Portfolio selection with probabilistic utility

Inspired by statistical physics, we present a probabilistic approach to portfolio selection. Instead of seeking the global extremum of some chosen utility function, we reinterpret the latter as a probability distribution of ‘optimal’ portfolios, and select the portfolio that is given by the mean value with respect to that distribution. Compared to the standard maximization of expected utility, this approach has several attractive features. First, it significantly reduces the excessive sensitivity to external parameters that often plague optimization procedures. Second, it mitigates the commonly observed concentration on too few assets; and third, it provides a natural and consistent way to account for the incompleteness of information and the aversion to uncertainty. Supportive empirical evidence is derived by using artificial data to simulate finite-sample behavior and out-of-sample performance.