ambiguity aversion as well as relative ambiguity aversion for each model across the set of different priors. The optimal solutions inferred by each prior are transformed through a generic absolute ambiguity function ψ. Then, the transformed solutions are mixed together through a measure π that reflects the relative ambiguity aversion of the investor for the different priors considered. This methodology is then illustrated through the study of an empirical example on European data.